Acquaintance, Ontology, and Knowledge: Collected Essays in Ontology 9783110327014, 9783110326659

Citation preview

Fred Wilson Acquaintance, Ontology, and Knowledge Collected Essays in Ontology

Philosophische Analyse Philosophical Analysis Herausgegeben von / Edited by Herbert Hochberg • Rafael Hüntelmann • Christian Kanzian Richard Schantz • Erwin Tegtmeier Band 18 / Volume 18

Fred Wilson

Acquaintance, Ontology, and Knowledge Collected Essays in Ontology

ontos verlag Frankfurt I Paris I Ebikon I Lancaster I New Brunswick

Bibliographic information published by Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliographie; detailed bibliographic data is available in the Internet at http://dnb.ddb.de

North and South America by Transaction Books Rutgers University Piscataway, NJ 08854-8042 [email protected]

United Kingdom, Ireland, Iceland, Turkey, Malta, Portugal by Gazelle Books Services Limited White Cross Mills Hightown LANCASTER, LA1 4XS [email protected]

Livraison pour la France et la Belgique: Librairie Philosophique J.Vrin 6, place de la Sorbonne; F-75005 PARIS Tel. +33 (0)1 43 54 03 47; Fax +33 (0)1 43 54 48 18 www.vrin.fr

2007 ontos verlag P.O. Box 15 41, D-63133 Heusenstamm www.ontosverlag.com ISBN 978-3-938793-58-9

2007 No part of this book may be reproduced, stored in retrieval systems or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use of the purchaser of the work

Printed on acid-free paper FSC-certified (Forest Stewardship Council) This hardcover binding meets the International Library standard Printed in Germany by buch bücher dd ag

Dedication to Judy Klein and to the memory of Lenny Klein

Preface “Empiricism”: This is now a chapter-heading term – perhaps it always was. It provides a label for the philosophy that was developed in the seventeenth and eighteenth centuries by the British philosophers, Locke, Berkeley and Hume. It developed in the context of providing an adequate philosophical basis for the new science – the new empirical science – of Kepler, Galileo and Newton. It provided a concept of reason that on the one hand contrasted with the a priori reason of the Aristotelians and the rationalists, and, on the other hand, provided an understanding of the nature of the sorts of inference and judgement that structured the new science.1 The basic framework of this new “empiricism” was that the only entities allowed into one’s metaphysical account of the world, into one’s ontology, were entities that are presented to one in our ordinary experience of the world, either sense experience or inner awareness – entities with which we are, to use the common term, “acquainted” in ordinary experience. This new “empiricism” was not only a metaphysics and epistemology. It also defended an account of the nature of human reason. It located the roots of reason in our ordinary experience of things, that is, observation by means of our senses and by means of our inner awareness. The older concept of reason had argued that there are a priori forms and concepts that provide an ontological structure of objective necessary connections in the world. When reason grasps these a priori forms its grasps with apodictic certainty just how the world hangs together, and in particular, the lawful structure that describes how things in the world change and move. The new empirical reason, in rejecting these a priori entities, rejected all claims to grasp an objective necessary structure in the world. Knowledge of the lawful structure of the world rested solely on our ordinary experience of the world. But that ordinary experience is always limited to but a sample, where our judgements of law aim to describe the population. On the new account of reason, then, our judgments about the lawful structure of the world are always fallible.2 1

Cf. F. Wilson, The Logic and Methodology of Science in Early Modern Thought (Toronto: University of Toronto Press, 1999).

2

Cf. F. Wilson, The Logic and Methodology of Science and Pseudoscience (Toronto: Canadian Scholars Press, 2000).

III But as the new “empiricism” replaced a science or reason that purported to enable the mind to move from the realm of sense to a realm that purports to lie beyond sense, a realm for philosophers of a priori concepts and reason, and of God, it was also replacing a concept of reason that made a comfortable place for religion. In eliminating such objects from the purview of reason, the new empiricism and its revised concept of reason was judged by its critics as a form of scepticism. In its own terms the empiricists made of reason a fallible cognitive capacity, but for its criticism it was worse than fallible, it was sceptical, denying all forms of knowledge that could be deemed humanly important. It was “mere empiricism.”3 It was against empiricism seen in this way that the idealist philosophers of the nineteenth century reacted, defending a version – a strengthened version – of the older form of reason, a reason that was far from sceptical and accommodated itself to needs of a religious worldview. It was not merely a felt need to accommodate religion, however, that motivated the idealists such as T. H. Green and F. H. Bradley, nor merely a felt need to attain knowledge more secure epistemologically than the fallible knowledge of the empiricists. They were after all philosophers, and the basis of their objections was a perceived philosophical or ontological deficiency. Specifically, the empiricism of the seventeenth and eighteenth centuries had inherited from its predecessors a view of relations that made in the end for an inadequate account of the logical structure of the world. The idealists recognized this inadequacy and developed a metaphysics that met this challenge: it provided for structure. But this structure was located in a metaphysics and a concept of reason that was contrary to the empirical reason that had been developed by the defenders of the new science in the seventeenth and eighteenth centuries. It provided an account of reason in which the latter grasped objective necessary connections. These connections provided the mind with apodictic certainty. It also provided the mind with a notion of structure and reason in which all things in the world are connected with all other things. In fact, on the idealist account, reason is holistic, and all differences among things disappear in the end into a meta3

For an attempt to separate the falliblism and the scepticism, see F. Wilson, Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1998).

See also F. Wilson, “Science and Religion: No Irenics Here,” Metaphysica, 7 (2006), pp. 159-170. Also F. Wilson, “William Wordsworth and the Culture of Science,” The Centennial Review, 33 (1989), pp. 322-392.

IV physical unity.4 This may have satisfied the aspirations of the religious for a form of reason more adequate to their view that “mere empirical” reason, but from the point of view of many it was itself a form of scepticism. For it denied the reality of the differences that in our ordinary experience we perceive among things. There was therefore a reaction to idealism in favour of the basic framework of empiricism. But this response was a more adequate empiricism. The idealist philosophers had seen a basic gap in the older empiricism: it could not account for the relational structures among things. If empiricism was to be defended then an account of such structure that at once met the idealist objections and at the same time provided for structures had to be developed. This was done by the logical atomists, Bertrand Russell in particular, and by the logical positivists. It was these philosophers who provided an ontologically adequate account of relational structures within the framework of the older empiricism that took from granted that our knowledge of the world, fallible as it is, is rooted in our ordinary experience of things, in sense and inner consciousness. With their results, empiricism could no longer be taken to be “mere empiricism” – though it remained inhospitable to the religious aspirations that motivated the idealists. However, there were among those who defended logical positivism some who located within their own framework an apparent gap in the account of concept formation. Specifically, on their view, one could not provide an adequate analysis of disposition concepts.5 Others disagreed.6 But the result was that those who perceived an inadequacy rejected the label of ‘logical positivism’ and adopted instead the term ‘logical empiricism’. The 4

See F. Wilson, “The Ultimate Unifying Principle of Coleridge’s Metaphysics of Relations and Our Knowledge of Them,” Ultimate Reality and Meaning, 21 (1999), pp. 243-61; “Ultimate Reality and Meaning in the Art of T. S. Elliot and Piet Mondrian,” Ultimate Reality and Meaning, 23 (2000), pp. 339-376; and Socrates, Lucretius, Camus – Two Philosophical Traditions on Death (Lewiston NY and Queenston ON: Edwin Mellen Press, 2001),. 5

Cf. R. Carnap, “Testability and Meaning,” Philosophy of Science, 3 (1936), pp. 417471, and 4 (1937), pp. 1-40. 6 Especially G. Bergmann; cf. his, “Comments on Professor Hempel’s ‘The Concept of Cogntive Significance’,” Dedaelus, 80 (1951), pp. 78-86; re-printed in his The Metaphysics of Logical Positivism (New York: Longmans Green, 1954), pp. 255-267. .See also F. Wilson, “Dispositions: Defined or Reduced?” Australasian Journal of Philosophy, 47 (1969), pp. 184-204.

V use of the term ‘empiricism’ was taken to denote their commitment to empirical science, the use of ‘logical’ to denote their commitment to Russell’s logic. However, the logic was not simply Russell’s; it was not simply a logic that allowed the empiricism to provide an adequate account of the relational structure of the world. The logic added features to Russell’s that aimed to solve the problem of analyzing dispositions. These features involved what was in fact a return to the holism of the idealists, a new holism which unfortunately did not escape the problems of the idealist’s holism. In fact, it is not clear how much of the older empiricism remained in the philosophy of those who called themselves “logical empiricists.”7 But the latter term continued in use. The upshot is that many who reject important features of the older empiricism, its logical atomism in particular, nonetheless continue to refer to themselves as empiricists. Empiricism is a good thing, atomism bad. The aim of the studies that follow is to as it were re-connect empiricism and atomism. This centrally involves a systematic criticism of the various forms of holistic accounts of meaning that have become common. There are, in the first place, those theories that derive in part from Dewey and his version of empiricism, in particular certain features of the philosophy of Wilfrid Sellars. There are, in the second place, those theories that derive from the theories of language developed by the later Wittgenstein. As for the positive thrust of these essays, that derives, it will be clear to the reader, from Gustav Bergmann, and his students. Herbert Hochberg and Reinhardt Grossmann have both assisted with their comments, from which I have always benefited, even when I am critical of their work. The studies also aim to make clear that the earlier empiricism of Locke, Berkeley and Hume did, as the idealists argued, have a serious problem with the ontology of relations. It is argued that Russell provided the response that enables one to meet the deficiencies of the older empiricism without descending into the holistic response of the idealists. The studies provide a systematic attempt to understand the notion of “acquaintance” on which any adequate empiricism must rest and on which any rejection of holism based on an appeal to acquaintance must therefore also rest. Finally, the studies make the specific point that a defence of empiri7

Cf. F. Wilson, “The Notion of Logical Necessity in Carnap’s Later Philosophy,” in A. Hausman and F. Wilson, Carnap and Goodman: Two Formalists (The Hague: Nijhoff, 1967), pp. 97-225. See also “Implicit Definition Once Again,” in the present volume, below.

VI cism against holism and an adequate account of mind requires acceptance of realism as opposed to nominalism. Already Plato had argued that the properties of things as given in sense experience are as particular as the things of which they are properties. Such a view may be characterized as nominalistic. Nominalism in this sense creates the problem of sameness, the problem of locating what it is about things that leads one to refer to them as being the same. Plato’s solution was to argue that the properties of things share in or imitate Forms. These Forms are entities that lie outside the world of sense, transcending ordinary experience. More recently, Frege has introduced these Forms in a way that attempts to solve certain problems in the philosophy of mind. It is evident that such forms, as entities that transcend the world of sense experience and inner awareness cannot be admitted into any ontology that pretends to be empiricist. Any empiricist oncology that aims to solve the problem of the sameness of presented properties must reject any account that requires the introduction of transcendent entities. But it can do this. All it must do is recognize that the sameness of properties is something that is given in ordinary experience. It must, in other words, reject the nominalism that motivates Platonism, and accept instead a realist account of properties. The first study, “Acquaintance, Ontology and Knowledge,” presents the basic empiricist argument against holism and against Platonic nominalism. The appeal is to the empiricist’s Principle of Acquaintance” (PA), which stipulates that no entity or sort of entity is to admitted into one’s ontology unless it is presented in ordinary experience, either in sense experience or in inner awareness. A lengthy Appendix deals specifically with the thought of Wilfrid Sellars, undertaking to show, to put it briefly, that his “myth of the given” is itself a myth, and that his arguments against an empiricist philosophy and an empiricist philosophy of mind are unsound. The second study, “Hume and Derrida on Language and Meaning,” examines the rejection of Platonism by two philosophers, David Hume and Jacques Derrida. Both appealed to what is in effect the empiricist’s Principle of Acquaintance: Derrida’s argument against Platonism is essentially a simpler version of Hume’s argument against Aristotelianism and rationalism. Derrida’s case would have been the stronger if he had read Hume. Derrida does not progress beyond the negativism and relativism of “deconstruction.” Hume, in contrast, develops a new empiricist account of ration-

VII ality to replace the older account of reason, replacing “rational intuition” of forms or essences as reasons for things with the this-worldly causal reasoning of the then-new science. The third study, “Empiricism: Principles and Problems,” examines the case for empiricism, noting in particular the central place of an adequate account of relations in any empiricist account of the world. The problem of relations was raised by Descartes; Hume replied, but not adequately. The views of James, Russell, Bradley and Bergson are discussed. The issue is how to have structure in an empiricist’s world while not falling into an unacceptable holism. Study four, looks at the work of Berkeley, looking at some recent interpretations, and in particular trying to come to terms with how it is that Berkeley moves from empiricism to idealism. Study five looks at the issue of relations from another angle. “Bradley’s Account of Relations and Its Impact on Empiricism,” examines Bradley’s critique of the earlier empiricism, Hume’s in particular. Bradley argues that this empiricism is inadequate for not being able t account for the existence of structure in the world. He proposes an account of the world which allows for such structure. But the resulting account of relations leads to a holism and a metaphysics in which the only reality in then is the Absolute. The result is a world that has structure in a way but loses it in the Absolute, and which is therefore unacceptable to any adequate empiricist account of the world. The inadequacy pointed out by Bradley of the older empiricism, but the empirical inadequacy of his own position helped move Russell to provide an alternative account of relations, one that is compatible with the empiricist’s Principle of Acquaintance, accounting for structure on the one hand but not implying any sort of holism on the other. This study addresses the problem of relations in the empiricism of Locke, Berkeley and Hume. It argues that Bradley, the idealist defender of the Absolute, was not merely the antithesis of empiricism, as some have claimed. His holism and consequent Absolutism was a reasoned response to a real inadequacy in that empiricism. His case forced those who wished to defend empiricism to address the issue of relations. The study argues that Russell provided this response: it was Bradley’s contribution to empiricism to force on the empiricists this radical re-thinking of relations.

VIII Study six, “Moore’s Refutation of Idealism,” looks at his argument that the idealist principle that esse is percipi is mistaken. The idealist defends his or her principle by appealing to a substance ontology and a holistic account of relations; Moore criticized both. It suggests that Moore’s argument is much misunderstood, with the consequence that its continuing importance goes unrecognized. The examination of the problem of relations in an empiricist philosophy is continued in the seventh study, “Burgersdijck, Coleridge, Bradley, Russell, Bergmann, Hochberg: Six Philosophers on the Ontology of Relations.”The traditional account that was taken over by the classical empiricists can be found in the textbooks of the age; that in Burgersdijck is examined. Coleridge was perhaps the first to point out the inadequacy of this account of relations. The critique was continued by idealists such as Bradley. Bradley’s critique is again examined together with Russell’s response. The continuing debate about the place of relations in an empiricist ontology can be found in the work of Gustav Bergamnn and Herbert Hochberg. The study looks at the claims that these philosophers make, and what these imply for empiricism. Study Eight, “Bareness, as in ‘Bare Particular’: Its Ubiquity,” turns to an issue in ontology that confronts one who accepts that properties are universals and denies that they are “tropes,” that is, properties that are said to be as particular as the things they characterize. The problem is that of individuation: if an individual thing is characterized by properties and these are universals, then what accounts for the difference of two different individuals all of whose properties are the same? One answer is that there is a particular in each which individuates the things. This particular is in itself not a property and has therefore often been said to be bare. But bare particulars have been thought to be like Lockean substances, things that are “I know not what’s,” and which are therefore incompatible with empiricism. This study argues that there are entities which individuate and which can therefore reasonably be called particulars, but that we are in fact acquainted in experience with these entities, so that they are admissible to any empiricist ontology. As for “bareness,” it is argued that this characterizes properties also, and is thus no objection to particulars, no more than it is an objection to properties. Study nine, “Universals, Bare Particulars and Tropes: The Role of a

IX Principle of Acquaintance in Ontology,” pursues some of these themes. It returns to another theme of the first study, namely, the problem of the nature of properties to which empiricists are committed. If one defines nominalism as the view that the properties of things are tropes, that is, are as particular as the things of \which they are properties, then, this study argues, such a nominalism is incompatible with empiricism: the appeal to the Principle of Acquaintance commits the empiricist to the position that one and the same property can exist in several different individuals, that is, commits on to realism, the view that properties of things are universals rather than particulars. In fact, it proposes, that a nominalism of tropes can lead to a nominalism in which properties and things are simply blobs, where all sameness is conventional. However popular such a view might be at present, it is simply unacceptable. But then, what is the relation of properties, understood as universals to the things, particulars, of which they are the properties? The tenth study, “The World and Reality in the Tractatus,” argues that this is one of the issues that is raised in Wittgenstein’s rather puzzling distinction in his Tractatus Logico-Philosophicus, between the world and reality. But there is another issue for empiricism which has so far been avoided, almost studiously. This is problem of the nature of negation. We do have truths to the effect that “x is not F.” On the one hand, they really are true. This seems to commit one to the position that there are in reality not only positive facts, that is, facts represented by true statements to the effect that “x is F,” but also negative facts, those facts that make true statements of the form “x is not F.” But it would seem, on the other hand, that all we find in the world as we ordinarily experience it are positive facts. The difficulty for empiricism is clear: the basic principle, the Principle of Acquaintance, seems to require us to reject negative facts even though we need the latter for any adequate ontology. The present study looks at the problem of negation, suggesting that Wittgenstein’s discussion proposes a solution that is compatible with empiricism. The eleventh study deals with further aspects of the same issue through a critical evaluation of some of the claims made by R. Grossmann in his important The Categorial Structure of the World. Grossmann places his discussion squarely in the context of an appeal to acquaintance, raising problems about what really we can said to be acquainted and therefore what really must appear in an empiricist ontology. Grossmann in his book

X raises issues about negative facts. He also argues that there are general facts over and above individual facts. These, he proposes, are in fact given to us in acquaintance and ought therefore be admitted into an empiricist ontology. This study argues to the contrary that they are not after all given in our ordinary acquaintance with things in the world. The world is not as populated with kinds of facts as Grossmann supposes. Study twelve, “Bergmann’s Hidden Aristotelianism,” deals with another issue raised by Grossmann, whether there are general facts in addition to individual facts. It is argued, against Bergmann (and Russell), that there are no such general facts: there are of course general truths, but for all that no general facts. It is suggested that to hold that there are such facts is to return to a form of anti-Humean Aristotelianism that violates the empiricist’s Principle of Acquaintance. Study thirteen, “Human Action and a Natural Science of Human Being,” discusses in the philosophy of mind and of human action, defending the position that there is a science of human being no different in kind or in method from a science of stones or of planets. This discussion raises issues of the nature of acquaintance, and, related to that, issues of meaning and of language. Some of these themes are taken up in the fourteenth study, on “Marras on Sellars on Thought and Language.” Wilfrid Sellars has been one of the most articulate defenders of a holistic account of meaning, some aspects of which have been treated in the first study. Sellars bases his arguments for holism on an account of thought and language. A. Marras has commented perceptively on these, and the present study looks at the views of both Sellars and Marras from an empiricist perspective. Issues about “intentionality” are raised here, and will be discussed further in succeeding studies. The world contains properties and it contains particulars: Study fifteen discusses one problem this raises, and also takes up issues in the philosophy of mind raised in study fourteen. Properties are one sort of thing. Individuals or particulars are another sort. How does one account for the division of the world into these sorts? The property of being a property or the property of being a particular seem to be odd properties indeed. G. Bergmann once argued that these special

XI sorts of property do in fact exist – they must be admitted into any empiricist ontology – but that they are ineffable – their existence can be shown as it were in language, but cannot be expressed. Study fifteen, “Effability, Ontology and Method,” argues that Bergmann is correct in his contention that these special properties must indeed be admitted into an empiricist ontology but that he is wrong to hold that they are ineffable. The discussion takes us further into the problems of linguistic meaning and of meaning in the sense of “intentionality,” that is, the way in which mental acts mean or intend states of affairs. This is important. If one takes the view that intentionality is real and holds, correctly that thoughts can intend negative no general facts, then on the account of relations defended earlier one would seem after all to be forced to admit negative and general facts into one’s ontology. Bergmann does seem to be committed to this position too. “Effability, Ontology and Method” develops an account of intentionality in which this consequence does not follow. This account is based on considerations of linguistic meaning. Study sixteen, “The Aboutness of Thought,” discusses the nature of the “intentional” relation that holds between mind and its object. It is argued that this “relation” can be understood in terms of linguistic meaning, but that the usual ways of doing this, e.g., those of Sellars and Ttchener are not in themselves successful. Careful attention is paid in particular to the views of Bergmann and of J. Searle, though those of the latter are less fully worked out that the former. The views of L. Addis are also examined. The problem of linguistic meaning has be explored by J. Katz and his reflections lead him to the onclusion that one needs to locate linguistic meanings in a realm of Platonic entities that transcend our ordinary empirical acquaintance with things. The study seventeen, “Language and (Other ? ) Abstract Objects,” critically appraises Katz’s argument as these are developed in his Language and Other Abstract Objects. An appendix looks at the issue whether the work of Chomsky has any relevance for philosophy: it is argued that it does not. This study, following the first and the fourteenth and fifteenth studies, criticizes materialist philosophies of mind and defends behaviourism, that is, methodological behaviourism, as the framework appropriate to a genuine science of psychology. The eighteenth study, “Implicit Definition Once again,” criticizes

XII the doctrine that the meanings of terms is governed by the axioms in which they appear. This view, the doctrine that the axioms “implicitly define” the terms they contain, was criticized early in the last century by Frege, but has more recently been revived by Sellars. This study, criticizes both the original doctrine, following Frege’s argument, and also a more recent version of it proposed by W. V. O. Quine. An appendix examines other points where philosophical positions are thought to receive support from “implicit definitions,” in ontology and the philosophy logic, to mention two of these places: again it is argued that the doctrine is spurious and the claimed support quite absent. The ninteenth study critically examines M. Dummett’s Origins of Analytical Philosophy. Dummett’s work criticizes Frege’s (and Katz’s) Platonistic account of meaning on empiricist grounds, by appeal to a Principle of Acquaintance. But Dummett then goes on to argue that the alternative account of the meaning of language and also of the meaning or intentionality or thought has to be holistic – a position that we argued in the first study to be inconsistent with the appeal to acquaintance. The present study argues that there is an account of meaning, both the meaning of language and the meaning of thought, that on the one hand does not require Frege’s Platonism but on the other hand, contrary to Dummett (and Sellars), also does not require a reversion to holism. It is suggested that one of the grounds that seem to lead to a holistic account of meaning is a commitment to nominalism: reject nominalism and accept a realistic account of properties and one need not accept holism. It is argued that the publicity of meaning, which was Frege’s concern, can be solved by a realistic account of properties.

Our understanding of the world and of human being depends upon our achieving a science of human being in the sense in which there is a science of stones or of oysters. There is no knowledge of ourselves, of the world, and of ourselves in the world beyond such a science. Claims to the contrary are not only mistaken, but are dangerous. There are many errors in philosophy, most of them innocuous. But those that claim to have insight into the nature of human being that is other than that of science are dangerous. They are dangerous because they support what is in fact an illusion, the illusion that there are other forms of

XIII knowledge which are equally good if not better, e.g., the appeal to some special form of knowledge that can be found in religion. Such claims can found views of the world that are clearly ideological, and can support moral claims that do not pass the only legitimate test, that of experience. We are all familiar with such ideologies and such moral claims. Relativism with regard to matters of fact is also dangerous for the same reasons. The following studies aim to support the view that there is a science of human being in exactly the same sense as there is a science of stones or of oysters. They endeavor to show how nominalism can lead to the denial of such claims and can also lead to forms of holistic relativism. It is argued that such nominalism and such a holism can be excluded only by embracing an empiricist’s Principle of Acquaintance and a realism and philosophy of mind founded on this Principle. Perhaps the present collection of essays will help in the defense of this enlightenment image of the world and our place as humans in that world. In any case, it is to be hoped that they will contribute to the furtherance of this ideal and make the world a somewhat better place.

April 2007 Toronto, ON

XV

Acknowledgments The studies in this volume appeared originally in the following:

1. “Acquaintance, Ontology and Knowledge,” The New Scholasticism, 54 (1970), pp.1-48. 2. “Hume and Derrida on Language and Meaning,” Hume Studies, 12 (1986), pp. 99121 3. “Empiricism: Principles and Problems,” in W. Sweet, ed., Approaches to Metaphysics (Dordrecht: Kluwer, 2004), pp. 265-302. 4. “On the Hausmans’ ‘New Approach to Berkeley’s Ideal Reality’,” in R. Muehlmann, ed., Berkeley’s Metaphysics: Structural, Interpretative, and Critical Essays (University Park, PA.: Pennsylvania State University Press, 1995), pp. 67-87. 5. “Bradley’s Impact on Empiricism,” in J. Bradley, ed., Philosophy after F. H. Bradley (Bristol: Thoemmes Press, 1996), pp. 251-82. 6. “Moore’s Refutation of Idealism,” in P. Coates and D. Hutto, eds., Current Issues in Idealism (Bristol: Thoemmes Press, 1996), pp. 23-58. 7. “Burgersdijck, Bradley, Russell, Bergmann: Four Philosophers on the Ontology of Relations,” The Modern Schoolman, 72 (1995), pp. 283-310. 8. “Bareness, as in ‘Bare Particular’: Its Ubiquity,” in H. Hochberg and K. Mulligan, eds., Relations and Predicates (Frankfurt: Ontos Verlag, 2004), pp. 81-112. 9. “The Role of a Principle of Acquaintance in Ontology,” The Modern Schoolman, 47 (l969), pp. 37-56. 10. “The World and Reality in the Tractatus,” The Southern Journal of Philosophy, 5 (1967), pp. 253-260. 11. Critical Review of R. Grossmann, The Categorial Structure of the World, Canadian Journal of Philosophy, 16 (1986), pp. 163-180. 12. Review of R. Tuomela, Human Action and Its Explanation, Dialogue, 21 (1982), pp. 571-577.

XVI 13. “Marras on Sellars on Thought and Language,” Philosophical Studies, 28 (l975), pp. 91-102. 14. “Effability, Ontology and Method,” Philosophy Research Archives, 9 (1983), pp. 419-470. 15. “The Aboutness of Thought,” in T. M. Lennon, ed., Cartesian Views: Papers Presented to Richard A. Watson (Leiden: Brill, 2003), pp. 151-164. 16 Critical Review of J. Katz, Language and Other Abstract Objects, Canadian Journal of Philosophy, 14 (1984), pp. 663-673. 17. “Implicit Definition Once Again,” The Journal of Philosophy, 62, (l965), pp. 364374. 18. Critical Review of Michael Dummett, Origins of Analytical Philosophy, Canadian Journal of Philosophy, 27 (1997), pp. 377-406. Editors and publishers are to be thanked for permission to use this material.

TABLE OF CONTENTS One: Acquaintance, Ontology, and Knowledge I II III IV Appendix One: The Mythology of the Myth of the Given: the Holism of Wilfrid Sellars Appendix Two: Things Seen and the Seeing of Them Appendix Three: How Not to Lose Your Mind Two: Hume and Derrida on Language and Meaning Three: Empiricism: Principles and Problems Four: On the Hausmans’ “New Approach to Berkeley’s Ideal Reality” Five: Bradley’s Account of Relations and Its Impact on Empiricism (I) The Earlier Empiricism (a) Relations (b) The Argument against Necessary Connections (II) Bradley’s Critique of Empiricism (a) Relations (b) Bradley’s Alternative (III) Russell’s Reply to Bradley (a) Russell’s Account of Relations (b) Russell’s Criticisms of Bradley (c) Empiricism and Necessary Connections Once Again (IV) Bradley vs. Russell (V) Conclusion

Six: Moore’s Refutation of Idealism Seven: Burgersdijck, Coleridge, Bradley, Russell, Bergmann, Hochberg: Six Philosophers on the Ontology of Relations 1. Burgersdijck and the Logic of Relations: The Tradition 2. Colerdige on Relations 3. Bradley and Russell against the Tradition 4. Russell's Second Argument against the Tradition 5. Bradley’s Alternative to the Tradition 6. Russell’s Criticism of Bradley 7. Russell’s Account of Relations 8. Bergmann’s New Account of Relations 9. The Russell-Hochberg Emendation Eight: Bareness, as in ‘Bare Particular’: Its Ubiquity I II III IV V VI VII VIII Conclusion Nine: Universals, Bare Particulars and Tropes: The Role of a Principle of Acquaintance in Ontology A. Some Terminology B. Two Philosophical Problems C. How to Answer Philosophical Problems D. PA and the Problem of Sameness E. PA and Nominalism F. PA and the Problem of Difference G. Tropes and Blobs H. Particulars Presented

Ten: The World and Reality in the Tractatus Eleven: Grossmann on the Categorial Structure of the World Twelve: Bergmann’s Hidden Aristotelianism Thirteen: Human Action and a Natural Science of Human Being Fourteen: Marras on Sellars on Thought and Language Fifteen: Effability, Ontology, and Method: Themes from Bergmann’s Ontology Sixteen: The Aboutness of Thought Seventeen: Language and (Other ? ) Abstract Objects Appendix I II III IV V VI Eighteen: Implicit Definition Once Again I II Appendix I II III

IV Nineteen: Dummett’s History: Critical Review of Michael Dummett’s Origins of Analytical Philosophy

One Acquaintance, Ontology and Knowledge* For philosophy, an indispensable distinction is that between philosophical statements and commonsensical statements. Thus, we can contrast “There are only books in that room” with “Only universals exist” and “There are no people in that room” with “There are no material objects.” The statements about books and people (be they true or false) have a clear and unproblematic sense. In contrast, the statements about universals and material objects are problematic: they involve “paradox, absurdity and opacity.”1 On the face of it they seem to be monstrously false, at least if they are taken in any straightforward way; for, there are books as well as universals (whatever the latter may be; certainly they don’t seem to be nice ordinary things like books and oysters and shadows and rainbows and after-images and even electrons), and even if one denies the existence of material objects one seems also at admit in some sense that material objects such as books and chairs and rainbows do as a matter of fact exist. Yet, given the philosophical context in which one finds them, in the contexts of the controversies among various philosophers who defended and attacked these prima facie absurd and paradoxical claims, one wants to say more, one wants to say that they are not simply false, and that they are therefore not to be understood in any straightforward sense. But then that simply raises the question: Just how are they to be understood? And the answer to this question is not at all clear. Ordinarily, in order to locate the meaning of a word, we would look into its context. How is it used? we ask. But the context in which we find words used with characteristically elusive and paradoxical senses is itself equally unclear. Context doesn’t seem to help. It is this problematic, paradoxical, perplexing nature of philosophical statements that sets them apart from the statements of commonsense. Indeed, as an initial characterization of what is commonsensical, one wants to say that it is just that which is philosophically non-problematic. But this negative characterization must be replaced by a positive one. That is the *

Originally in: The New Scholasticism, 44 (1979), pp. 1-48.

2 function of one’s metaphilosophy. But even at this point in our discussion, one thing is clear. In order adequately to deal with the philosophical statements, they must be rendered commonsensical, which is to say that they must be explicated. At least, they much be explicated to the extent to which they do indeed have a commonsense core. For, there may be no recoverable commonsense core, and in any case, one will inevitably lose precisely that which made them problematic and non-commonsensical. At the same time, an adequate metaphilosophy will in its own way capture this problematic aspect also, for it ought to have some account of exactly wherein lies the peculiarity of philosophical statements. One thing is clear about philosophical statements right at the beginning. It is this, that the core of the philosophical tradition has to do with ontology: the most characteristic philosophical statements are to the effect that certain kinds of things, and only those kinds, exist, or that certain other kinds do not exist. The emphasis must be on ‘kinds’ and ‘exist.’ The latter locates the interest as ontological. The former locates the interest as categorial. Ontological statements appear in the first instance as philosophical, not commonsensical. Thus the perplexingness of much of traditional ontology. What is required is that the ontological statements be rendered commonsensical by explicating them. And, to repeat, this requires one to have a metaphilosophy. What this study proposes to do is examine one such metaphilosophy. This metaphilosophy makes essential use of the notion of a Principle of Acquaintance (PA).2 It is this upon which I propose to concentrate. It seems to me that the use of a PA has been much misunderstood. And not only by those who hasten to dismiss its use as a form of primitive empiricism. It is not my primary purpose to defend the metaphilosophy to be discussed – though inevitably part at least of such a defence will emerge. Rather, I wish to examine some examples of the use of PA with a view to determining exactly what role the principle plays in the metaphilosophy that it defines. Indirectly, of course, this by itself will amount to a partial defence at least; for, as it turns out, and as we shall see, many of the objections to the use of a PA in ontology have been based on a misunderstanding of the role of a PA. Our discussion should free the position of at least the objections which are based upon the misunderstandings we are going to try to remove.

3 –I– The metaphilosophy we are to consider begins as we have suggested any metaphilosophy ought to begin, from the distinction between philosophical and commonsensical uses. It holds that ontological statements are, in some very broad sense, factual, and delimits commonsense by means of a PA. A commonsensical statement is one which is grammatically well-formed and whose (non-logical or descriptive) words refer to entities of a kind with which we are acquainted in ordinary experience, either sense experience or inner awareness.3 This latter requirement is one imposed by PA: this principle requires that the words refer to entities which are or could be presented. The ‘could be’ is important. Ontology is categorial. It follows that all that is required is that the categories of entities admitted not violate PA.4 A categorial statement is commonsensical, then, only if it mentions only categories of entities, members of which are presented to one. An ontological statement which mentions entities of a category which violates PA is therefore to be rejected as false. The crucial point about this falsity is that it is, if you wish, radical falsity. A state of affairs which is presented may not be veridical. But the entities are still presented, and are in categories which are presented. Such categories are therefore not excluded by PA. PA excludes only those categories which are never presented, even in nonveridical presentations. This is the main point to be made in the present study. Before continuing with this theme, some other features of the metaphilosophy deserve mention. First, we should note that an adequate metaphilosophy ought not itself to rely on unexplicated or still problematic philosophical notions. Thus, it cannot be a principle of one’s metaphilosophy that there are substances. The aim of a metaphilosophy is to eliminate the problematic features characteristic of traditional ontology. In this context it should be made clear that the metaphilosophy we are examining, one based on PA, does not, by virtue of its appeal to acquaintance or presentation violate this rule. To the contrary, it takes ‘acquainted’ and ‘presented’ in a commonsensical and philosophically nonproblematic sense. Seeing, hearing, etc., are acts of acquaintance or presentation. We see trees, hear trains, and so on. We also more narrowly see colours, hear sounds. This distinction that we have just noted lines up with the realism-phenomenalism controversy. As used here, ‘acquaintance’ and ‘presentation’ are not restricted to the narrow sense, which would have the

4 appearance at least of excluding realism – though it should also be said that for the most part what we shall say will be neutral between the broader and the narrower uses. Instead, the contrast that must be drawn is that between these non-problematic cases of “acquaintance”, on the one hand, and, on the other, such things as Plato’s “reminiscence” – the sort of act through which one is supposed to be acquainted with Forms. But just as the Forms violate PA – or so we shall argue – so also do the supposed acts of “reminiscence”: such acts simply do not exist. One clue as to what has gone wrong lies in the fact that Plato cannot rest content with experience, as he – and ourselves – rest content with experience to, if you wish, discover, for example, acts of seeing: experience does not suffice for reminiscence as it does for seeing or for sensing, an argument needs to be offered in which the conclusion is “reminiscence must exist”. Such an argument will be rejected as somewhere unsound by one who accepts PA as the leading principle of his or her metaphilosophy. The claim therefore that there are acts of reminiscence, in Plato’s sense, is radically false – not simply false in the way in which the perception of Macbeth of a dagger is false – that perception is indeed false, but the dagger is still presented in ordinary commonsensical experience – whereas the non-existence of both the Forms, which are “known” only by reminiscence and reminiscence which is “known” only through argument is radical in that speaking of such things and such ways of knowing has no sense, that is, no sense that can be explicated in terms of a commonsense delimited by PA. Another point about the adequacy of a metaphilosophy is this: to be adequate a metaphilosophy must provide one with what may be called diagnostic tools: it must suggest the source of the distinctive perplexingness of unexplicated philosophical statements and uses. The answer which the present metaphilosophy gives is that the crucial words, especially “exist,” constantly shift from categorial discourse to non-categorial discourse; the resultant blur gives rise to paradox. The paradox is removed by systematically disentangling the categorial from the non-categorial uses. Again, for one’s categories to not violate PA is not sufficient for an adequate ontological inventory, or, what amounts to the same, an adequate categorial scheme. Two further criteria must be satisfied: first, in terms of the categories one must be able to solve or resolve the various explicated ontological problems; and, second, one must be to, as one says, “account for commonsense.” Note, in the first criterion, it is explicated ontological problems which are mentioned. Such a problem, since it is by hypothesis explicated, will not be philosophically problematic in a way that an unex-

5 plicated problem is: it will not involve “paradox, absurdity, and opacity.” Indeed, it may turn out to be utterly trivial Thus, one might claim that once the problem of sameness, the problem which gave rise to the whole discussion of universals from Plato on, is properly explicated, it is resolved by pointing out that two red objects have the same (shade of) colour, and that therefore the colour is, in the appropriate sense, a universal.5 But, of course, though such an answer makes a point that is in a sense trivial, that does not mean that it is also uninteresting.6 Nor does it do anything of a diagnostic nature, explaining how something so obvious could come to involve “paradox, absurdity, and opacity,” how, in other words, it could come to be philosophically problematic.7 Note also, in the second criterion, that the phrase “account for commonsense” is misleading. It suggests that there is something about ‘commonsense’ which is problematic and therefore needs ‘accounting for.’ But within the context of the metaphilosophy that we are considering, this is not so. What is commonsensical is precisely what is not problematic, i.e., philosophically problematic. It is the philosophical uses that need accounting for, not the commonsensical ones. Unpacking the phrase, the second criterion states, on the one hand, that all entities must fall into one category or another, and, on the other hand, that the structure of the categories must reflect the ways in which the entities in the categories are connected to one another. Both aspects of this second criterion are necessary. A system of categories may consist of categories all of which are compatible with PA, while still being inadequate in the sense that not all entities are categorized. One is reminded here of materialism, which claims that there are no mental things, no mental acts. (And of course there are mental acts – mental acts as well as the blood on Macbeth’s knife and the chair upon which I am sitting – mental acts such as seeing and believing and supposing and approving and of course including mental acts of acquaintance – they are there just because I am acquainted with them. To be sure, there are some philosophers who deny the existence of mental acts, including mental acts of acquaintance.8 These philosophers naturally reject any metaphilosophy based on PA.9) Or, a system may fail to “account for commonsense” because it may assert entities are connected in ways in which they are in fact not connected. Consider what some call a relational view of mental acts.10 We have the category of mental things and that of non-mental things. A member of the former is presented with one of the latter if it is related to it in the manner distinctive pf presentation, where the relation in question is in

6 the same category as the relations, such as to the left of, which relate nonmental things to each other. Since such a (“descriptive”) relation obtains if and only if both the relata exist, it follows that the relation of presentation will obtain if and only if what the mental thing is presented with exists. It becomes categorially impossible that there be a presentation the object of which does not exist. Within this categorial scheme error – that is, a presentation as existing of an object that does not exist – turns out to be impossible: one can draw no distinction between appearance and reality, commonsense has not been accounted for. That we make mistakes and that we have false beliefs is part of commonsense; by itself this is not philosophically problematic. But it can become such a problem within the context of an inadequate categorial inventory. Indeed, to suppose mental acts are relational does so transform it: for it makes error impossible. Nor can the metaphilosophy we are discussing give any answer but one as to why such a transformation occurred – that is, why it turned out that something ordinary and commonsensical, such as mental acts and the fact of error, became something philosophically problematic. The reason this has occurred is clear: PA has been violated. And, not surprisingly, this turns out to be correct: an inspection of the mental acts of presentation reveals no such descriptive relation.11 Is this way of handling the situation adequate? Such questions provide the test of a metaphilosophy. A metaphilosophy purports to provide a method for the removal of philosophical perplexity. If the answer to the question just asked is in the negative, then the method has been unsuccessful, and the metaphilosophy is inadequate. The test of a method is success.12 However, the issue of the adequacy of the metaphilosophy is not one that I wish here to pursue. The aim is not to defend the metaphilosophy directly, only indirectly, by delineating the role within it of PA. So let me return to the main point I wished to make. I propose to begin with two illustrations. On the basis of these I shall try to draw the moral I believe to be implied. The first example involves two categories which are permitted by PA. Consider a red patch in one’s visual field. This red patch, and any entity of the same kind I shall call a thing. Things thus form an ontological category, and, moreover, a category permitted by PA, since it is perfectly clear that things are presented. Now, there is about this particular thing that we are considering a feature by virtue of which we justifiably predicate ‘red’ of the patch, and any entity of the same kind, that is, any feature of things by virtue of which predications are made of presented things, I shall

7 call a property. Just as things are presented, so are properties of things. That means that properties form an ontological category permitted by PA. The two categories of thing and property are commonsensical: both are permitted by PA. Note that nothing that has just been said depends upon one taking the presentation to be veridical. Even if it is false that some presented thing has a property it is presented as having, it is still presented as having that property, and by virtue of its being presented as having that property we justifiably predicate as we do of it. Thus, for example, suppose that our patch happens to be in fact green. It is still presented as red. By virtue of this feature it is presented as having we justifiably predicate ‘red’ of it. To be sure, ‘red’ is in fact false of it, and ‘green’, let us say, in fact true of it. But this does not affect whether the categories we have introduced using the red patch as an example are not permitted by PA: they are permitted by PA simply because they are presented. (Let me say here that I am well aware of all the pitfalls which reside around my use of ‘justifiably predicate’ in the last few sentences. Thus, we may justifiably predicate ‘green’ of an object presented as red provided that we know that it is in fact green Yet I think the basic point is clear enough that it is not necessary to explore and clarify the qualifications that can and, in certain contexts, must be made, but which can, in the present context, safely be ignored. But we shall come back to these issues in Appendix One, below.) The second example, to which I now turn, provides the required contrast. It will be best to begin with two formulae. Here is the first: Things and properties are given in sensory presentation, or, more briefly, are given in sense. Here is the second: Whatever is given in sense is particular. Consider now two different things, such that ‘red’ can be predicated of each. The property of the one thing, by virtue of which ‘red’ can be predicated of it, is, according to the second of the two formulae, is given in sense and is therefore particular. This means that it is different from the property of the other thing by virtue of which ‘red’ can be predicated of it. This raises the question: By virtue of what about the properties makes it to be such that they both justify the predicating of ‘red’ of the two things? This, of course, is the problem of sameness; and it is categorial, since the issue can be raised with respect to any members of the category of property. One answer to the problem of sameness is the Platonic. According to this answer, the property of the one thing and the property of the other both justify the predication of ‘red’ of each of the two things because (1) ‘red’

8 names a transcendent form, the Red Itself, and (2) the two properties both participate in the Red Itself. Platonists argue that, if the transcendent forms did not exist, then the obvious truth, that ‘red’ can justifiably be predicated of the two things, would be false. PA immediately excludes the Platonic Forms.13 The Platonist’s answer to the problem of sameness remains irremediably to be rejected by virtue of the fact that one is not presented with the transcendent entities it postulates. One is not acquainted with them, whether the presentation be veridical or non-veridical. It is false that there are Platonic Forms, not merely in the sense that, though I am presented with a red thing, that thing really is green, but in the radical sense that they are never – not ever – presented in ordinary experience, even falsely. Here we have the desired contrast between the category of property and that of Platonic Form. One who accepts PA thus infers that, since the conclusion of the Platonist’s arguments is false, radically false, these arguments must be unsound: they must have fallen into error. So the point where the error has appeared must then be uncovered; the required diagnostic remarks must be made. Nor, of course, does appeal to PA show one where this unsoundness lies. In the case of the Platonist, the direction in which one must at this point move is fairly clear. The problem of sameness takes off for the Platonist after the enunciation of the two formulae we noted above.14 Now, that things and properties are given in sense is non-controversial, commonsensical. Thus, the first of the two formulae is non-problematic. But the second of the two formulae, that everything given in sense is particular, is not commonsensical, it is philosophical, and therefore one that requires explication. As it stands, it states that whatever is given in sense is particular. The rejection of Platonist has led some15 to challenge this premise, before the problem of sameness as understood by the Platonist can even get off the ground. In that way, the need for transcendent Forms will vanish. These people have held (a) the same property can be in several things; (b) things are particular; and (c) properties are always presented as being in things. (a) automatically leads to the disappearance of the Platonist’s problem of sameness. (b) and (c) explicate the commonsense core of the formula in question. Others have accepted the Platonist’s premises but have denied that the conclusion follows.16 Rather, the problem of sameness is resolved when it is realized that the property in the one thing and the property in the other both justify the predication of ‘red’ of the two things because the two properties stand in the internal relation of exact similarity.17

9 This move has, however, been challenged, by Moore for one, on the basis of an appeal to PA.18 But these are not the lines I am supposed here to be pursuing. Return to the radically false claim that there are Platonic Forms. Since these postulated entities are, on the one hand, not presented, not even in non-veridical presentations, and are, on the other hand, supposed to be those entities by virtue of which we are justified in predicating one term of two different things, an immediate consequence is a radical scepticism: we can never know if we are justified in predicating a term of a thing. It is not merely that we are presented with a red thing and wonder whether it is in fact red, whether ‘green’ might not be true of it after all. More drastically, we can never identify the property we are presented with as red. Knowing that something is red is excluded by the fact of the transcendent nature of the category of entities which were introduced precisely in order to account for ‘red’ being justifiably predicable of the thing.19 Parasitic upon the introduction of entities which are not permitted by PA is a radical, philosophical scepticism. It is not a matter of doubting whether we know this or that particular fact; it is not a matter of this sort of commonsensical scepticism; but rather it is a radical scepticism in which it comes to be doubted whether anything at all is known or even knowable. Conversely, adhering to a PA removes this sort of radical or philosophical scepticism. It does not follow that a PA, considered as a principle for ontology, provides for the removal of commonsense scepticism. For, the use of a PA in ontology did not require that the presentations be veridical. Thus, a PA, considered as a principle for ontology, provides no foundation for empirical knowledge. This is the moral I wished to draw.20 There has often been a PA to which appeal is made to provide a foundation for empirical knowledge. Beliefs or statements about that with which one is acquainted or, perhaps, directly acquainted (the difference makes no difference here) are held to be intrinsically veridical, or, as one has come to speak, “incorrigible.”21 About such facts as one knows by authority or by inductive inference or, indeed, by any sort of inference, one can raise doubts of a commonsensical sort. But, by virtue of the intrinsic nature of this incorrigible sort of awareness, this sort of commonsensical scepticism cannot be raised with respect to statements or beliefs about that with which one is acquainted. Such beliefs or statements are held, for that reason, to provide a foundation for empirical knowledge. The possibility of raising doubts of any ordinary sort is, of course, commonsensical. There is nothing philosophically problematic, that is,

10 nothing which involves “paradox, absurdity, and opacity,” about doubt or the possibility of error, or error itself. Error therefore constitutes no philosophical problem, and, from the standpoint of the metaphilosophy we are considering, is not to be denied, or even overcome, but only “accounted for.” Now, we do of course correct errors we make. This possibility must be “accounted for” in one’s ontological inventory: it too is part of commonsense. Those who defend the use of a PA as providing a foundation for empirical knowledge argue that such correction of error is possible only because there is in fact an “incorrigible” foundation for empirical knowledge, and that this is provided by PA. The point is that the doubt, or scepticism which such an appeal is designed to remove is commonsensical; it is not philosophically problematic in the way in which the radical scepticism parasitic upon the introduction of transcendent entities is philosophically problematic. A PA considered as a principle for ontology removes scepticism. A PA used to provide a foundation for empirical knowledge also supposedly removes scepticism. Only, the scepticisms are of two different sorts: the former is radical, and parasitical upon the introduction of a category of transcendent entities; the latter has nothing philosophically problematic about it at all. But if one were not careful to make the relevant distinctions, then if one were to attack the idea of a PA as providing a foundation for empirical knowledge, one might be led to also believe that one was showing that a PA had no role to play as a principle of ontology. And this is what I want to suggest is going on in recent philosophy. We have attacks on the “given.” These are arguments to establish that there are no “incorrigible” beliefs or statements. It is concluded that therefore a PA has no role to play in philosophy. This latter conclusion simply does not follow, and arises only because of the mentioned confusions. The idea of a foundation for empirical knowledge has been defended by Ayer.22 Recent attacks on the “given” have on the whole taken his work as a point of departure.23 That is unfortunate, because the attack on Ayer has indiscriminately spread over onto the work of Moore and Russell. These philosophers also appealed on occasion to what amounts to a PA. Sometimes, especially in Russell, the appeal was of Ayer’s sort.24 At other times, however, it was not; on these occasions it was explicitly ontological. And these latter uses of PA have not been touched by the arguments against the “given”. The trouble was that neither Moore nor Russell was as clear on these matters as one might like. They much too readily blurred

11 that distinction between a radical scepticism and a commonsensical scepticism, and the use of a PA to remove the former and the use of it to remove the latter. Nor was Berkeley, who was the first to more or less explicitly appeal to a PA as a principle of ontology, much clearer on these matters. What I propose to do in the rest of this study is to try to disentangle some of the threads that the mentioned philosophers have tangled together. On the one hand, this will provide illustrative support for the claims I have made about the role of a PA in ontology. On the other hand, it will render impotent those attacks on the use of a PA in ontology which depend for their force upon an attack on the “given.”25

–II– We apply predicates to things. We do so on the basis of the properties the things are presented as having. Let us consider properties only insofar as they are, as one says, qualitatively different, so that different predicates are applied to things on the basis of qualitatively different properties, or, synonymously, so that different predicates refer to qualitatively different properties, and so that the same predicate refers to properties which are qualitatively the same. Restricting ourselves in this way simply brushes aside the issue whether in the case of properties numerical and qualitative difference coincide. Some thing being red and the same thing being not green are two different states of affairs. The philosophy of logical atomism, as defended by Moore and Russell, claims to be commonsensical, where commonsense is delineated by means of a PA. Whether they were always consistent on this point, that is, whether they ever suffered from a transcendental lapse, is neither here nor there.26 So long as we restrict ourselves to entities of the categories of thing and property we shall be on safe ground. For, as we saw, these categories are in fact permitted by PA. The point in which we are now interested is that the philosophy of logical atomism made a rather specific claim about states of affairs, of the sort just indicated, a claim which was defended by Moore and Russell against the holists whom they saw, correctly, as their opponents. The claim is this, that states of affairs are logically and ontologically independent. This means, in the case of our example, that the predication of ‘red’ of the thing need not involve the denial of ‘green’ of that same thing. Nor is the point just made intended to be

12 a merely syntactical comment, that one has not yet decided to lay down a set of rules for moving from one statement to the negation of another. Rather, the states of affairs are themselves logically independent. In other words, the claim is ontological in nature, about the entities in the category of property, that is, the entities on the basis of which we apply predicates to things: it is the claim that properties are logically and therefore ontologically self-contained.. Now, we justifiably predicate ‘red’ of a thing when we identify the property which that thing is presented as having as the property to which ‘red’ is used to refer. Since the predication of ‘red’ is not to involve the predication or the denial of ‘green’, and since this is to be interpreted as a thesis about the properties on the basis of which the predication is made, it follows that the identification of a property as the referent of ‘red’ will not involve the referent of ‘green’. In other words, it is not a condition necessary on the basis of the nature or natures of the entities involved, that, in order to know what property a thing is presented as having, one need know or be acquainted with any other property. Or, to put it succinctly, properties are logically and ontologically self-contained and, as such, are wholly presented, in themselves: they are what they are and not anything else, not even partially anything else. This is the claim of logical atomism. If this claim is true, then it follows directly that in order to know that some thing is red it is not necessary to know that it is not, say, green. In other words, the statement (1) (x)(x is red e ~ x is green) is not a statement which is true by virtue of the meaning of the word ‘red’: it is not true ex vi terminorum, at least in any sense of ‘meaning’ which might be relevant to ontological problems. At least, so the logical atomists maintain. Holists, of course, disagree. The contrast that the atomists make is to cases which involve explicit definitions. Suppose we have the definition (2) ‘Fx’ is short for ‘fx & gx’ where, let us suppose, ‘f’ and ‘g’ are interpreted undefined predicates. Then (3) (x)[Fx = (fx & gx)] is “true by definition.” Suppose it is a true generality that (4) (x)[fx = gx] Since (5) (x)[fx = (fx = gx)] follows analytically from (4), so does

13 (6) (x)(fx = Fx) And because (4) is true, so is (6). But (5) is synthetic, and, since (6) is, by (2), simply short for (5), it too is synthetic. Consider, now, the following states of affairs about some thing a: (7) fa (8) ga (9) Fa The state of affairs (7) is connected to both the states of affairs (8) and (9). It is connected to (8) because, first, (4) and (7) entail (8), and, second, (4) is (we are supposing) true and is connected to (9) for similar reasons involving (6). But the connection in each case is external since the generalities (4) and (6) are synthetic, and not true ex vi terminorum, or, what amounts to the same thing, the connection is external since the state of affairs (7) by itself entails neither (8) nor (9). Contrast this with the connection between the state of affairs (9) and (8). (9) is connected to (8) because, first, (3) and (9) entail (9), and, second, (3) is true, indeed true by definition. But this last point gives the contrast. For, here the connection between the states of affairs is internal: (3) being true by definition amounts to (9) by itself entailing (6). Ex vi terminorum, the state of affairs (9) logically implies the state of affairs (8). Those who disagree with the ontological claim of the logical atomists and hold that the state of affairs (10) This is red logically implies or somehow objectively necessitates the state of affairs (11) This is not green must hold that (1) gives the meaning of the terms ‘red’ and ‘green’. However, (1) is not true by definition, ex vi terminorum, since there simply is no linguistic convention of the sort (2) which permits the elimination of one of the predicates of (1), as ‘F’ was eliminable from (3). The incompatibility of red and green is not a matter linguistic convention, or at least, not merely a matter of linguistic convention, as it is merely a matter of linguistic convention that bachelors are unmarried males. If, therefore, it is held that (1) gives part of the meaning ‘red,’ the meaning must be real meaning, as opposed to merely linguistic meaning, that is, something which has to do not merely with ‘red’ but with the referent of ‘red’.27 Holists thus hold that predicates have not only a linguistic meaning but also a real meaning, the latter conferred upon them by the intrinsic nature of the properties to which they are conventionally used to refer. What is involved in real meaning cannot, however, be recognized simply by con-

14 templating the internal connection between red and green. For, red is incompatible with any other colour, and so there is an internal connection between red and each of these. And also, of course, between red and all the various pitches, tastes, etc. However, red is a colour, so the latter is part of the real meaning of red, and, since whatever is coloured has a shape, having a shape will be involved in the real meaning of red, and, finally, since being round is a shape, this property, and also every other shape, will be involved in the real meaning of the property red. In short, the net of real meanings extends itself to the point where every property will be internally related to every other property.28 This aspect of the holistic doctrine of real meaning has the following consequence. Suppose some thing is presented as having the property red. Simply noticing the presented property will not be sufficient to justify predicating ‘red’ of the thing. To be sure, it will in a way be a necessary condition but it will not by itself be sufficient. Since being not green is part of the real meaning of being red the thing being not green will be another logically ( = ontologically) necessary condition for the thing being red. One will have therefore to notice not only whether the presented thing is red but also whether it is not green. But it does not stop here, since the net of meanings connects every property, and, in our case, connects red specifically to every other property. A condition logically sufficient to justify the predication of ‘red’ of the presented thing will be at hand only after all the various real meaning connections will have been noticed. One can know that a thing has one property only if one knows all the other properties it has, or, more accurately, all the other properties it must have and all the other properties that it cannot have. One can know one fact about a thing only if one knows every fact about it.29 Note that this is not a matter of verifying whether it is in fact the case that the thing does have the property in question. Rather, it is a question of identifying the property it is presented as having. Before one can verify the fact it is necessary to know or identify the property it is presented as having, and the holist’s position with respect to properties is simply that before such an identification is made, let alone a verification, the whole network of real meanings must in its details be present before consciousness.30 What follows from the holist’s position is radical scepticism. The holists make it impossible to understand how we know any fact at all. When we doubt that the thing is red, we doubt not merely that the thing is red, that is, that the thing is red as presented but that is it red, that is, red in all its connections to other things and properties, its connections to all other

15 things and properties. With respect to any presented state of affairs it is possible to raise the question – without being able to answer it – whether the thing has any specific property or others. For, it will be impossible to answer this with respect to any specific property until one can answer it with respect to all properties – which, of course, one cannot do. Unless, of course, one is the Absolute. Unfortunately, one is not the Absolute, no one of us mere mortals, and for us scepticism – a radical scepticism – is raised with respect to the properties things are presented as having. This in turn makes it impossible to verify those facts.31 This radical philosophical scepticism is a safe sign that a transcendental lapse has occurred. I shall suggest that an appeal to PA may be used to exorcise this radical scepticism. It does not follow that PA is providing a foundation for empirical knowledge. The appeal to PA that will remove the radical scepticism will not require that our acquaintance with states of affairs be incorrigible. The appeal to PA is this: red is wholly presented, that is, by virtue of itself the presented property is what it is, and it is not what it is by virtue of its relations to other properties, whether they be presented or not. A necessary condition for one to justifiably predicate ‘red’ of a thing is that one has identified the property to which ‘red’ is used to refer. On the basis of what is such an identification made? Simply on the basis of what one is in fact presented with, and not on the basis of any relations in which these entities might stand to others. We identify the property as red solely on the basis of the property itself. We know what property it is without having to know how it is related to other properties. Being acquainted with the presented property is by itself logically sufficient for the identification. And here, ‘logical’ has the force of ‘ontological.’ In short, the property is what it is by virtue of itself and not by virtue of its relations to other properties.32 Hence, the suggestion that entities in the category of property are connected to each other by real meanings is one that violates PA. So much the worse for real meanings and so much the worse for the philosophical scepticism parasitic upon real meanings.33 If this seems a rather cavalier way of dealing with holism, it must be said in reply that it is no more cavalier than the way in which Platonism is usually dismissed. However, as we have already noted, an appeal to PA can only be the beginning of the enterprise. Categories unacceptable to PA are introduced for reasons. The errors in the reasoning on these things, the errors in the reasoning that lead to the acceptance of transcendent entities must be diagnosed. Once this is done, the dismissal of holistic real mean-

16 ings will appear much less cavalier. What one might say on these points in regard to holism, I shall indicate in a moment, but only indicate, since the point of the present study lies elsewhere. I want first to make some remarks which are both relevant to this latter point, and will pave the way for some of the subsequent discussion. To say that red’s being what it is, is not a matter of its being related to any other property, is to say that (1), asserting that whatever is red is not green, is a synthetic statement. Holists have often advanced a certain argument to establish that such generalities are in fact logical truths.34 Logical atomists have replied to this argument.35 But these replies have at times not been as careful as one would wish. The argument and these replies are what I would now like to discuss. Begin with an example. Consider the law of nature that (12) For any sample of wax, if it is heated, then it melts and the following states of affairs: (13) This is wax (14) This being heated (12), (13), and (14) entail (15) This melts. What might very well happen is this. One judges that (13) is true. One observes the state of affairs that serves to verify (14). But one also observes that the sample does not have the properties which define melting; i.e., one observes that the facts verify (19) This does not melt Equivalently, observations falsify (15). In such a case, we would in all likelihood give up the judgment that (13) is true. We would, in the light of our subsequent observations, withdraw that judgment and replace it with the judgment that (13) is false, or, equivalently, with the judgment that (17) This is not wax is true. In this age of chemical miracles, where it is possible to produce substances which are different yet externally alike, the example is not implausible, and, indeed, others like it are, as a matter of fact, often true. They are useful to the holists. For, it is on the basis of such examples that they erect the following argument against logical atomism: In the example, first (13) and (14) are asserted and then so is (16). The joint affirmation of this is inconsistent with (12). Their joint affirmation requires the rejection of (12), that is, it is so required provided that (12) is synthetic, making a factual

17 claim. But, in point of fact (12) is not rejected. Rather, (12) is asserted in the face of the factual data. What is rejected is one of the factual claims, namely (13). From the fact that (12) continues to be asserted in the face of contrary evidence, from the fact that (12) is not rejected, it is concluded that (12) is not a synthetic proposition, that, rather, it is a necessary truth. This conclusion is arrived at by consideration of an analogous, or supposedly analogous case, one which clearly involves a necessary truth. Consider the generality, true by definition, that (18) All bachelors are unmarried and suppose we have reason for asserting (19) John is a bachelor Then subsequently we discover that after all (20) John is married The factual datum (20) does not lead us to reject the generality; rather, we reject the singular statement with which we began, namely, (19). The generality (18) continues to be asserted even in the face of contrary factual data; we give up one of the factual claims instead of it. It is this parallel between what happens with the clearly analytic(19) and what happens with (12) which leads people to conclude that the latter, like the former, is analytic.36 It is clear that the conclusion of this argument follows only if the premise is added that the non-rejection in the face of contrary factual evidence is the criterion of necessary truth. Unless that premise is added, the argument is a non sequitur. Only if this premise is added does the analogy between the cases constitute a justification of the conclusion that the holists draw. If the non-rejection is not the criterion but at best only a sign that the statement is a necessary truth, then there may be alternative explanations of the fact, the psychological fact, of non-rejection, which constitutes the analogy between the two cases. The holists conclude from their analogy that (15) is a necessary truth, that, in other words, there are real meanings. If the above appeal to PA is correct this is to introduce transcendent entities. The argument involves a transcendental lapse. Within the context of the metaphilosophy we are considering, the argument must be unsound since its conclusion is radically false, violating PA. The error would surely lie in the hidden premise: non-rejection is not the criterion of the necessary. To be sure, it is a sign of a necessary truth; indeed, it is this commonsense truth, misunderstood by the holist, and introduced as a hid-

18 den premise, which lends the argument its plausibility. Substituting ‘criterion’ for ‘sign’ changes the commonsense truth into a commonsense falsehood. That it is false that non-rejection is the criterion of necessary truth is made clear even by (18). For, what makes (18) a necessary truth is, first, the metalinguistic truth (21) ‘Bachelor’ is short for ‘unmarried man’ which tells us that (18) abbreviates (22) All unmarried men are unmarried i.e., (x)[(Ux & Mx) e Ux]; and, second, the fact that the propositional function which forms the matrix of (22) is a truth table tautology37: these two facts, and not the psychological response of non-rejection, are what make (18) a necessary truth.38 Because a person knows these facts, he or she will not reject (18) in the face of contrary evidence. But equally well, he or she may fail to reject a generality for other reasons. And these other reasons may be compatible with the generality being synthetic. The logical atomist can reply as follows to the holist’s analogy between the fate of (12) and that of (18). (13) is rejected instead of (12) since, first, observations have been so made that we are certain that (14) and (16) are true; second, there is overwhelming evidence that the synthetic generality is (12) is true; and, third, our observations led us to asset but not with certainty, that (13) was true. Because our belief that (13) was true was less than certain, it was (13) which was rejected. Because our belief that (12) is true was certain, it was not rejected. But that is compatible with (12) also having the logical status of synthetic. The holist’s argument has thus been shown to be invalid.39 Bergmann has claimed40 that in order to show the argument to be invalid one must hold that certain judgments, namely, those about objects with which one is “directly” acquainted, are absolutely certain, never open to revision, that is, are “incorrigible.” This conflicts with the just-offered alternative to the holist’s account. All that this latter alternative required was that some beliefs be more certain than others at the time at which the rejection is made. This is clearly compatible with idea that what was once certain may well not continue to be so. If the metaphilosophy we are considering is correct, then error is commonsensical. So is the correction of error. The discussion just given of the wax example establishes, I believe, that the correction of error requires

19 that one appeal only to beliefs or judgments which are certain and more certain than others. No appeal to “incorrigible” beliefs or judgments is needed.41 Beliefs and judgments which are certain and have varying degrees of certainty are clearly permitted by PA. The account of the correction of error therefore introduces only entities compatible with PA; i.e., the account is commonsensical, as indeed it should be according to the metaphilosophy. Again, according to this metaphilosophy, holists deny commonsense by introducing real meanings, by holding that laws of nature are necessary truths. They argue that there must be real meanings since otherwise the admitted piece of commonsense that (18) is not rejected in the face of contrary factual evidence is something which cannot be accounted for. It is this argument that Bergmann is discussing. He denies the conclusion of the holists, the conclusion that there are real meanings, and takes it to be commonsensical that laws of nature are synthetic generalities. But he argues that such a denial can be made only if there are “incorrigible” judgments; he argues that one will have to accept the real meanings of the holists unless there is a level of perception (“direct acquaintance”) at which a generality must be rejected in the face of contrary perceptual judgments, a level at which it is logically impossible for the generality to be accepted, a level, in other words, of “incorrigible” perceptual judgments. In short, he argues that, since it is commonsensical that laws are synthetic there must be “incorrigible” beliefs. But this conclusion follows only if one accepts the confused premise of the holists, that the non-rejection of a generality is the criterion for its being a necessary truth. As a consequence, accepting this premise, he, as much as the holists, suffers a transcendental lapse. In this context, namely, that of correction of error, to introduce “incorrigible” judgments is to introduce transcendental entities. For, as we just saw, all that commonsense requires in this context is judgments which are more certain than others; nor is there any reason to suppose that where error is corrected there are the entities Bergmann argues must be there, namely, “incorrigible” judgments. Thus, even if there are such judgments, to introduce them as does Bergmann is to in effect introduce entities as transcendent as the real meanings of the holist. Bergmann accepts the commonsense that laws of are synthetic and contingent. His justification for this is in terms of a PA, along the line suggested above: properties are wholly presented, each as wholly itself.42 This is the ontological use of PA. But Bergmann gives it another use: he requires that it also function as providing a foundation for empirical knowl-

20 edge. We are required to be able to tell on the basis of acquaintance not only that what is presented is red but also that it is in fact red. This is what the transcendental lapse leads to, and in effect it amounts to a confusion of meaning and truth. In order to avoid this Bergmann must abandon the hidden and confused premise that he shares with the holists.43 The reply to the holists consists, on the one hand, of the claim, based on PA, that properties are wholly presented, presented as being themselves and not other things, that being green is not part of the meaning, real or linguistic, of being red; and consists, on the other hand, of replying to that argument that demands either real meanings or “incorrigible” judgments by denying the hidden premise. At no point will such a reply demand the use of PA as a foundation for empirical knowledge. One of the main drives behind holism is the desire to make statements about colour incompatibilities into necessary truths.44 The critics of holism must argue that these generalities are synthetic and contingent. To refute holism it suffices to find some such generalities that are synthetic and contingent. This was seen by Moore, at least if I understand him.45 But the full demands of the atomist’s position were realized, so far as I know, only for the first time by Frank Ramsey.46 It was he who pointed out that such statements as (1) were synthetic and contingent and that that was all there was to it. This, however, only poses the now vitally important diagnostic problem. There must be some point to what the holists were or are trying to say. It cannot be for no reason at all that they were inclined to include (1) among the truths of logic, rather than among the class of mere generalities. Clearly, an alternative account of the commonsense, that the holists were trying however confusedly to capture, must be given before such a diagnosis is possible. I want to turn now to one such account, that of Hume. However, my point is not to provide the mentioned diagnosis; that, unfortunately, would take us far afield. Nor does my point lie in the fact that I believe Hume’s account to be correct. I turn to it, rather, because it is of immediate interest in the context of the metaphilosophy we are examining: the argument for the account is essentially in terms of PA, as indeed Hume himself was well aware.

21 – III – Consider again the generality that red and green are incompatible. The essence of the Humean position is that this generality may be translated into the symbols of logic by (1) (x)(x is red e ~ x is green) The non-Humean, in contrast, holds that it is best represented by (23) (x)(x is red C ~ x is green) where ‘C’ is a non-truth-functional, “nomological” connective. Hume’s point was essentially an appeal to PA. All that we are presented with in ordinary acquaintance or experiencing of the world are things which are red and not also green. We are not presented with anything besides these “constant conjunctions.” In particular, we are not presented with any special connection between the states of affairs that could plausibly be represented by ‘C’. This being so, the incompatibility is represented by (1), not (23). If (23) purports to be a perspicuous ontological representation of the way things are then the representation represents not the way the world is but a transcendental lapse. To take one of many who have argued the same way, let us look at the case for a “nomological” tie that was made by A., C. Ewing.47 Ewing has suggested48 that from the fact that we experience no connection, and, in particular, no connection that might be represented by “C”, it does not follow that there is no such entity. For, from the fact that Aristotle had not observed any black swans, it does not follow that there aren’t any such creatures. This point of Ewing’s, however, is radically misguided. The “nomological” pseudo-relation49 C is unlike chairs and swans. The latter are, after all, physical things, and we are in fact acquainted with entities in this category. But we are not acquainted in ordinary experience with entities in the category of pseudo-relation. Indeed, this makes the “nomological” pseudo-relation different in kind even from Hume’s missing shade of blue. With respect to the latter: first, we are acquainted with entities of the same category, namely, colours or, more generically, properties, and, second, we can give a definite description of the shade, namely, as the one between these other two shades of blue on the colour wheel.50 But, to repeat, nothing like this can be done for C, simply because the category is not presented.51 Ewing admits as much, at least for physical causation.52 The “tie” between cause and effect is, he allows, not observed. Note what follows from this: a scepticism whether we ever know any causal relations or not.

22 All we observe are constant conjunctions; we never observe “nomological” “ties,” not even fallibly; we therefore can never even say what a nomological relation is, that is, identify one, let alone know when we have in fact discovered that one obtains in this case or that. A radical scepticism develops with respect to statements of, say, colour incompatibilities.53 This is indeed an odd consequence, since one of the reasons usually given for holding that there is a special “nomological” pseudo-relation is that only in that way can one avoid the sceptical consequences of Hume’s position,54 or, more accurately, what are held to be the sceptical consequences of the position. But of this, more in a moment. However, first for a further comment on Ewing. He himself notes that these consequences can be avoided since he holds that, though the “tie” is not presented in physical causation, instances of the category and therefore the category itself are presented in mental causation.55 This transforms the case into one analogous to the missing shade of blue. Unfortunately, Ewing does not direct out attention to an instance of a “nomological” “tie.” All he succeeds in drawing our attention to is the fact that volitions are intentional, that they intend, in Brentano’s sense, the states of affairs they are causally efficacious in bringing about. As many have pointed out,56 one event intending another is one thing, its causing the other is a second. Ewing has not, therefore, uncovered for us the “nomological” “tie” we were seeking. Ewing does not succeed in avoiding the radical scepticism about causation that we have noted above. Nor has such a radical scepticism failed to have its effect on the history of philosophy. The Aristotelian account of causation is non-Humean; it involves a special “nomological” connection.57 Within this framework, transeunt causation remained forever a mystery.58 In particular, it plagued the discussion of the relation between mind and body.59 How could a spiritual and active substantial mind affect the non-spiritual substantial body, and conversely? Malebranche adopted occasionalism, or, at least, made explicit Descartes’ occasionalism.60 In effect, this is to adopt a Humean approach to the issue: there is no special tie, only constant conjunction. But since what Malebranche wanted was a special non-Humean connection, his position amounted to an adoption of scepticism,61 specifically, the radical philosophical scepticism parasitic upon the demand for a transcendent causal “tie.” If, however, one gives up the demand, and adopts the Humean view of causation, then one is committed to no such scepticism. The observed constant conjunctions between mind and body are the causal connections, and that is all there is to it. The mystery disappears.62

23 It is very strange, then, to discover that many believe Hume’s view to have sceptical consequences. To the contrary, it is an attempt to avoid a philosophically problematic scepticism. Many times the objections to Hume are most naïve: for example, we have it that if Hume is correct then everything is mere accident,63 as if the term ‘accident’ was not itself loaded, being used in this context philosophically and for that reason requiring explication, if indeed it can be explicated, since its function clearly is to smuggle in the transcendent “nomological” “tie” for which the non-Humean lusts. It is of course true that the Humean must account systematically for all the various causal distinctions we make. Thus, in order to distinguish chance from regularity, he or she will distinguish statistical from non-statistical generalizations. In order to distinguish explanation from mere forecasting, he or she will distinguish more and fewer of the relevant interacting variables in the system with which one is concerned, the more gappy from the less gappy, the more imperfect from the less imperfect.63 Mainly though, he or she must account for the distinction between post hoc and propter hoc. This he or she does by indicating the different psychological attitudes which attach to generalities of the latter sort but not to those of the former.64 Ontologically no difference is admitted between laws and “accidental” generalities.65 The difference is not objective, lying in their (onto)logical form, but rather is psychological. The generalities which are laws are precisely those which we use to make predictions and which we use to support the assertion of subjunctive conditionals: this is their defining characteristic.66 (It is in terms of such attitudes that the truths of logic and those of colour incompatibilities come to be grouped together in the same class; this is how the Humean explicates the commonsense core of the position which the holists wish to defend.67) Laws and “mere” regularities are alike in this, that they both involve nothing more than constant conjunction, i.e., are represented in an ontologically perspicuous language as Russellian formal implications; such constant conjunctions are compatible with PA. The two are unlike in that laws involve a psychological attitude that “mere” regularities do not; such attitudes are compatible with PA. Thus, within the framework of the metaphilosophy we are considering the Humean account of causation is not philosophically problematic; rather, it is commonsensical. Upon this philosophically non-problematic Humean position, induction is open. Causal relations, and, indeed, all matter of fact generalizations are never “known for sure,” since, on the one hand, we know on the basis of acquaintance only some instances of the generalization, while, on the

24 other hand, “knowing for sure” is so defined that a generalization can be “known for sure” to be true only when one has come to know on the basis of acquaintance all the instances of the generalization. The demand that inductive generalizations when known to be true are known for sure becomes self-contradictory.68 In this sense, every inductive generalization is open to doubt. We can be sceptical, in this sense, of every inductive judgment. But this doubt, this scepticism, is not philosophically problematic: it is commonsensical. So, from the viewpoint of the metaphilosophy that we are considering, such doubt constitutes no philosophical problem. To be sure, one’s philosophical categories must allow one to account for it; but it is certainly not anything that one need try to overcome. At the same time, the problem of causality, if it is looked at in nonHumean (e.g., Aristotelian) terms, does give rise to scepticism, that is, philosophical scepticism, parasitic upon the introduction of a transcendent causal “tie.” It is clear that it can be easy to confuse such a scepticism with the sort of commonsense, philosophically non-problematic scepticism to which the Humean is committed, and then to demand that the Humean solve these sceptical issues just as one can demand of the Aristotelian that he or she solve the ones that arise within his or her philosophy. Only, we have just seen, this is a confusion, at least within the framework of the metaphilosophy that we are considering, and given that framework, the Humean can legitimately resist the demand. Sceptics seem to have taken these positions seriously. Thus, Sextus, for one, seems to have accepted this sort of metaphilosophy, and did not take the openness of induction to constitute his scepticism. He insisted that a sceptic could order his or her life according to the natural signs he discovered in the world — that is not scepticism with regard to induction, or at least not the radical scepticism about transcendent causal ties – it is just that he or she should never affirm, or deny, dogmatically any inductive generalization. His (Sextus’) scepticism was constituted rather by the fact he suspended judgment upon all philosophical claims about transcendent entities and connections behind the presented signs.69 Hume, too, held this position.70 At least so I would argue against those who would make a radical and dogmatic sceptic of him, one who rejects not only transcendental entities but also calls into question our commonsense view of the world. I do not propose to defend this claim in detail. I shall merely quote one passage and suggest that if the ‘dogmatism’ in it is so read as to mean the introduction of transcendental entities, and the ‘scepticism’ is so read as to mean the philosophically problematic scepti-

25 cism that is parasitic upon the introduction of such entities, then we obtain a statement which rejects the transcendental entities and the philosophical scepticism but accepts as non-problematic and commonsensical the scepticism connected with the openness of induction. The sceptical and dogmatical reasons are of the same kind, though contrary in their operation and tendency; so that where the latter is strong, it has an enemy of equal force in the former to encounter; and as their forces were at first equal, they still continue so, as long as either of them subsists; nor does one of them lose any force in the contest, without taking as much from its antagonist. It is happy, therefore, that nature breaks the force of all sceptical arguments in time, and keeps them from having any considerable influence on the understanding. Were we to trust entirely to their self-destruction, that can never take place, until they have first subverted all conviction, and have totally destroyed human reason.71

I suggest that we might well follow Hume and accept PA, and do our ontology within nature; then neither philosophical scepticism nor dogmatism will subvert our understanding. Nor was Hume the first to make such a point. I have mentioned Sextus, noted Bayle, and praised Hume. I must also mention Berkeley, to whose rejection of material substance we now turn.

– IV – Berkeley was perhaps the first to adhere more or less consciously to PA as a principle in ontology. He employed it as an effective weapon against Locke’s resolution of the problem of perceptual error. And this is, of course, an ontological problem, at least to this extent, that one’s ontological categories are such as to permit a distinction between appearance and realty, between fake and real, between the reality or possible reality of the dagger Macbeth perceives and its unreality or possible unreality. Locke’s account of perceptual error is this.72 One is acquainted with sensible objects; this is non-controversial: everyone agrees that we are presented with sensible entities. This is the world as we ordinarily experience it. These sensible phenomena Locke calls “ideas”. On the Lockean ontology, (and ignoring inner awareness,) one is acquainted with nothing else. Then there are perceptual objects. These perceptual objects are entities with which we are not acquainted; however, they are the causes of the

26 ideas with we are acquainted. These perceptual objects, these causes of our ideas, are represented to us by the ideas they cause. Since perceptual objects are never presented to us, they constitute a category of transcendent entities. (Exactly how Locke characterizes these objects is not relevant for what we are about.) Ideas may represent these transcendent objects correctly or incorrectly: the truth and falsity of our ideas are determined by the objects that they represent and which transcend the sensible world, that is, our world as we ordinarily experience it. If the idea presented to us correctly represents a perceptual object, then no perceptual error occurs. If the idea incorrectly represents a perceptual object, then error occurs. The presentation of incorrectly representing ideas is explained, in principle, in a straight-forward scientific, mainly physiological, way. This clearly solves the ontological problem of perceptual error. Locke’s categories are such that perceptual error is possible: a distinction between appearance and reality can be drawn. However, this solution is obtained only at the cost of introducing a transcendent entity. And with this transcendent entity comes a radical philosophical scepticism. Because one of these categories is that of a transcendent entity, it becomes impossible in terms of these categories to ever come to know when we are in error. Never being presented with the perceptual object, we are never able to discover whether the ideas with which we are presented do in fact correctly represent it to us.73 Berkeley is out to slay this radical scepticism.74 His central move (he has several parts to his overall argument) is an appeal to PA. The scepticism is parasitic upon the representationalism, that is, upon the perceptual objects being transcendental. The objects, in other words, which are supposed to be there as the causes of our impressions and ideas, are outside of and transcend our ordinary experience. The scepticism can easily be destroyed – perhaps too easily. The full story is more complicated, but for our purposes we can restrict our analysis to the simpler formulation. So, the point here is this, that the scepticism can be removed by PA used to exorcise the transcendent objects, what are, on the Lockean scheme, the perceptual objects. But the Berkeleyan use of PA establishes that, whatever sort of thing perceptual objects might be, it is certain that they are not Lockean objects: this is the upshot of the Berkeleyan use of PA.75 This removal of the transcendental perceptual object of Locke by appeal to PA does not require that our acquaintance with the impressions that are held to represent the perceptual object are somehow “incorrigible” acts of experience. Lockean objects are, Berkeley argues, “unintelligible.”76

27 The removal of the radical scepticism at which Berkeley aims therefore does not require that our acquaintance with sensations, that is, with ideas of sense, be “incorrigible”; in other words, it does not require that our acquaintance with ideas of sense provide a foundation for empirical knowledge. It would be perfectly consistent if Berkeley were to hold with respect to sensory experience the position introduced above in the context of the wax example: he could hold that we might in the light of further evidence come to revise our beliefs about previously presented sensory objects. Of course, this is not the position Berkeley in fact holds. He believes that empirical knowledge does have a foundation of “incorrigible” awareness. But his reasons for believing this have nothing to do with the rejection of the radical scepticism parasitic upon Locke’s account of perception. His reasons for holding that there is a foundation for empirical knowledge are quite independent of, and have nothing to do with, his attack on scepticism. This assertion requires two comments, the first to indicate that the only scepticism about which he was worried, was a radical philosophic scepticism; the other comment to indicate the independent reasons he had for holding that awareness of sensations did indeed provide a foundation for empirical knowledge. That Berkeley was not worried about commonsense scepticism is readily established by the fact that he had no hesitation whatsoever about accepting the Humean regularity analysis of causation so far as the causal connections among the presented entities is concerned.77 The commonsense scepticism that this involves did not worry him; that, owing to the openness of induction that the Humean view entails, we might err in the causal inferences we make, did not lead Berkeley to reject the view. But he did reject Locke’s account of perception. The conclusion is, I think, that it was the scepticism parasitic upon Locke’s view that he attacked with PA. He did, after all, hold to a non-Humean view of causation so far as the creation of the objects presented to us was concerned; that was how he arrived at the existence of a God.78 Nor, more importantly, did he exorcise the active self.79 Indeed, Berkeley’s own ontology is everywhere haunted by this machineless ghost. Specifically, it is precisely this ghost that provides the crucial component for the claim the acquaintance with ideas of sense is “incorrigible.”80 Berkeley’s minds are substantial. To the Aristotelian tradition from which this sort of ontology derives, properties – the sensible properties of things – inhere in substances. When they do, they are truly predicable of the substance in which they inhere. It is part of the anti-Platonism of this

28 Aristotelian tradition to hold that there are no separable properties, or, what amounts to the same, that properties must inhere in substances.81 Berkeley construes the connection between his substantial minds and the ideas of sense with which they are acquainted in terms of the inherence tie.82 For mind to be acquainted with, or presented with, an idea of sense, is for that idea to inhere in the mind; for one to be acquainted with a property is for that property to be in the mind in the sense that it inheres in that substance. It follows directly that all acquaintance with ideas of sense is veridical, that sensory experiences are “incorrigible.” Here is the source of Berkeley’s claim that acquaintance with sensory ideas provides a foundation for empirical knowledge. Too, it is Berkeley’s idealism.83 Ideas of sense are related to substantial minds via the tie of inherence. From the Aristotelian tradition comes the view that properties must inhere in substances. Ideas of sense must, therefore, inhere in substances, and since the only substances allowed by the Berkeleyan scheme are active, that is, mental substances, it follows that the ideas of sense must inhere in minds: in short, and in Berkeley’s formula, for ideas of sense, esse is percipi. Indeed, it is his idealism that leads him to construe acquaintance in terms of inherence; otherwise the view has little to recommend it.84 For, if inherence is understood on the Aristotelian pattern, then what inheres in the mind will be predicated of it. One will have to predicate of the substantial mind the ideas of sense; a Berkeleyan mind will therefore turn out to be, say, blue, or green, or what have you, depending upon what one is acquainted with at a given time, Here, however, we have surely failed, with a vengeance, to “account for consciousness.” Colours, etc., are just not properties of mental things, and that is all that there is to it. Nor was Berkeley unaware of this issue. He explicitly denies that ideas of sense are to be predicated of his substantial minds. Rather, predication is to be understood in part-whole terms; the sensory quality is predicated of the complex of which it is a part.85 Berkeley thus avoids challenging commonsense, but only at the cost of divorcing inherence and predication. The consequence of this divorce is that he has no longer any reason for being an idealist, that is, no reason for holding that ideas of sense must inhere in mental substances. In order to avoid Lockean abstract ideas, which is what separable properties become when separated, then one must hold that there are no properties apart from the things of which they are predicated. For the Aristotelian, and more generally for the defender of the substantialist tradition,

29 this amounts to saying that the properties must inhere in a substance. The corresponding principle in Berkeley’s ontology would be the principle that there are no properties apart from complexes of properties. The different principles reflect the different accounts of predication. Berkeley wishes to avoid the separable properties that become Lockean abstract ideas: he wishes to avoid that sort of Platonism. In order to secure his position, he holds that ideas of sense must inhere in substances.86 This has no warrant, given his account of predication. Or, in other words, his idealism has no warrant. When this idealism is given up, his main reason for holding that acquaintance is inherence, and for the “incorrigibility” of sensory experience, also disappears. I say “his main reason” deliberately: for, clearly, one of his reasons is that inherence is the only tie or connection which the tradition bequeathed to him. He was therefore more or less forced to account for acquaintance in terms of it. What is required is a complete re-thinking of the idea of the mental, and of the “aboutness” or intentionality of acts of acquaintance.87 What this re-thinking of the mental does not require, as our wax example showed, is at least one of the features of Berkeley’s account, namely that some acquaintance is “incorrigible.” Berkeley slays the radical scepticism entailed by Locke’s account of perceptual error. This cannot be the end of the matter for our mitred empiricist, however. After removing Locke’s transcendental objects, Berkeley himself is faced with the ontological problem of error. He must introduce categories, compatible with PA, and such that perceptual error is possible. The crucial term is ‘perceptual object.’ What he suggests is that a perceptual object is a “pattern of ideas,” which is to say, a pattern of phenomenal objects.88 Thus, the only entities that occur in perceptual objects are entities of a kind with which we are acquainted. The pattern is a lawful one, where the laws are treated in the fashion of Hume. The pattern is therefore one that is learned through experience, though what one thereby learns is always correctable in the light of further experience: the learning is the learning of an inductive generalization, and induction is always open.89 Berkeley’s category of perceptual object thus does not violate PA.90 Berkeley goes on to argue that this solves the problem of perceptual error. Perceptual error occurs if one has a false belief about the pattern, that is, if one has not correctly learned the laws definitory of various kinds of perceptual objects, and on the basis of which one predicts which further ideas of sense will be presented to one.91 Berkeley’s own account of perceptual error thus reduces it to the

30 sorts of error that are involved in inductive inference.92 We see why he must proceed in this way. The property of being a cherry tree, for example, cannot be wholly presented in sense experience. If it were, then it would inhere in the substantial mind to which it was presented. In that case, the presentation would be veridical, and it would be impossible to falsely perceive a cherry tree. But at the level of cherry trees error is indeed a fact of life, and Berkeley is therefore forced to some other solution to the issue of the status of the property of being a cherry tree, one that allows that this property and other properties of perceptual objects that are like it, need not be wholly presented in the way in which it is required that simple properties be wholly presented. It is therefore clear that he, or at least one, need not offer an account of perception such as the one he does defend, one in which perception is a variety of induction. One could offer, alternatively, the simple view that besides the sensing of sense impressions, there are perceptual acts in which one is presented with physical object properties such as that of being a cherry tree. All that one needs is an account of mental acts over and above the sensing of sense impressions – where one’s account of such acts allows that they can be erroneous but also that we can discover such error. For, what one requires of one’s ontology is, on the one hand, that no perceptual presentation be “incorrigible,” that all be “corrigible,” and is, on the other hand, that it (the ontology) not be such as to render categorially impossible our being able to revise perceptual beliefs we come to think are erroneous: the wax example establishes that this really does suffice for one’s ontological account of error. Unfortunately, the Bishop of Cloyne’s inadequate categorial structure, and specifically his inadequate account of the act, prevents him from proceeding in this fashion and forces him to make inductive and perceptual error one of a kind. In the light of these remarks, some comments on a reading of Berkeley by Harry Bracken are in order. According to Bracken, The sceptic is stood prepared to eliminate all claimants to knowledge; in particular those who distinguished ideas from things, appearances from realities, esse from percipi, were ever open to sceptical questioning as to how they bridged this distinction.... [I]n good part Berkeley’s dictum esse est percipi, stems from Berkeley’s sensitivity to this sceptical problem and is intended to resolve it by his revolutionary stroke of identifying ideas with things. Deprived of the handle afforded by the thing/idea dichotomy, scepticism stands disarmed.... The reason Berkeley’s refutation of scepticism is bound up with his appeal to common sense is that there must be no instance in which the philosopher tells the gardener that what the latter holds to be the case about the sense

31 world is not really the case. The philosopher must never reopen the appearance/reality dialectic by saying that the world is really anything other than what the gardener takes it to be.

But Berkeley must face up to the problem of error, Bracken suggests. Berkeley’s response to this problem is what Bracken calls a “sense datum” analysis of perceptual objects, that is, the latter are construed as collections of ideas of sense (sense impressions) where the collections exemplify different patterns or regularities depending upon the sort of object it is – the cherry tree pattern is different from that of the oyster. The result [Bracken argues, with respect to the sense datum analysis] is disastrous to his [Berkeley’s] claim to hold, with the “vulgar,” that “those things that they immediately perceive are the real things”[93], and hence he falls guilty ... of scepticism as he himself defines it... [I]n spite of his own intentions, Berkeley is driven towards scepticism by urging ... an argument that amounts to saying everything is as it is, but we must wait an indeterminate period to find out what it really is, for what it is, is what we expect it to be .... Berkeley refutes scepticism by a dramatic denial of “screens,” to use Jessop’s apt phrase, interposed in perception between knower and known – that is, by identifying ideas with things. But, as I have tried to show, Berkeley’s invocation of sense-datum analyses involves his postulating a temporal screen, the imposition of which entails the abandonment of the idea-thing “naive realism” of the appeal to the gardener and the refutation of scepticism dependent upon it.94

One might begin by asking whether the gardener does have the “naive realist” view which Bracken attributes to him or her. Does the gardener really believe that all his or her presentations are veridical, that all his or her perceptual beliefs are true, that there is no distinction appearance and reality? Surely not, must be the immediate reply. Error, as such, is not philosophically problematic; it is to the contrary very much a part of our commonsense view of the world. Specifically, as we have seen, it presents no perplexities that can only be removed if there are perceptual beliefs which are not subject to error, which are, in other words, “incorrigible.” To open up the appearance / reality distinction will not by itself lead one to adopt a view that is in any way at variance with that of the “vulgar,” and insofar as Bracken suggests otherwise he is mistaken. Error, being part of commonsense, is something “to be accounted for,” not denied. It is when one tries to account for it that one finds philosophical perplexity creeping in. In particular, attempts to account for the appearance-reality distinction after the fashion of Locke, in terms of a category of transcendent entities, introduces a radical philosophical scepti-

32 cism. It was in order to exorcise this that Berkeley collapsed the perceptual-object / presented-object distinction as it was found in Locke. However, to collapse that distinction, Bracken not withstanding, does not require one also to ensure somehow that inductive inference is, sometimes at least, infallible, nor does it require that Berkeley eliminate the possibility of ordinary perceptual error. Berkeley did attempt to account for the commonsense fact of perceptual error, for (in other words) the distinction with regard to perceptual objects of the appearance and reality. This is why he offered his analysis of perceptual objects as patterns of sense data. Perception was then analyzed as a species of inference, and perceptual error as a species of inductive error. This account of perceptual error involves what Bracken refers to as the introduction of a “temporal screen.” The point to be made against Bracken’s criticism is that such a “temporal screen” is not by itself philosophically problematic. That is, such a “screen” does not introduce a philosophically problematic scepticism, only the piece of commonsense that inductive generalizations are never conclusively verified. This being said, that Bracken is mistaken in holding that the commonsense Berkeley backslides into scepticism because he fails to overcome the openness of induction, one must also say, I think, that it is also true, that he (Bracken) has pointed out a serious flaw in the views of the Bishop of Cloyne. If Berkeley is correct, then when one sees a tree, what goes on is this: first, one identifies some combination of sense data (ideas of sense, in Berkeley’s terms) and the relations in which this combination stands to other, perhaps similar, combinations; and second, on the basis of what one has learned inductively, one infers that in the future certain other sense data will be presented, and that under appropriate conditions still other sense data would be presented. Both these things are involved in the fact of one’s awareness of a tree in perceptual experience, upon Berkeley’s account of perceptual experience.. However, – and this is the point – , this is undoubtedly false as an account of being presented with a tree. We no more infer in the case of a tree than we do in the case of something red.95 We predicate ‘tree’ of the presented object on the basis of the property the thing is presented as having. The property of being a tree is as wholly presented to a perceiver as the property of being red. Just as we must reject the idea that there is any inference involved in identifying a presented object as red, so we must also reject the notion that there is any inference involved in identifying an ob-

33 ject as a tree. This point, about the property tree being wholly presented, I take to be a piece of commonsense, and more specifically something that can be answered by appeal to PA. Nor need the presentations of trees be veridical for the ontological appeal to be successful. Berkeley does not account for this piece of commonsense – which seems to be the main point behind Bracken’s contrast of the Berkeley of commonsense and the Berkeley the philosopher who equates perception and inductive inference. It should be remarked that, although Berkeley cannot, as we saw, account for this piece of commonsense within his categorial structure, he was nonetheless still sensitive to the problem, for I believe that this was one of the inadequacies of his system which he tried to handle in terms of his (unfortunately somewhat obscure) doctrine of “notions.”96 Nor, finally, do I want to be understood as claiming that Berkeley’s account of perceptual objects, as patterned collections of sense impressions, is wrong. To the contrary, it seems to be right, or on the right track: the principle that he is defending, that the world is as it appears to be, while needing qualification, is surely right.97 And if it is, then things we perceive are bundles of sense impressions. But this is compatible with the following. Consider our tree. In perception we see this tree, its brown bark, its green leaves, the toughness of the trunk. So, in perception we are presented with some sense impressions and these are presented in an act of perception as being of a tree – and if that act is veridical, then those impressions given in sense when we perceive the tree are in fact impressions of a tree. They are thus perceived as being of a tree; and that characteristic, being of a tree, is presented to the perceiver and is wholly presented. And if that pattern is there, in the world which we sense and which we perceive, then the act of perception is veridical. It is of course true – and this is the point I wish to make – that the pattern could be attributed to the impressions we sense and to many others which we do not sense on the basis of inductive inferences – that follows from the (ontological) claim that the perceptual object just is a patterned bundle. However, it is also true that the impressions we sense are situated relative to other impressions, most not sensed; that the perceptual judgment in which they are taken to be of a perceptual object (in our case, a tree) which is itself not given as an inductive inference; that the facts which constitute the patterned bundle do each of them exemplify the property of being of a tree; and each of them may, in appropriate circumstances, be perceived non-inferentially to be of a perceptual

34 object pattern – the pattern which can be known inductively can also be presented non-inductively, wholly, as a property of sensed particulars. The property of the sensed particular is a determinable property. We do not sense the other side of the tree nor the parts within the outside of the tree. But there are other parts there, particulars that are not sensed but could be, unsensed sensiblia; and there are properties which they have, sensible properties which are however unsensed by me at this moment, where these properties are similar to or resemble generically the properties of particulars which I do sense or which I have sensed; and there are relations in which they stand to one another and to the particular which I am sensing, relations which resemble generically relations which are or have been presented to me. The property presented in perception is not a sensible property, nor are sensible properties and relations part of that property: the property is the property of there being such entities there in the world, but it is not one of those entities. The tree in the sense of the whole bundle of sensible particulars is not presented to me, that is, the tree as it is in all its particularity and all the specificity which those parts have. Or, to put it yet another way, the particular parts which, when bundled, constitute the tree, and all the specific properties which those particulars have are most of them not presented to me – in fact, none are given save, of course, those sensible parts which are given to me in sense experience and which form the impression of the tree. Thus, while we are not wholly presented in sense with the tree, that is, with the tree as an ordered bundle, past, present and future, of sensible parts, we are wholly presented in perception with the fact that this, a sense impression, is (of) a tree. The perceiving is non-inferential; unlike an inference, which is a movement of thought, it is a single, unified, act. It is nonetheless logically speaking like an inference with regard to the tree in being inductive, making a claim as it were about unsensed, past, present and future, parts of the bundle which is the tree. In our unified acts of perceiving we are directly aware of, that is, non-inferentially aware of, the tree, the bundle, as a whole. Not all are aware of this point. Thus, to take one example, Quinton98 wishes to argue that the ontology that Russell, following Berkeley and Hume, had developed, an ontology in which ordinary things are construed as patterns of sensible appearances, is mistaken. He wishes to argue that the bundle view, what he calls “phenomenalism,” is false. He proposes that we are as directly acquainted with ordinary objects as we are with the sensible appearances of those objects, and concludes that therefore the former

35 are not “mere” patterns of the latter. He does not say that he is advancing a sort of substance ontology, but he does deny that the position he wishes to defend is a form of representationalism: to the contrary, he holds, there is no veil of appearances that comes between us and ordinary objects, making the latter beyond our cognitive capacities. Quinton attacks the very notion that one can speak of the appearances of things as entities – as “sense data” or “ideas of sense” (Berkeley’s expression) or “impressions” (Hume’s expression). It is wrong, he argues, to infer that since things appear in certain ways, there are entities which have the properties presented by those appearances. A statement like (24) This tree appears to be brown about ordinary objects of perception does not imply that there is on the one hand the tree and on the other hand something else referred to as a sense impression or sense datum, where this sense datum is experienced as brown. To the contrary, a statement like (24) is simply a guarded way of making the same assertion as (25) This tree is brown which attributes a colour not to a sense impression but to an ordinary object. When used in everyday and ordinary contexts, (24) expresses the same judgment as (25), but gives it to be understood that the person using it is less than fully certain about its truth.99 Now, what needs to be noted is that Russell or Berkeley or Hume can grant this point. The argument that they offer for sensible appearances turns on the simple point that when we perceive in normal circumstances that the tree is brown we are presented with an expanse, an area if you wish, which is coloured brown. In this sense, what we experience is a patch that is brown: there is an impression, there in what we experience, which is brown. What we do when we use (25) to express our perceptual judgment is locate that brown patch as part of a pattern of similar impressions, actual and possible, that is, as actually sensed and as possibly sensed; it, the brown patch or area which is there, present to me, is presented in the perceptual judgment as part of a patterned whole which we describe as a brown tree. (25) describes the object of our perceptual judgment. It is open to one defending the bundle account of ordinary things to hold that one can make such an assertion in a way that indicates the certainty of one’s judgment, using (25), or in a guarded way, using (24). Nothing about Quinton’s account of the relation between (24) and (25) implies something inconsistent with the bundle analysis of ordinary things as collections of sensible appearances.

36 Quinton has another argument against the bundle account. We are told that it is wrong to hold, as the bundle account does, that in perceptual judgments we locate as part of an ordinary object an impression given to us in sensible experience, because in such perceptual judgments we not only are not attending to the sensory core, the impression, but are often hard put to give an accurate description of that core: we are much more certain about the perceptual judgment than we about the sensory core of that judgment.100 Thus, we are told that ...experience cannot be the sole object of acquaintance since it is not the case that in every perceptual situation we are aware of it.101

But again, why should the defender of the bundle view not grant this? In fact, Russell makes much the same point during his discussion of the analysis of ordinary things into patterns of sensible impressions: The painter wants to know what things seem to be; the practical man and the philosopher want to know what they are; but the philosopher’s wish is stronger than the practical man’s, and is more troubled by knowledge as to the difficulties in answering this question.102

Only rarely do we attend to the sensory impression; usually it is the object of perception to which are attending. But none of that implies that the object of perception is not a patterned set of the ways in which it appears in and out of sensible experience. Nor does it imply that in perception we are not locating in such a pattern the sensible impression which we are experiencing. Quinton thinks all this is relevant to refuting the bundle analysis because he construes the latter as based upon an epistemological argument. Thus, “...while no knowledge of material objects is direct, all or only knowledge of [sensible] experience is direct. In more linguistic terms, while no statements about objects are basic, all or only statements about experience are basic.” Here, Quinton tells us that he means by ‘basic’ that a piece of knowledge is basic only if it can expressed in a direct statement, one which is not justified indirectly by some inference from other judgments but directly: it is a statement of something non-inferentially known. There is, then, as Quinton sees things, an argument for judgments about sensible appearances being basic. ...beliefs about objects are never certain, beliefs about experience are always

37 certain, and ... for any uncertain belief to be even probable something else must be certain.

From this it follows that material object statements must be inferable from premises about the more basic judgments about sense impressions. And since material object statements are thus evidentially dependent upon statements about appearances, the former must, in their logic, be statements about complex patterns of sensible impressions. But the premises of this argument are, he proposes, all false, and he therefore concludes that there are no grounds for accepting the bundle analysis of ordinary objects.103 Quinton argues that it is false that beliefs about objects are never certain: for, we quite often do have beliefs about material objects that are as certain as anything. But the same is true of many inductive inferences. That the sun will rise tomorrow is based on inductive inference, but it is as certain as any judgment we will make. Similarly, many perceptual beliefs are as certain as any judgment might be, e.g., the judgment that I am now sitting at a desk composing an essay on ontology. Why should one defending the bundle analysis not agree with this? Just because a perceptual judgment has the logical form of an inductive inference, it simply does not follow that it is not certain. Quinton also argues, as we have seen, that beliefs about experience, i.e., about sense impressions, are often uncertain. But as we have also seen, there is no reason why a defender of the bundle view ought not to agree with this also. After all, as we have suggested, more often than not it is the perceptual object to which we are attending and not to the impression we have in our sense experience. Finally, Quinton argues that there can be judgments that are uncertain but are not based on inferences from prior judgments that are in themselves certain. Again, why should the defender of the bundle account have to disagree? To be sure, inductive inferences are, logically speaking, uncertain: as Price made clear, our perceptual judgments make ampliative assertions that go beyond the instances which could have formed their evidential basis. In this sense, every perceptual judgment goes beyond the instances on the basis of which we learned to make such judgments. But, taking these things phenomenologically, perceptual judgments are not inferences, nor, in terms of the psychology of the case, can they be construed as being the upshot of inferences, of a conscious movement from premises to conclusion. But perceptual judgments made in full daylight are more certain than perceptual judgments made at twilight. Among the things that we

38 have learned is that the latter sort of judgment must be made more guardedly than the former. And as was made clear with the example of statements (24) and (25), the more guarded statement will be made using the language of appearances. As we emphasized with (24) and (25), however, one cannot conclude from this, as Quinton attempts to do, that the ontology is not that of the bundle view. By taking the argument for the ontology of the bundle view to be based on an inference from what must be epistemologically basic in a chain of inferences, Quinton provides himself with an easy target. If he is correct in his analysis of the case for the bundle account, then his position is quite indisputable: no case has been made. But the argument for the bundle account is quite different. To be sure, Russell does base his case on something like the Cartesian method of doubt. But this is not the Cartesian method of doubt, which has as its aim the clearing away of those idols that lead us into error, and prevent us from discovering or uncovering the true and infallible foundations of all knowledge. Russell instead uses the method to lead us to those features of the world in which we live that are not the result of an acquired capacity to recognize those features as falling under a pattern that points to other features of the world that are not presented to one in one’s sense experience. There is nothing incompatible between a bundle ontology, on the one hand, and, on the other, there being perceptual judgments which are cognitively more securely based than the corresponding awareness of the sense impression which that judgment locates as an aspect of a material object, located in space. Chisholm has advanced a similar though stronger argument for the same conclusion as Quinton, that perceptual object statements about material objects cannot be arrived at by inference from statements about sense impressions. Chisholm considers the possibility of inferring statements about material objects from statements about sense impressions. He draws the conclusion that perceptual judgments about material objects must be as inferentially basic as judgments about sense impressions. He indicates that “...it is very difficult to think of any proposition about the ‘external world’ which is probable – more probable than not – in relation to any set of propositions about the way in which one is appeared to [in sense].” That is to say, it is very difficult to think of a set of statements of this sort: one of them is a synthetic statement, attributing some property to a material thing; the others are statements of the form, “I am appeared to [in sense] in such and

39 such a way” ...; and, finally, the statement about the material thing is probable – more probable than not – in relation to the statements about appearing.104

Chisholm defines ‘empiricism’ to mean that the only basic statements are those which describe our sense impressions, how we are appeared to in sense. He is suggesting that if empiricism in his sense is true, and if there are no material object judgments which are confirmed by statements about our sense impressions, then we can never make judgments about material objects that are even the least bit probable. His argument thus aims at justifying the conclusion that if empiricism in his sense is true then it will never be reasonable to accept any material object judgment: scepticism about material objects is the only thing that the empiricist can conclude. He suggests that Hume, for one, subscribes to empiricism in his (Chisholm’s) sense and that he (Hume) draws this sceptical conclusion, when he asserts that “it is in vain to ask, whether there be body or not?”105 This is perhaps not fair to Hume; after all, there are a variety of reasons, some good, some bad, why he might hold that it is vain, and unreasonable to ask whether there be body, but let us leave Chisholm have his reading of Hume. In any case, Chisholm himself aims to avoid scepticism about material objects, and therefore holds that perceptual object judgments ought to be taken to be as equally basic in the line of epistemological justification as judgments about sense impressions.106 Now, the defender of the bundle account might well accept that perceptual object judgments are as certain as judgments about sense impressions, and that they might therefore be taken as equally secure epistemologically. In that sense, such a one might well agree with Chisholm. But Chisholm takes his argument to establish that perceptual objects cannot be understood ontologically as patterns of (sensed and unsensed) sense impressions; he takes it to establish that a realistic ontology of the bundle kind is in error: besides sense impressions and patterned collections of sense impressions, there are also unanalyzable material objects. This raises the ghost of the representationalism that Berkeley set out to slay. Chisholm escapes that scepticism only by insisting that we have in our perceptual judgments a secure awareness that takes us beyond the world of sense experience to a world of material objects unrelated to the objects of sense experience. Berkeley had already criticized this position as requiring us to have concepts not derived from sense: the Berkeley-Hume attack on abstract ideas as the vehicles that are supposed to enable us to judge about material objects construed as unrelated to what we know by way of sense experience is equally an attack on the concepts or whatever that are unre-

40 lated to sense but are, according to Chisholm’s account of the world, the vehicles by which we non-sensibly experience things like chairs and tables and oak trees. But the issue before us now is whether Chisholm’s argument establishes that material objects could not be patterns of sense impressions. The argument is that a material object statement is compatible with any sense judgment that we make, and that the latter cannot therefore support the former. That being so, the former could not be judgments about patterns of sense impressions. As R. Firth has argued against Chisholm, it is simply not true that judgments about sense impressions never count as confirming or disconfirming hypotheses about the nature of the material thing that they are taken to be of. Firth, debating Chisholm’s argument, considers the case where he perceives something moving in the shaded underbrush of his garden. It seems to him that what he is seeing is a cat. But he is not sure. So he examines more carefully how he is struck by that thing; he examines more carefully the impression of it that he is given in sense. Upon examining that impression, he decides that it is not really a cat after all, it is a dog, of about the same size as a cat. So Chisholm is wrong to say that I never rely upon an examination of my impressions to help me decide among alternative hypotheses about the sort of thing that impression is an impression of. And if this is the logical relation in some cases, then it is the same logical relation in other cases, cases where I never do rely upon a careful examination of my sense impressions to determine the kind of material object to which I attribute those impressions. Just because I am not doubting what sort of material object I am perceiving, it does not follow that that perceptual judgment does not stand in the logical relation to the sense impression of an hypothesis about the pattern into which that impression fits. That is, from the fact that the judgment is somehow evidentially certain, it does not follow that judgment is not about what is ontologically analyzed into a pattern of sensed and unsensed sense impressions.107 We may conclude, then, that Chisholm has not provided grounds for rejecting a bundle ontology of perceptual things as patterns of sensed and unsensed sense impressions. Chisholm wrongly infers that a bundle ontology is mistaken and that we need to return to a sort of representationalism because he rightly notices that some perceptual object judgments are more certain than judgments about our sense impressions. But in fact, there is nothing incompatible between a bundle ontology, on the one hand, and, on the other, there being perceptual judgments which are cognitively more se-

41 curely based than the corresponding awareness of the sense impression which that judgment locates as an aspect of a material object, located in space. The point is that the bundle account of ordinary things – trees and tables, shadows and rainbows, and straight sticks that appear bent in water – is an ontological account based on a Principle of Acquaintance. It is not thereby also a position to the effect that our sensible experience of sense impressions alone is the basic starting point for epistemology. Ontology and epistemology are different things, and a starting point for the former need not be a starting point for the latter: the use of a PA in ontology does not commit one to taking ontologically basic entities to be epistemologically basic.

42 Appendix One The Mythology of the Myth of the Given: the Holism of Wilfrid Sellars Wilfrid Sellars has presented a case against empiricism. As part of this case, there is further a defence of materialism. This is held to be his major contribution to philosophy: both of these things are supposed to be good things. Unfortunately, Sellars’ case against empiricism and for materialism, the epigones notwithstanding, is simply wrong. Sellars’ case is developed in a central argument of his essay on “Empiricism and the Philosophy of Mind.”108 This is often celebrated as showing that standard empiricism is wrong and wrong-headed, and fundamentally erroneous in its arguments. It is celebrated too as showing how one can be a materialist, which is assumed to be a good thing. The celebrations, however, are premature. The essay is basically flawed. Unfortunately, the basic flaws in this essay and in its line of argument are rarely noted. The central argument is to the effect that all observation is “theory laden”, and as part of that thesis there is a further thesis – this is the holistic thesis that all concepts are “theory laden.” These theses constitute Sellars’ attack on what he called the “myth of the given,” that is, the thought common among empiricists – though less common than Sellars and his epigones have supposed – that our systematic knowledge of the world, both the external world and our inner life, rests on a foundation of incorrigible awareness, sensation on the one side and inner awareness on the other (the latter often misleadingly called introspection109). Sellars’ critique can be understood as resting on a distinction between the materials of sense and the inputs to the processes of reason. The inputs to the processes of reason are conceptualizations. We may react to sensory experiences with such conceptualizations, but the sensory experiences and the conceptualizations are not the same thing. Inasmuch as there is always the possibility that our conceptualizations are mistaken in some manner, there can be no foundation of indubitable truths about sensory experience such as the foundationalist imagines. Sellars’ critique of the “myth of the given” is seen by many to be significant for at least two reasons. First, it is thought to be connected with the idea that there is not simply a, that is, one mind-body problem, but in

43 fact two. One of these is what Sellars termed the “sensorium-body” problem. This is the problem of how it can be that sensory qualities can be in brains. The other mind-body problem is what Sellars called the “mindbody” problem proper. This is the problem of how thoughts can be in brains. The problematic nature of Sellars’ critique is recognized immediately: for it seems commonsensical that sensory qualities, e.g., the quality green that is given in sense when we perceive a tree, are not in our brains, that is, they are certainly not qualities of our brain states – the states of our brains are states of matter which is grey. It also seems commonsensical that our thoughts – our believings, our denyings, our supposings, our lovings, our hatings, our likings and dislikings, and so on – are also not states of matter. When I catch myself thinking, as Descartes caught himself thinking when he reached the cogito in his meditations, what I am thereupon aware of is an event which lacks the properties of material things like brains. But it is the point of Sellars’ argument to attack the idea that since the “given” is a “myth” we do not in the end have to take seriously the properties of these events as they are given to us: we can dismiss the given, and with some alacrity attribute other properties to these events, and identify them as being “really” brain states. The other reason that Sellars’ critique of the “myth of the given” is thought to be significant is the conceptual holism which is an essential ingredient of that critique. Sellars insists upon the distinction between conceptualizations and the sensory experiences to which they are responses. This is surely not problematic: there is such a distinction. But it is thought that this undermines the notion that our knowledge rests on firm foundations, that is, that it helps establish that the “given” is a “myth.” The reason behind this thought is the idea, that the conceptualizations with which the mind responds to sensory experience are sensitive not only to the intrinsic nature of that experience but also to the larger system of concepts to which they belong. Now, this too need not be problematic; there would appear to be no reason why the empiricist need reject it. For by itself it does not imply holism, which is incompatible with empiricism. To be sensitive to other parts of systematic conceptualizations of the world is not necessarily holistic.110 It is Sellars’ contention that it does involve a holistic account of meaning, however, and it is precisely here that the empiricist will disagree. The colour green, Sellars concludes, is not simply what it appears to be, it is not as it were logically self-contained, but derives its being, is what it is, through its connection to other things, e.g., through the fact that not only is

44 it not red but also that whatever is green is not red. In fact, Sellars argues, our concept of green through which we respond to green sensory inputs derives its meaning – its conceptual meaning – not at all from the entity to which it refers but rather wholly from its connections with other concepts, from such connections as the incompatibility relations like nothing that is green is red. The connection with the two mind-body problems now becomes clear. Our concept of a sensory quality is given its meaning not from its referent, the quality, but from its connections to other concepts. If we change those connections then we can have the concept green become the concept of an entity which is located in the brain. So qualia may be as we experience them, but we will conceptualize them as properties of brain states. So also our thoughts: with appropriate theories – conceptualizations – of our thoughts, these too will become entities which are located in our brains. The holism leads to materialism. And that of course is seen, nowadays, by many, especially Sellars’ students, to be a good thing. There is only one problem with this and that is that materialism is false. Whatever our theory or Sellars’ says about this, our experience tells us otherwise: qualia are not brain states and neither are our thoughts. And holism is false too: green is what it is in itself, and not through its relations to other things, and the concept green derives its cognitive content from what it refers to, namely the quality green, and not from the inferential connections it has to other concepts.111 We should turn, then, to Sellars’ claim that all observation is “theory laden.” Here we might begin with the difference between differential responsiveness, on the one hand, and observation, on the other: in what does this difference consist? what is added as it were to differential responsiveness to make it into observation? what is the difference between the capacity of a photocell that responds differentially to red things and our cognitive capacity to perceive red things? An intermediate case between these two might be Locke’s parrot which has been trained to say “This is red” when it is confronted with red things. Perhaps another intermediate case is that of a prelinguistic infant, if there be such. Now, awareness is never simply awareness of a bare thing or of a thingless property. Awareness is always of something as something, the object of awareness is always of something

45 complex, a fact and not an unattached simple: it is always in this sense propositional. If you wish, in awareness things come classified. More must be said,, however, before we can say awareness is observation. For, our friend the photocell also classifies stimuli, since it too responds differentially to them: connected to another mechanism, it lets out one beep if confronted with something red, two beeps if confronted with something green. In just this sense a chunk of magnetized iron classifies objects with which it is confronted, distinguishing them into classes of things which it attracts, things which it repels, and things with regard to which it is neutral. Indeed, an old nail can distinguish the environments in which it is located, separating them into two classes accordingly as it rusts in some and not in others. Sellars proposes that what distinguishes observation, and, more generally, perception is that it has a conceptual component not present in mere differential response. Perception, and so also observation, requires not just differentially responsive classification, but differentially responsive classification under concepts. The difference between a person and the photocell, when each is confronted by something red, is that the person does, and the photocell does not, have a concept, the concept red, which the person is differentially disposed to apply. The difference lies in the understanding which the person has and the photocell or the piece of iron or the nail or Locke’s parrot all lack, namely, the grasp, if you wish, of cognitive contents. It is this cognitive capacity, this ability to correctly apply concepts, this understanding that characterizes an observer, and distinguishes him or her from a photocell. But what is it to understand, to grasp a concept? Sellars argues that there is something basically linguistic to having or grasping a concept. When the person responds to the red thing with which he or she is confronted, that thing is not only distinguished by the person but also he or she responds by classifying it as red: the red thing evokes a verbal response in which the word ‘red’ (of a similar term in another language) is applied to the thing to which one is responding in respect of its quality to which one is also responding: “this is red” one says (perhaps to oneself) of that thing in virtue of that quality. So does the parrot. What distinguishes the person from the parrot is that for the person the word ‘red’ occurs as part of a broader linguistic framework. The word ‘red’ of course has connections to the world: it is the quality red of red things that evokes the application of the word ‘red’. These are world-word connections. Then there are word-word connections. There is, for example, the connection to ‘green’ by virtue of which one never applies

46 ‘green’ to something to which one has applied ‘red’. And there are wordworld connections. For example, there are the connections with speech acts, such as saying, perhaps overtly to others, “Lo, something red”, or there the connections with other acts, such a stopping when a warning signal is recognized as red. Sellars also refers to these as language-entry transitions, intra-linguistic transitions, and language-exit transitions. Among the speech acts which conclude the language-exit transitions are those in which a claim is endorsed, acts in which one expresses the judgment that the claim ought to be accepted as true. The claim will arise as the conclusion of a language-entry transition. The claim will be connected linguistically to other items, e.g., other predicates. Among these linguistic connections will be the norms about the endorsement of perceptual claims, that is, rules about when to judge that the perceptual claim ought to be accepted as picturing the way the world really is. The point is that the language-exit transition in which the claim is endorsed is made in conformity with these normative rules. It is therefore straightforwardly an action. Sellars sees this as crucial. He is questioning the notion that empirical knowledge has its foundation in knowledge of a set of particular facts, those that are privileged as being known in a direct awareness. He suggests that “if observation reports are construed as actions, if their correctness is interpreted as the correctness of an action, and if the authority of an observation report is construed as the fact that making it is ‘following a rule’ in the proper sense of this phrase, then we are face to face with givenness in its most straightforward form.” (“Empiricism and the Philosophy of Mind,” p. 296) For, he says, on these stipulations “one is committed to a stratum of authoritative non-verbal episodes (‘awareness’) the authority of which accrues to a super-structure of verbal actions, provided that the expressions occurring in these actions are properly used.” (p. 296) But there is no privileged stratum, the acceptance of which is justified solely by the fact that it is given in a direct awareness. Sellars carefully rejects the notion that a token, or utterance, of an expression, say “This is green,” for example, “expresses observational knowledge” if and only if the circumstances are of a certain kind, i.e. the utterance “is a manifestation of a tendency to produce overt or covert tokens of ‘This is green’ – given a certain set – if and only if a green object is being looked at in standard conditions” (p. 297). This won’t do, he says, because “it is the knowledge or belief that the circumstances are of a certain kind, and not the mere fact that they are of this kind, which contributes to bringing about the action” of an observation

47 report (p. 296). One’s grasp of the concept red (or the concept green) which is used to characterize the situation to which one is differentially responding is more than the mere capacity to respond differentially to red (or to green) things, in that you have the capacity to use such a concept as part of a linguistic network, a conceptual scheme if you wish. Indeed, this is what makes it a concept. In using a concept one knows these connections. In particular, one knows what the inferential significance of applying the concept red is – one knows what follows from something’s being red, for instance that it is coloured, that it is not green, that it is not a prime number or an electron or a chair, and so on. The response, essentially linguistic, that one is differentially disposed to produce is inferentially articulated. It occupies a place in a network. The world, the object to which one is responding, provides the occasion for applying the concept. This occasion can be in itself a reason for applying the concept. But for the occasion to be a justifying reason one needs to place the claim in the context of norms – cognitive norms – that justify endorsing the claim elicited on that occasion. There can, however, be other things which justify endorsing perceptual claims. The network can provide other reasons justifying endorsing such claims. For, in terms of the network there can be other sorts of reasons that can be given for applying (or not applying) the concept in that situation. Inference, of course, like endorsement, is a normative affair. To say what is a reason for what is to say something about what judgments one ought to make, not necessarily what judgments one does in fact make. Sellars develops these notions with some care. His is, for example, not a causally reductive account of the fundamental semantic relations of inference and incompatibility which ignore the normative aspect. But it is a causal account, for he argues that among the regularities that describe our linguistic uses are those in which characteristically normative judgments about what ought to be, bring it about that persons conform (more or less) to what ought to be. There is an apparent circularity in appealing to our “knowing” what follows (i.e. what it would be correct to infer) from what in order to know or grasp a concept: in order to know the meaning of a concept one must know the meaning of rules which already involve that concept. But the circularity is only apparent, since the knowledge in question is knowhow, a practical ability to discriminate, that is, differentially respond to, good and bad inferences. The responses that you are differentially disposed to produce in this case will involve the application of further concepts, whose articulation as parts of the network must also be mastered in prac-

48 tice. The understanding that distinguishes the person’s perception from the photocell’s differential responsiveness and the parrot's, is an inferentially articulated normative status. Thus far, there is little with which an empiricist need disagree, Sellars and his epigones notwithstanding. There might be some empiricists who do disagree with some parts of this account of our perceptual judgments. Maybe such contrary views were held by the early Wittgenstein. But that should not count for much. In fact, most empiricists would not disagree with Sellars’ account. These ideas are not uncommon. One finds these notions in, for example, Russell’s The Analysis of Mind and in his later Inquiry into Meaning and Truth, both of which are likely sources for Sellars’ thought on these issues. But one can find the core notions already in Locke’s Essay concerning Human Understanding, where this is the way the conceptualization of the person is distinguished from that of the parrot, and in Hume’s Treatise of Human Nature, which makes clear the role of language in perception and makes clear, too, that these connections are not merely a matter of (causal) regularities but are also normative.112 Related to these claims of Sellars about the connection between concepts articulated in a linguistic network and perception is an account of the relation between perceiving how things are and how things seem to be. It might be tempting to say in response to this way of making out the difference between us and the parrot that there is in common at least the givenness of the qualities of things. The parrot does not grasp the concept of red, because its responses do not for it play a role in a normative network involving, for example, incompatibility relations (as opposed to the nonnormative incompatibility of, for instance, being unable to respond simultaneously to something as red and as green). Nonetheless there is still some kind of perception that the parrot has, in common with persons, and that the photocell does not. At least the parrot can in a way grasp at least apparent redness as a recurring feature, a recurring way in which things are present in its awareness of things in the world. As Sellars sees it, this then becomes the starting point for the empiricist understanding of perception. There is a distinction between how things are and how things seem to be. And we can be wrong about how things are, but not about how things seem. The awareness of the bare seemings of things is what we have in common with the parrot. Sellars’ empiricist concludes that how things seem constitutes that which we know or understand immediately, without inferences, without involving connections to the network, inferences that go beyond what is simply given in awareness, in-

49 ferences which, precisely because they go beyond what is given, might go wrong. Sellars’ empiricist concluded that such seemings could serve as a foundation for knowledge and understanding, because grasp of seemings presupposes no non-immediate or inferential knowledge or understanding of parts of the conceptual network. Sellars developed a view which he took to be contrary to that of his empiricist: this is the view that perception of how things are requires both reliable differential responsive dispositions and the capacity to also respond with a judgement in which a claim is endorsed, that is, a judgment which involves the adoption of a normative attitude toward a claim in which an inferentially articulated concept is applied to the thing evoking the differential response. There is the thing and its quality. There is a response to that thing qua so qualified – “this is red” – this response is a sentence which represents the state of affairs to which one is differentially responding. In this sentence a concept which is inferentially connected to a whole body of other concepts is applied to the thing. But there is further an endorsement of that claim, to the effect that the claim ought to be accepted as true – “this (I assure you) is (that is, really is) red”, where the endorsement derives from the network in which the concept is located inferentially. Perception normally involves a claim that is endorsed – the ‘is’ in the spoken claim that “this is red” not only represents the linking of the quality to the thing qualified, it also expresses one’s endorsement of the claim as correct or true. Perceptual judgments about how things look or appear to be are parasitic as it were upon judgments about how things are. Learning to use correctly the concepts applied in reports of how things look or seem to be involves learning to withhold the usual endorsements of the normal response. Thus if I have good reason to believe that the thing in my hand is not red, but am responsively inclined (by my training and internal wiring) to call it red, I can express the responsive disposition without endorsing it by saying that it looks red. The extent of my willingness to endorse such dispositions, for instance in a case of bad light or brief exposure, may vary. This explains the possibility of merely generic, or qualitative lookings (e.g. how a polygon can look merely many-sided, without there being any definite number of sides it looks to have), which are hard to assimilate on other approaches. Again, the empiricist need not disagree. But we shall come back to the example, of which Sellars makes much. It is, however, perhaps worth pointing out how Sellars, again not im-

50 plausibly, uses this account of ‘seems’ judgements or judgments of how things appear as opposed to how things are. Specifically, he suggests that this accounts for the apparent incorrigibility of judgments of how things seem. According to Sellars’ story, the incorrigibility of ‘seems’ judgements simply reflects the fact that they withhold the usual endorsements of ‘is’ judgements. Where there is no endorsement, he suggests, there can be no mistake. The claim of the perceptual judgment is made and then immediately as it were withdrawn. Since the endorsement is immediately withdrawn the accusation that one is wrong is forestalled: since one has made no claim about the way the world is, or at least, endorsed no such claim, one cannot be faulted for being wrong about the world. So the ‘seems’ judgment cannot be erroneous. At the same time, however, Sellars points out, incorrigibility, or rather, incorrigibility in this one sense is revealed as something rather trivial, and certainly unsuited to the role of being a foundation, a secure foundation, for empirical knowledge. On this account of the language of “seems” or “appears,” one must be able to use the inferential connections among concepts (and so must endorse some inferences) even to grasp the concept “looks red”, that is, one must have the linguistic know-how regarding inferences in the context of the endorsements one is withholding. Thus a ‘seems’ judgement is not an immediate cognition, that is, non-inferential. Perception of how things look, as of how things are, requires mastery of a network of inferentially connected concepts. The lesson Sellars draws is this: One can not have a language or conceptual scheme in which all judgements can only be arrived at noninferentially. There are perceptual judgments that are non-inferential, but they must involve concepts which are part of a network which can provide inferences that will justify those judgments if challenged. It is possible to have a language or conceptual scheme in which all perceptual judgments can be arrived at both non-inferentially (on some occasions) and as conclusions of inferences (on other occasions). Take a dot the colour of which is scarlet. “This [the dot] is red” may express either a commitment arrived at non-inferentially by observation (perception) or one inferred from one expressed by “this is scarlet.” The same claim can be made either as a report or as the conclusion of an inference. The content does not differ in the two cases. Since the claim that is a report must also be one that could be arrived at inferentially, the report must itself involve concepts so connected in the network that an inference with that report as a conclusion must also be possible. In that sense, one cannot have a language in which there are re-

51 ports which are completely non-inferential: all observation is theory-laden. But again one must point out that the empiricist need not disagree. There is only one thing which is odd. That is this: why would anyone, e.g., Sellars, think that the empiricist would build on such a shaky foundation? Sellars’ empiricist is in fact a straw person: it is hard to think of a single philosopher who would fit his description of an empiricist. Why should an empiricist not accept his view of observation as involving the application of concepts inferentially connected to other concepts in a linguistic network and as involving acts of endorsing, where such endorsement presupposes various cognitive norms? Why would Locke disagree? Why would Hume? Why would Russell? At any rate, Sellars draws his conclusion: So much the worse for empiricism. Sellars’ critique can be understood as resting on a distinction between two sorts of inputs into our perceptual judgments, that is, those among our judgments of sense which we count as observations. There are, on the one hand, the materials of sense and there are, on the other hand, the inputs from language, and, within that, from the concepts we use and their involvement with the inferential norms that constitute the processes of reason. The inputs from the processes of reason are conceptualizations. We may react to sensory experiences by applying our concepts, that is, we may react by conceptualizing the experience in certain ways, e.g., as red. But the sensory experiences and the conceptualizations are not the same thing. Inasmuch as there is always the possibility that our conceptualizations are mistaken in some manner, there can be no foundation of indubitable truths about sensory experience such as the foundationalist imagines. We may grant that Sellars has, for the reasons he gives, shown that a certain sort of foundationalist empiricist is off on the wrong track. But, we must ask, is there in this account of observation anything that a non-foundationalist empiricist must reject? There seems not to be ay such thing. So Sellars is thus far wrong to think he has shown empiricism to be false. He is alas, as we have suggested, too hasty. There is an important point where he does come into conflict with empiricism. He argues that the observation claim that (a) This is red as elicited directly by a sensory context might well not be endorsed, and that its contradictory be endorsed instead. For suppose we knew that the lighting conditions were not normal, that the object, the “this”, was being viewed in a red light, and that we knew that the thing was in fact green. So

52 we have, for whatever reasons, but assumed to be good, endorsed (b) This is green Furthermore, our “conceptual network” of which the concept green is a part tells us that (c) Whatever is green is not red Since this is the way the world is, red and green are incompatible, we ought so to speak: (c) is normative. Endorsing (b) and inferring via (c), we conclude by endorsing that after all (d) This is not red Now, there is a difficulty here. Why not accept both (a) and (b) and maintain that we have a counterexample to (c)? The point comes out perhaps more clearly with another example. As we have seen, Sellars rejects the notion that a mere utterance of an expression, such as (*) This is green “expresses observational knowledge” just in case that the circumstances which evoke the utterance are of a certain kind or sort, i.e., just in case that the utterance “is a manifestation of a tendency to produce overt or covert tokens of ‘This is green’ – given a certain set – if and only if a green object is being looked at in standard conditions” (“Empiricism and the Philosophy of Mind,” p. 297). This won’t do, he says, because “it is the knowledge or belief that the circumstances are of a certain kind, and not the mere fact that they are of this kind, which contributes to bringing about the action” of an observation report (p. 296). He then argues that for an evoked utterance of a sentence like (*) by someone – Sellars calls this person “Hypothetical Q. Jones” – to count as “expressing observational knowledge,” not only “must it be a symptom or sign of the presence of a green object in standard conditions, [call this condition (I)], but also [call this condition (II)] the perceiver, that is, Jones, must know that “tokens [utterances, perhaps internal] of ‘This is green’ are symptoms of the presence of green objects in conditions which are standard for visual perception” (pp. 297-298). He comments that The point is specifically that observational knowledge of any particular fact, e.g. that this is green, presupposes that one knows general facts of the form X is a reliable symptom of Y. And to admit this requires an abandonment of the traditional empiricist idea that observational knowledge ‘stands on its own feet’. (p.298)

Notice that he still hasn’t given us any grounds as to why he thinks an em-

53 piricist cannot agree with him. Now, since Sellars is proposing that knowledge of particular facts presupposes knowledge of general facts, such as that in (II), one can ask about the status of the knowledge of these general facts. Sellars does say that with respect to knowledge itself ...in characterizing an episode or a state as that of knowing, we are not giving an empirical description of that episode or state; we are placing it in the logical space of reasons, of justifying and being able to justify what one says. (pp. 298299)

It would seem that Jones can justify his knowledge of the general fact in (II), that is, the fact that utterances of ‘This is green’ are symptoms of the presence of green objects in conditions which are standard for visual perception

by citing prior particular facts as evidence, as inductive reasons (cf. p. 299). Thus we might have (&)

Utterance A of “This is green” at time t by Jones was accompanied by the presence of a green object in standard conditions of perception

and (&&) Utterance B of “This is green” at time t1 by Jones was accompanied by the presence of a green object in standard conditions of perception

and so on. There is a regress looming here, since these statements will count as evidence only if yet further general facts are known. Moreover, the regress is vicious. Sellars is sensitive to this point. For, he admits that ...does it not tell us that observational knowledge at time t presupposes knowledge of the form X is a reliable symptom of Y, which presupposes prior observational knowledge, which presupposes other knowledge of the form X is a reliable symptom of Y, which presupposes still other and prior observational knowledge, and so on. (p. 298)

But those who would think this are simply mistaken, he argues. He explicitly denies that any regress is involved. He makes the point that, while the correctness of Jones’ statement that he now knows a general fact of the

54 form X is a reliable symptom of Y [such as that in (II) for example] ...requires that Jones could now cite prior particular facts as evidence for the idea that these utterances are reliable indicators, [this does not lead to a regress because] it requires only that it is correct to say that Jones now knows, thus remembers, that these particular facts did obtain. It does not require that it be correct to say that at the time these facts did obtain. And the regress disappears. (p 299)

Now, Sellars has indeed avoided one regress here, a “temporal regress” as it were. What Selllars is saying is that the observational claim can be evoked in a differential response to the world and endorsed without that requiring an inference. But it can also be challenged: it might be wrong. If challenged, then further endorsement is necessary, this time by inference. Such an inference might well involve a general fact like that mentioned in condition (II). If a challenge is to be met, then our observer Jones must know this fact or this sort of fact. Even if it is not needed and not used to defend the endorsement or to criticize it, that knowledge must be available in the context, there to be used if needed. So, even if the observational endorsement is not a matter of inference, the claim must, even when made without inference, be embedded in an inferential network of reasons. One cannot have a language or conceptual scheme in which all judgements can only be arrived at non-inferentially. There are perceptual judgments that are non-inferential, but they must involve concepts which are part of a network which can provide inferences that will justify those judgments if challenged. This is why the concepts used in the claim must already be located inferentially in a network of concepts. In that sense, one cannot have a language in which there are reports which are non-inferential: all observation is theory-laden. Not only can one not have a language in which all observation reports are non-theory-laden, one cannot have a language in which there are any observations which are not theory-laden. So Sellars avoids the “temporal regress.” But the question still remains, how can Jones justify his knowing of the particular facts (&), (&&), etc., which in turn is taken to justify his knowing the general fact in (II). In other words, there is also logical regress which Sellars seems to ignore. Of course, the straw empiricist with whom Sellars is disputing might suggest that the required knowledge of particular facts is self-justifying, in that it is expressed in ultimate and authoritative observation reports, utterances which merit endorsement simply by virtue of the fact that they are evoked. But Sellars already has argued that there is no knowledge of par-

55 ticular facts that in this sense “stands on its own feet,” and has insisted “knowledge of any particular fact ... presupposes that one knows general facts of the form X is a reliable symptom of Y”. (p. 298) Robert Grimm has suggested that Sellars is trapped113: either Sellars falls back into what he [Sellars] calls the “Myth of the Given” or the justification of observation reports is circular or it involves a vicious regress. Sellars has not improved upon his empiricist. Grimm argues that, on the one hand, Sellars cannot have Jones justify his knowledge of the particular facts (&), (&&), and so on, by appealing to his knowledge of the general fact mentioned in (II), since this would involve him in circularity: Jones would be justifying his knowledge of the particular facts by his knowledge of the general fact in (II), which knowledge he had initially set out to justify by his knowledge of the particular facts now in question. Grimm argues that, on the other hand, if Sellars claims that Jones’ knowledge of these particular facts (&), (&&), etc., presupposes other knowledge of the form X is a reliable symptom of Y [other than the general fact in (II)], then he is involved in the “logical regress” that we have noted: the acceptance of this general fact then presupposes logically prior knowledge of other particular facts as justification, which in turn presuppose logically prior knowledge of other general facts of the form “X is a reliable symptom of Y” as justification, and so on. Grimm concludes that, since Jones, first, cannot justify his knowledge of particular facts (&), (&&), and so on as being self-authenticating without becoming embroiled in the so-called “Myth of the Given”, and since, second, he cannot justify his knowledge of these particular facts by his knowledge of the general fact in (II) without becoming involved in circularity, and since, third, he will fall into a vicious logical regress if he tries to justify his knowledge of these particular facts by knowledge of other general facts of the form “X is a reliable symptom of Y”, and finally, since these seem to be the only alternatives – therefore, it would seem that Sellars’ proposed alternative is no improvement at all over the “Myth of the Given” that he is concerned to refute. However, Grimm’s conclusion is too hasty. There are alternatives that Grimm has not considered. Sellars in fact has a way out. And there is yet another way. Sellars’ way is contrary to what empiricism requires. The other way out is not. Sellars’ way out is this. Grimm generates what is either a circle or a vicious logical regress in

56 justification by focussing on the evidence justifying the assertion of the required generality in (II). He takes it that the evidence that justifies endorsing the generality is based on a process of induction by simple enumeration. On that assumption, he is undoubtedly correct. It is, however, an assumption which is open to dispute. Sellars in particular would dispute the assumption. Sellars in fact argues that all generalities which are laws derive their endorsement from (material) rules of inference. The generalities like (2) or that in (II) are true ex vi terminorum, true by definition if you wish, but in any case true a priori. It is, to be sure, a pragmatic conception of the a priori, as C. I. Lewis would put it. But it is still a priori. The endorsement of the generality is a priori and not a matter of simple induction from instances. Since Grimm’s objection turns on assuming the latter, the objection does not apply: Sellars escapes by offering an alternative that Grimm does not consider. Sellars has put his point in this way: ...the material transformation rules determine the descriptive meaning of the expressions of a language within the framework established by its logical transformation rules. In other words, where ‘Ra’ is P-derivable [derivable by a material rule of inference] from ‘Na’ (in modal language Na necessitates Ra), it is as correct to say that ‘Na e ψa’ is true by virtue of the meanings of ‘N’ and ‘R’, as it is to say this where ‘Ra’ is L-derivable [derivable by the usual rules of logic] from ‘Na’. In traditional language, the “content” of concepts as well as their logical “form” is determined by rules of the Understanding.114

This allows him to escape from the Grimm dilemma (or trilemma). Thus, Sellars’ critique of the “Myth of the Given” is bound up with his conceptual holism.115 The reason why the distinction between conceptualizations and the sensory experiences to which they respond undermines the notion that our knowledge rests on firm foundations is that the conceptualizations with which the mind responds to sensory experience are sensitive not only to the intrinsic nature of that experience but also to the larger system of concepts to which they belong. But for Sellars this larger system finds its justification as a set of “P-rules”, material rules of transformation that define the concepts.116 The concepts are inferentially articulated in a network, but for Sellars in this network the propositions are all true ex vi terminorum. The network is one thing: who would deny it? who would deny that concepts are bound up with other concepts through intra-linguistic connec-

57 tions? To suggest that the connections forming the network all hold ex vi terminorum, so that they define the concepts that they contain, is quite another thing. The empiricist might well grant Sellars the first, but he or she must reject the second: the holism is incompatible with empiricism. Sellars, however, is convinced he needs it in order to provide a ground for escaping the Grimm dilemma. He needs a ground other than induction by simple enumeration for affirming or endorsing the connections in the network. That those connections hold ex vi terminorum provides that ground. There are, however, problems with this way of avoiding the Grimm dilemma. Sellars may escape between the horns, but this new alternative is in fact disastrous. It is the holism which is the problem. Sellars contrasts his position with the traditional position in which “the form of a concept is determined by ‘logical rules’ while the content is ‘derived from experience’,” a position which “embodies a radical misinterpretation of the manner in which the ‘manifold of sense’ contributes to the shaping of the conceptual apparatus ‘applied’ to the manifold in the process of cognition.” He continues: The contribution [of sense experience] does not consist in providing plums for Jack Horner. There is nothing to a conceptual apparatus that isn’t determined by its rules, and there is no such thing as choosing these rules to conform to antecedently apprehended universals and connexions, for the “apprehension of universals and connexions” is already the use of a conceptual frame, and as such presupposes the rules in question.

So when Sellars holds that laws and generalities are all true ex vi terminorum, he means to assert a very strong thesis. This is the thesis that the cognitive content of a concept is given by the rules of inference in the network in which it occurs. The cognitive content of a concept is exhausted by the intra-linguistic connections in the network. In particular, the worldword or language-entry connections do not contribute to the cognitively relevant meaning of the concept. That is, reference does not contribute to the cognitive meaning of a concept. The role of the environment is not to determine the meanings of concepts. Rather, The role of the given [the environment which evokes observation claims] is ... to be compared to the role of the environment in the evolution of species ...

though Sellars thinks enough of the human mind to add that

58 ...it would be misleading to say that the apparent teleology whereby men “shape their concepts to conform with reality” is as illusory as the teleology of the giraffe’s lengthening neck. After all, it is characteristic of modern science to produce deliberately mutant conceptual structures with which to challenge the world. For primitive thought the analogy is much less misleading.117

I don’t want to pursue Sellars’ own views on the role of the environment in bringing about conceptual change, though I will say that it seems rather odd indeed, at least if I grasp the point of the evolution metaphor and the point about deliberately producing mutants. For he seems to be suggesting in the “natural selection” metaphor that changes in the network are not a matter of reasons but come from outside the realm of reasons: it seems to be that nature selects the realm of reasons, and reason has nothing to do with this selection. I would prefer an account of reason in which reason, in the light of experience, adjusted itself to the world, to that of Sellars, in which the world seizes reason and without consulting it, forces it into some sort of conformity: I like reason to be involved in the way the network of reasons adjusts itself to the world as we experience it. Be that as it may, there are deeper problems with Selllars’ account of the meaning of concepts. Sellars holds that it is part of the meaning of green that (c) Whatever is green is not red and that laws like those in (II) tokens [utterances, perhaps internal] of ‘This is green’ are symptoms of the presence of green objects in conditions which are standard for visual perception

hold for this concept. The colour green, Sellars is holding, is not simply what it appears to be, it is not as it were logically self-contained, but derives its being, is what it is, through its connection to other things, e.g., through the fact that not only is it not red but also that whatever is green is not red. This is where the empiricist disagrees. For such a one, we are in acquainted in experience with green as a quality of things. This quality is in fact wholly presented. When presented with it, there is nothing in that presentation through which I am also somehow presented with the fact (b) that

59 the presence of green is incompatible with the presence of red, nor when I am presented with green am I presented with a fact like that in (II) which guarantees that when I am presented with the state of affairs that something is green then that fact of presentation guarantees that the state of affairs presented really does exist, or is itself a fact. So the quality green with which I am presented is logically self-contained, it is what it is in itself and not by virtue of it connections to other qualities and things. Whatever is is: this may be a trivial proposition. But it is what we must insist upon. For it really is true and its truth in incompatible with Sellars’ holism (or that of any other idealist). This is how the world is presented to us, and it is how we must take that world. That is, the things of the world are presented to us, and they are presented as logically selfcontained: that is how we must take them. Or so the empiricist, appealing to PA, must insist. So much the worse for Sellars: what he is implying the world is like is wrong, the world is simply not as he implies it to be. On the basis of an appeal to PA, then, Sellars’ conceptual holism is to be rejected. In that case, however, we seem after all not to have escaped from the Grimm dilemma: Sellars’ way of going between the horns turns out to be unacceptable. Unacceptable because the ground on which he found his way through, that is, the holism, is itself unacceptable. What might the empiricist say? It was in fact already said by Hume. The relevant points we have already noted briefly in part (III) of the main body of this essay. Sellars argues that the appeal to P-rules or material rules of inference which make the intra-linguistic connections of the network all hold ex vi terminorum is necessary if one is to have subjunctive conditionals. ... P-rules are indispensable to any language which permits the formulation of material subjunctive conditionals...P-rules are essential to any language that contains non-logical or descriptive terms.118

Hume has argued otherwise.119 He makes the case against objective necessary connections of the sort that Sellars is defending, though of course Hume’s opposition is the tradition of the substance philosophy, in both its Aristotelian and rationalist versions. For the substance tradition, certainty of inferences from samples to populations is achieved through a grasp of the essences or natures of things.120 Causal activity of a thing in accordance with its essence guarantees that always in similar circumstances it would again behave in similar ways: the objective connections that yield the ne-

60 cessity are part of the real definition of the essence of the thing changing. Hume of course argues vigorously against this view121, adopting arguments from Malebranche.122 Indeed, he almost quotes Malebranche verbatim.123 Malebranche argued that there are no objective necessary connections other than those of God’s causal activity: “les causes naturelles ne sont point de véritable causes. ... Il n’y a donc que Dieu qui soit véritable cause, et qui ait véritablement le puissance de mouvoir les corps”.124 Hume refers to this “Cartesian” doctrine,125 and argues that it is untenable (Treatise, p.160): if, as these philosophers hold, there are no objective necessary connections among bodies because we have no impression of such a connection, then neither do we have any idea of it – ideas being derived from impressions – and therefore we cannot have an idea of God that includes within it the idea of causal power or activity. “We never therefore have any idea of power”.126 Objectively considered, then, there is no distinction between an accidental generalization and a causal generalization: both are of the simple form “all A are B”. This is Hume’s first definition of ‘cause’: “An object precedent and contiguous to another, and where all the objects resembling the former are plac’d in like relations of precedency and contiguity to those objects, that resemble the latter”.127 This is “cause” defined as a philosophical relation.128 But there is a distinction between post hoc and propter hoc. As Hume says, “there is a NECESSARY CONNEXION to be taken into consideration”.129 In the substance tradition the distinction is grounded objectively: connections propter hoc are a matter of objective necessities grounded in the real definitions of the essences of things, connections post hoc are not. So also in Sellars: connections propter hoc are a matter of objective necessities grounded in conceptual ties that hold ex vi terminorum, connections post hoc are not. But, as Hume has argued, there are no such objective necessary connections: a thing is what it is, logically separable from other things. This is the way the world is, as we are acquainted with it. The distinction between post hoc and propter hoc therefore cannot be objective; it is rather, Hume argues, subjective. This is the thrust of the second definition of “cause”, which asserts that a generalization is causal just in case we are prepared to use it in counterfactual assertions and in predictions: “An object precedent and contiguous to another, and so united with it in the imagination, that the idea of the one determines the mind to form the idea of the other, and the impression of the one to form a more lively idea of the other”.130 This is “cause” defined as a natural relation.131 The element of necessary connection, then, which is an in-

61 gredient in the idea of cause, is the propensity of the mind to make inferences in the case of causal connections which is absent in the case of accidental generalities.132 (Such determination is itself a case of causal determination, and is also, of course, subject to the Humean analysis.133) What the second definition mentions is, on the one hand, prediction – the mind is so determined that the impression of the one [leads it] to form a more lively idea of the other–, and, on the other hand, the assertion of subjunctive conditionals – the mind is such that the idea of the one determines the mind to form the idea of the other. Where one has a connection of the sort (X) All A are B, then A and B are logically separable – this yields the first definition – objectively there is no necessary connection, nothing which makes such a proposition true ex vi terminorum. But nonetheless, some propositions of this sort we treat differently – we use them to make predictions and to support the assertion of subjunctive conditionals – where there are others we do not treat in this way. The former are by virtue of being so treated propter hoc, the latter are by virtue of not being so treated are post hoc. There is a necessity to the former that is absent from the latter. But the necessity lies in how they are used, and not in any real definition or set of propositions true ex vi terminorum. Laws like (b) and like those of (II) are, of course, precisely like (A). They have the logical form of matter-of-fact empirical generalities. But they are necessary in the Humean sense in that they are used to support predictions and the assertion of subjunctive conditionals. This provides us with another way to escape between the horns of the Grimm dilemma. Here was our example. Sellars argues that the observation claim that (a) This is red as elicited directly by a sensory context might well not be endorsed, and that its contradictory be endorsed instead. For suppose we knew that the lighting conditions were not normal, that the object, the “this”, was being viewed in a red light, and that we knew that the thing was in fact green. So we have, for whatever reasons, assumed to be good reasons, endorsed (b) This is green Furthermore, our “conceptual network” of which the concept green is a part tells us that (c) Whatever is green is not red Since this is the way the world is, red and green are incompatible, then we

62 ought so to speak: (c) is factually true but is therefore also normative. Endorsing (b) and inferring via (c), we conclude by endorsing that after all (d) This is not red Now, there is a difficulty here. Why not accept both (a) and (b) and maintain that we have a counterexample to (c)? Sellars’ argument is, in effect, that we don’t do this, but rather in accepting (b) we reject (a) because the generality (c) is a necessary truth: it is true by definition of the concepts which it contains, it is true ex vi terminorum. But there is no need for Sellars’ move to holism. One can instead accept the Humean move in which the connection embodied in (c) is indeed necessary, and in this sense protected against falsification, but which makes the necessity a matter of its psychological context. It is necessary in the sense that it is used to predict and to justify the assertion of contrary-to-fact or subjunctive conditionals. But logically and ontologically speaking, the concepts are logically separable and the things and qualities and relations to which they refer are ontologically separable. The cognitive content of a concept is given by the world-word connections, the semantic rules of reference: the meaning of a concept, at least for those at the observation level, is given by its referent. Since the referents are ontologically separable, the concepts are logically separable. And the general fact is, therefore, contrary to Sellars, a contingency. So we are not after all forced into Sellars’ holism as a means of escaping the Grimm dilemma: there is yet another alternative, the Humean, that allows an escape. This alternative, unlike Sellars’, is throughout empiricist: it allows that the language we are using is logical atomistic in form. At the same time it allows, with Sellars, that perceptual claims are reckoned to be observations by virtue of their occurring in a network that inferentially articulates the concepts appearing in those judgments. In other words, we can have both the logical atomism which is essential to empiricism, the logical and ontological separability of things and their qualities and relations, and also the non-foundationalist account of observation. And we see that Sellars’ inference from the latter to the holism that denies the former is simply wrong.

63 Appendix Two Things Seen and the Seeing of Them We have argued that there are sensings and there are perceivings. Sensings have as their objects sense data or sense impressions. We have noted that these sense impressions are part of a larger class of sensibilia. Sensibilia are simply those ways the world is in its sensible appearances. We see a round sphere. Given to us in sense experience is the whiteness of the side that is present to us. But there is another side which is, say, brown. This sensible appearance is there, ready at it were to be presented to anyone who views the sphere from the relevant side. We, however, who are not looking at that side, are not given this sensible appearance, we are given in sense only the whiteness and the shape of the side to which we are responding, it is this sensible material and not that other to which we are giving a discriminating response. There are, then, many sensibilia, many ways in which there are appearances of things. Sense data simply are sensibilia which happen to be sensed. But there are many, indeed, countless sensibilia which are unsensed. The sphere itself has many sensible appearances, some sensed, many unsensed. They are visual and tactile mainly, but also olfactory and gustatory, and perhaps even auditory (that crunch sound when you bite into it). In fact, the sphere is what it appears to be: there is no radical distinction between the sphere as it appears and the sphere in some inner reality. There is no distinction between, say, its substance which is its inner reality and the appearances which that substance presents to us. For, there is no such inner reality, no such thing as a substance which lies beyond or behind the appearances of the thing. Such substances which lie beyond or behind the way the world appears to be are excluded by the Principle of Acquaintance for any philosophy that purports to be empiricist. The appearances of the thing are its reality: the thing simply is a bundle of sensibilia – a structured bundle, to be sure, but bundle nonetheless. It is simply the way or ways in which it appears to us: that is its reality. Notice that rainbows and shadows and other ephemera and insubstantial things of this world also do appear to us in various ways. They too are bundles of appearances and every bit as real as the medium sized dry goods that we so often, perhaps following Aristotle, take to form the substance of the world. I like this sort of world. I like rainbows to be just as real as straight

64 sticks. And when the straight stick is put in water and, as one says (Professor Sellars has reminded us), appears to be bent, I like it that the bent shape that I sense is real, and just as real at the straight shape that is presented to me when the stick is out of water. Not only do I like the world to contain those shapes but I can explain how it is that they hang together: this is the job of Snell’s Law for the refraction of light as it passes from one medium to another. To be sure, I do also like medium sized dry goods, tables, chairs, trees, and oysters – I write on the table, sit on the chair, rub the bark of the tree and enjoy its shade, and there is a smoky bar where I eat oysters. I like but also step on shadows, chase rainbows, enjoy the shade, and squint through the smokiness. It is just that there are many sensibilia, many ways in which the world appears to us, and they are all equal, none of them privileged The objects of perception are many. They are tables and chairs, rainbows and shadows, creepy corners and well lit squares, sticks in water and sticks out of water, fleas and flea bites, the sun and sunburn, the soft touch of a lover’s hand. But the objects of perception are bundles of sensibilia, structured or ordered in many different ways – they are material objects let us say, but understanding that term in a sufficiently broad way. The object perceived has many parts, all of them sensible, which are, however, not sensed: the bundle which it is, is a bundle that contains parts that are unsensed. In sensing, I sense, say, a red patch. I say, “this is red”. What I sense is the patch and its quality red, and maybe its spatial location relative to other patches which I also sense. But these objects which I sense, though they may be bulgy, have no back to them – they are two-dimensional: I cannot walk around and sense the other side of the patch that I am now sensing: there is no other side. In perception, in contrast, there is another side to things – to many things anyway, to apples and chairs and the moon but not to rainbows or shadows. When I am aware of an apple or some other three-dimensional object to be met with in space,134 I am aware of the sensible appearances of a thing, but these are now located as the appearances of a thing located in 3-dimensional space and at a distance from me. When I say in perception, “this is red” I am locating the patch which I sense and its quality as part of a larger bundle. In perception there is always the material thing which is appearing to me. In perception I am aware not just of a sensible quality but also that the quality is part of a greater whole most of which is not presented to me. The sensible patch I sense in perception presents to me the tree. It is not just “this [the patch] is red” as

65 in sensing, it is that “this [the red patch] is a tree [the redness of which I am sensing]” – in perception I perceive as being of a tree the red patch which I sense, the red patch presents to me not only itself, as in sensation, but the tree as a whole; it represents to me all those parts which are not given in sense and presents them as so structured that they are of a tree. So I may know the tree, the material object, which is a bundle of sensibilia, in two ways. One way is by inference – by inference from the presented sense data to other unpresented or at least unsensed parts of the bundle of which they themselves are parts. Such an inference from sensed parts to unsensed parts is essentially inductive. The other way of knowing the tree, the material object, the tree, is by perception. In perception the object perceived is indeed a bundle, and since most parts of the bundle are not sensed, it shares with induction the fact that it goes beyond what is immediately present in sensation. In this sense, perception is like induction: it shares in the same logically hazardous move from what is sensed to unsensed parts. But inference involves a movement of thought, from premise if you wish to conclusion. In perception there is a unity: you do not infer that this is a tree; it is simply given to you directly, and without inference, that this is a tree. The material object which is the object of perception can, in other words, be given in two ways, one by inference, where it is not wholly presented, and one directly, wholly presented in a single unified act of perceiving. And when it is perceived, it is perceived as directly as is a sensedatum in an act of sensing. For some time it has been fashionable to challenge empiricism by arguing that it does not do justice to our awareness of the world as something “thick”, or at least as containing things which are “thick” The world of the empiricist is a world of ephemera, the world itself is hardly that, it is thick. So they say. Roderick Firth has explored this topic in detail.135 His discussion is worth looking at. He notes that the thesis that material objects can be observed as directly as sense-data has been used by some as grounds for criticizing certain aspects of empiricism. He cites John Wild, then in his realist phase, as one such philosopher.136 Wild suggests that what is actually given in perception is a “world of things”. Wild (and following him, Firth) quotes with approval a statement of C. I. Lewis which makes the point that the world given in perception is “thick”: “it is indeed the thick experience of the world of things ... which constitutes the datum for philosophical reflec-

66 tion”; in fact, Lewis goes on, “we do not see patches of colour, but trees and houses [Lewis misses rainbows and shadows]; we hear not indescribable sounds, but voices and violins”. Wild approves but then proceeds to criticize Lewis for giving up this “classic view of the given” (classic because Wild takes it to derive from Aristotle) for the more restricted one of Berkeley and other modern empiricists. Modern empiricism, Wild proposes, “abandons the aim of classic philosophy to describe the thick experience of the world of things as it is given. Instead of this, it singles out a certain portion of the given as peculiarly accessible or given in some special sense”. This special portion of perceptual experience is the sensory core, the sense-datum which is present when one perceives a material object. This alone, on the modern view, according to Wild, becomes what is given in experience. To be sure, there is in perception a sensory core that is in fact given: what Wild objects to is the claim that “the immediately given alone is given.” Firth also cites Hans Reichenbach who argues in his Experience and Prediction that material objects are immediately given in perception and has used this as an argument against “positivistic” theories of “reduction”. But Firth notes that Reichenbach's position is much more extreme than that of Wild. According to Wild, those things that are called ‘sense-data’ by modern empiricists really are part of what is given; what he (Wild) objects to, as we have just seen, is the view that sense-data alone are given. For, according to Reichenbach sense-data (what he calls “impressions”) are not given at all. “What I observe”, he says “are things, not impressions. I see tables, and houses, and thermometers, and trees, and men, and the sun, and many other things in the sphere of crude physical objects; but I have never seen my impressions of these things”.137 [Reichenbach is not entirely fair to the positivists, as Firth also notes. The latter is careful to point out that one could be a positivist and yet also hold that besides sense-data or impressions, material objects could also be given: not all positivists were as committed as Reidchenbach suggests to the “reductionist” thesis that Reichenbach sees himself arguing against. Thus, Firth notes, by way of example, R. Carnap in his essay on The Unity of Science, where Carnap, while characterizing himself as a positivist, questioned the view that Reichenbach refers to as “the positivist dogma”, that is, the view that it is sense impressions alone that are given: there is the alternative, Carnap suggested, that “material things are elements of the given”, and although “it is not often held to-day, it is ... more plausible than it appears and deserves more detailed investigation”.138 It is certainly

67 true that the “reductionist” thesis was characteristic of the positivists of a certain period. Perhaps Carnap was here feeling the influence of Wittgenstein, who by this time had altered his view from the phenomenalism of the Tractatus to the physicalism that appeared in the Philosophical Investigations.] There have been many statements of this sort in the literature. We have looked at similar claims by Quinton and Chisholm, above in the main body of this essay. As Firth points out, the crucial point was made already by William James in his Principles of Psychology, where he stated that a perceiving “is one state of mind or nothing”. It is not a complex of which a sense-datum is a literal part – as an inference based on a sense impression would have that impression as a literal part. We certainly ought not to say what is usually said by psychologists, and treat the perception as a sum of distinct psychic entities, the present sensation namely, plus a lot of images from the past, all ‘integrated’ together in a way impossible to describe. The perception is one state of mind or nothing.139

To be sure, we may look at a physical object in such a way that we attend to the sensory core and ignore those aspects of what is perceived that go beyond that sensory core. As James put it, in such a context what we apprehend approaches “sensational nudity.” When so attending to the core of a perception, e.g., of a painting, ...we lose much of its meaning, but, to compensate for the loss, we feel more freshly the value of the mere tints and shadings, and become aware of the lack of purely sensible harmony or balance that it may show.140

But it is still true that, while perceptions do have a sensory core, sense-data do not exist as literal parts or constituents of perceptions; a perceiving is, to repeat, a unity. But is also worth emphasizing that there are such things as sensations. They do exist as complete and independent states of mind. In fact, the sensory core of a perception is the same thing or entity as the sense-datum of which I am aware when I attend to the sensory core alone. Except that when it is merely sensed, it is sensed alone, while when it is the core of a perception it is located as part of the material object that is perceived. When a material object is perceived, it is given to the perceiver as a whole, in a unitary act of consciousness. But the material object itself is a complex whole, a bundle, not a substance, a structured bundle to be sure,

68 but still a bundle, a bundle of appearances of which the sensory core of the perceiving is a (metaphysical) part. It is not uncommon for this bundle account of material things to be accompanied by an account of perception in which perception is taken to be an inference. Perception is taken to involve, on the one hand, a direct awareness of a sense-datum and, on the other hand, an inference constituting the “perception” of the material object, as a mediated or indirect awareness of the object. Thus, for example, Locke in his Essay concerning Human Understanding suggests that perception is an inferential process which begins with awareness of a sense impression and ends with the “idea” of a physical object. According to this account, the perception of a material object always involves a sensation and a subsequent act of judgement; every perception, therefore, includes awareness of a sense-datum as a distinct, indeed a temporally prior or initiating distinct act or state of consciousness. To use Locke’s example, when we look at an alabaster globe, we are responding differentially to a material object. In that response an idea of sense, a sense-datum, is (usually) imprinted on our mind; it is the impression of a flat circle. But we have learned for past experience that the cause of this sensible appearance is a convex body, and so “judgment frames to itself the perception of a convex figure.”141 Locke admits that the inferential transition from sense-datum to judgement “in many cases by a settled habit ... is performed so constantly and so quick that we take that for the perception of our sensation which is an idea formed by our judgment; so that the one, viz. that of sensation, serves only to excite the other and is scarce taken notice of itself.”142 But he does not doubt that both the sensing as direct awareness and the inferential judgment always occur when we perceive a physical object and that they always occur one after the other. Berkeley makes much the same point. He eliminates substances by appeal to the empiricist’s PA, and leaves material things as bundles of appearances – eliminating the philosophically problematic scepticism parasitic upon the substance philosophy with its distinction between the way things appear and the transcendental substance which is the supposed reality behind those appearances and which gives the real truth of things, not just the truth as we see and hear and feel it. Berkeley’s account of perception occurs in his New Theory of Vision.143 Berkeley argues that the perception of a material object – an ordinary thing like a chair or an apple, not a substance – is a process of what is in effect inductive inference wherein a sensation – that is, a sense impression, or, as Berkeley would say, an idea

69 of sense – “suggests” a material objects to an observer. A sensation comes to be present in the mind and at that instant the mind becomes aware of it, and no sooner is this so than ... it withal perceives the different idea of distance which was wont to be connected with that sensation.

Thus, ...having of a long time experienced certain ideas perceivable by touch ... to have been connected with certain ideas of sight, I do, upon perceiving these ideas of sight, forthwith conclude what tangible ideas are like to follow.144

Berkeley does allow that there are times when ...we find it difficult to discriminate between the immediate and mediate objects of sight. ... They are, as it were, most closely twisted, blended, and incorporated together.145

So he does make a gesture towards the obviously correct phenomenological point that in perceiving a material object the act of perceiving is a unity; he does admit that the successive components of a perceiving may sometimes be hard to distinguish. At the same time, however, he never really doubts that in perception the act of perceiving is inferential, and that in every act of perceiving there are two successive events: the coming to be in consciousness of a sense-datum and the occurrence of an idea which it suggests. Hume endorses the same view. He argues that it is simply not true that ...no other faculty is required, beside the senses, to convince us of the external existence of body.

For, among other reasons, he argues that the perception of body requires inferences learned through experience: ...our sight informs us not of distance or outness (so to speak) immediately and without a certain reasoning and experience, as is acknowledged by the most rational philosophers.146

As for Locke and Berkeley, so for Hume: perceiving involves sensing on the one hand, and an inference on the other.

70 More recent philosophers who accept the empiricist-BerkeleyanHumean-Russsellian position, that things are bundles of appearances, have rejected the Locke-Berkeley-Hume suggestion that perceivings are inferences. Locke and Berkeley found it merely difficult to distinguish a temporally distinct state of direct awareness in every perception from the inferred concluding idea: most of those who think about these things have not only found it difficult to discern the distinction, they have found it quite impossible. They have pointed out that the fact that a series of perceptions may become increasingly refined or determinate as one examines something in ever greater detail does not constitute any proof that there exist separate states of direct awareness. Thus, C. I. Lewis has remarked that If the content of perception is first given and then, in a later moment interpreted, we have no consciousness of such a first state of intuition unqualified by thought, though we do observe alteration and extension of interpretation of a given content as a psychological temporal process.147

On this phenomenologically more reasonable position, a direct awareness of a sense-datum is a constituent of perceptual consciousness even though perceptual consciousness is not an inferential process. And so, when we perceive, say, an apple, there is present in consciousness a sensory element, the sense-datum – perhaps a round, bulgy red patch – where we can say of the patch that “this is red”. But also present in consciousness is an awareness that this, the patch, is or is of an apple: present to consciousness is the property of being an apple, where we can say in attending to the patch that “this is an apple,” and, at least in those parts that we can sensibly discern, a red apple. It is the presence before consciousness of this property, the property of being an apple, that distinguishes this perception, which is of an apple, from the perception of a tomato. What puts that property before consciousness is a property of the perceiving, it is the fact that the perceiving intends that fact, the fact that this, to which I am attending, is an apple. This account is phenomenologically more accurate than the account of perception found in Locke, Berkeley and Hume. On this more accurate account material things are bundles of appearances; some of these appearances are before consciousness as sense-data; the material object as a whole is before consciousness in our perceptual awareness; perception involves awareness of a sense-datum as a sensory core, but it is not a matter of inference; in perception there is a unified act of consciousness, a perceiving, in which one is presented at once with the sensory core and the fact that this core is or is of a material object.

71 This position is not uncommon. It is, for example, the position of H. H. Price, who argues that perception involves no causal inference, nor any inferential process whatsoever: The two states of mind, the acquaintance with a sense datum and the perceptual consciousness [of the object] just arise together.148

C. D. Broad also endorses this account of our perception of material things. Thus, allowing with respect to perception the sorts of qualifications that Broad is forever making, he does say that ... in a perceptual situation we are acquainted with an objective constituent which sensuously manifests certain qualities, and that this acquaintance gives rise to and is accompanied by a belief that the constituent is part of a larger spatio-temporal whole of a specific kind.149

More recently, Bergmann and Grossmann also accept something like this account of perception.150 We may conclude, then, that in perception we are presented with facts to the effect that “this sensory patch which is red is of an apple”, and that such facts are wholly presented to us. Like other entities in the world, they are what they are and not any other thing: here, as elsewhere, what is, is, and is not something else, not even partially something else.

72 Appendix Three How Not to Lose Your Mind What holds for properties like red holds for other properties, and in particular those properties constitutive of mental entities. I feel a pain in my big toe; I have a toothache. That property of being pained which is present in the two cases is what it is and not another thing. Like red, it is located in each example in a patch. In these cases, however, in contrast to what happens when I perceive something red, when I perceive the property of being a pain, I locate it not in a material object other than myself, but in that material object which is my body. In fact, when I locate the pain I locate it on the inside of my body. The sense datum, the patch of pain, and the fact of this part of my body being pained in this way, are facts that only I observe. They are, as one says, private. That creates problems for philosophers, mainly the problem of other minds. Be those problems as they may, these facts that I sense and that I perceive or observe are what they are. Their existence is not be denied: only someone taking some philosophical theory to its absurd conclusions would be silly enough to deny those facts or to say that they were other than they appear to be – a radical behaviourist perhaps, or maybe some sort of materialist would say such things, but nobody who took seriously the way the world feels and the way his or her body feels. Pains are generally localized in one spot in one’s body, though many spots can be pained. Hunger pangs, too, are localized, but occur only in roughly the same spot. Pleasures also come localized – the sweetness on one’s tongue of a candy, the pleasant feel of the touch of a baby. Some come unlocalized – the pleasure of an orgasm occurs throughout the body – so does the pain of suffering through depression. And then there are thoughts. Descartes was right when he recorded the sudden shift of attention from the perceived to the act of doubting: “I think!”, he exclaimed, not a little surprised – surprised, perhaps, at the fact that the thing of which he was now aware was different in kind, totally different in kind, from the things, the sensations and the material objects, of which previously he was aware. His “I think!”, his “Lo! a thought”, records the observation of a fact which is what it is and is not an other thing; it is a non-sensible non-material thing or fact given to us in our inner awareness of our own mental processes.

73 We have all had the Cartesian experience. We are observing something, perhaps something a little naughty, perhaps something forbidden, and then we catch ourselves doing this observing – self-consciously we become aware that we are observing these things, and, perhaps, feel a bit uneasy that our minds should be drawn to such things. We look away. Descartes took the mental facts of which he was aware in the cogito to exist but also, however, to be something more than ordinary facts. Having said “cogito”, he took this to describe an ordinary commonsensical fact, but took it also to be true that this ordinary fact implies an objective and necessary connection to something else, namely, that simple substance which he argued was his real self that lies behind its appearances. In the cogito he discovers those appearances which are his thoughts, but within them, he holds, there is about their reality a tie that reaches out to and connects them necessarily to another reality. From “cogito” he thought he could move by a necessary movement of thought – “ergo” – to his own existence – “sum” – an existence about which he was soon led to recognize that “sum res cogitans” – I am a thinking thing, that is, a substance. So one could not say of the fact that Descartes discovered or located in his “cogito” that it is what it is and not another thing. For, while that fact no doubt is what it is, its being is also determined by another, Descartes argued, or rather simply took for granted: it is, he held, in fact necessarily determined by another – necessarily determined in its very being by its connection to something else, something which it is not. So, it is what it is but also something which it is not: its very being is bound up with and is inescapable from this other being – this other being which it is not and yet also inseparably is. So Descartes is holding that the fact with which he is presented in the “cogito” is what it is, but one cannot say of it that it is not another thing – since its being is inseparable from that of another – another of which one can say too that it both is what it is but also (partially) something which it is not: each inferentially implicates the being of the other. Just as the holist, Sellars’ holist, for example, holds that red and green are not logically and ontologically self-contained but each implicates the being of the other through the principle that nothing that is red can be green, which is true ex vi terminorum. PA excludes the Cartesian substantial mind. It excludes too the supposed objective necessary connections that are thought to lead from “cogito” – I think – to “sum res cogitans” – I am a thinking thing, a simple substance, something not given in the “cogito” itself – something transcending the appearances which are captured in the “cogito”, but of which

74 those facts of which I am conscious in the “cogito” are the appearances – the “mere” appearances, the way the substance presents itself to consciousness, but not that substance itself, not the inner reality that lies beyond or behind, or perhaps below, but in any case a reality that transcends those appearances. Like any such “reality” that purports to transcend the appearances of things as they are given to us in sensible experience and inner awareness, such a “reality” is eliminated by the empiricist’s appeal to PA. Descartes infers from the facts as they are given in the “cogito” the presence of his substantial self, which is not given in the “cogito.” Descartes is correct: if one is to allow such a transcendent entity into one’s ontology, then appeal to PA will not suffice – the entity transcends our ordinary experience of the world, so a transcendental argument is required. But the argument is not convincing: Descartes convinced no one but himself. Perhaps he also convinced Cardinal Bérulle. But it was not long before the good Bishop of Rouen, Pierre Daniel Huet, had transformed “I think, therefore I am” into “I thought, therefore perhaps I was.” Transcendent entities inevitably generate scepticism, and, indeed, as we have been suggesting, a radical and inescapable scepticism. Not the mitigated scepticism of commonsense, not our own fallible understanding, which is the source of our mitigated scepticism – mitigated because, while it admits error is possible and sometimes actual, it also allows that there are reasons through which such error can be corrected. Rather, the scepticism parasitic upon the introduction of transcendent entities is, as we have argued, a radical scepticism under which nothing can be known, a scepticism where error is in other words pervasive and everywhere, and where there is no correcting that error. It is a world in which nothing is knowable, not you, not my desk, not about the reality of whether the apple I am eating is poisonous, not even myself: the thing which is seemingly closest to me turns out to be a reality, or perhaps an abyss, which forever eludes me. I cannot find the world, I cannot find my place in that world, I cannot even find my own self. I do not and cannot know, and that is or ought to be terrifying. With the introduction of transcendent entities, the world inevitably turns into this fearful and terrible place in which I can know nothing, and can do nothing about it, and in any case my own self turns out in the midst of nothingness to be itself a hollow of nothingness. All I can do is mess about in the sea of appearances, flail hopelessly about with only the assurance that so far as I can tell it is all pointless. The reality which alone would give it point is the transcendent reality which I inevitably discover is beyond the grasp of any human un-

75 derstanding. Hope can be sustained in such circumstances only by illusion. It is a dangerous way out. More appropriate is the determination to accept PA and banish the transcendent entities that generate despair and thereby generate the need for hope, but permit only a hope which can be based on illusion. Holism is such an illusion. So is materialism. The former is no doubt mistaken. The latter is just silly. It was Hume who solidly broke with the Cartesian self. His appeal was to PA. Here is the Cartesian position, as it is expressed – correctly – by Hume in the Treatise, in his discussion of “Personal Identity”: There are some philosophers [e.g., Descartes] who imagine we are every moment intimately conscious of what we call our self; that we feel its existence and its continuance in existence; and are certain, beyond the evidence of a demonstration, both of its perfect identity and simplicity.151

There are no necessary connections that lead us from what we experience to this self of the Cartesians: PA has eliminated any such objective necessary connections. So, if such a self is to be admitted into our ontology, it must be because it is given to us in experience. However, that substance is not to be found in experience, nor are the necessary connections that it is supposed to provide and which are supposed to lead us to it: in the mind as in material things, so far as experience is concerned, the parts are experienced as wholly themselves and therefore logically self contained and therefore in turn ontologically self-contained. “Unluckily,” says Hume, ... all these positive assertions are contrary to that very experience which is pleaded for them; nor have we any idea of self, after the manner it is here explained. For, from what impression could this idea be derived?, what must become of all our particular perceptions upon this hypothesis? All these are different, and distinguishable, and separable from each other, and may be separately considered, and may exist separately, and have no need of any thing to support their existence. After what manner therefore do they belong to self, and how are they connected with it?152

Hume throws a challenge – it is ironic, he expects the Cartesian, if he or she is honest, to have no answer. Or, if there is an answer, then it rests on an illusion. For my part, when I enter most intimately into what I call myself, I always

76 stumble on some particular perception or other, of heat or cold, light or shade, love or hatred, pain or pleasure. I never can catch myself at any time without a perception, and never can observe any thing but the perception.... If any one, upon serious and unprejudiced reflection, thinks he has a different notion of himself, I must confess I can reason no longer with him. All I can allow him is, that he may be in the right as well as I, and that we are essentially different in this particular. He may, perhaps, perceive something simple and continued, which he calls himself; though I am certain there is no such principle in me.153

The self is a bundle of perceptions: I may venture to affirm of the rest of mankind, that they are nothing but a bundle or collection of different perceptions, which succeed each other with an inconceivable rapidity, and are in a perpetual flux and movement. Our eyes cannot turn in their sockets without varying our perceptions. Our thought is still more variable than our sight; and all our other senses and faculties contribute to this change: nor is there any single power of the soul, which remains unalterably the same, perhaps for one moment. The mind is a kind of theatre, where several perceptions successively make their appearance; pass, repass, glide away, and mingle in an infinite variety of postures and situations.154

The Cartesian self is simply an illusion, however tempting an illusion it might be. There is properly no simplicity in it at one time, nor identity in different, whatever natural propension we may have to imagine that simplicity and identity. 155

This is not to say, however, that we have no idea of the self. Hume in fact introduces the idea of the self into his discussion of pride and humility, which begins just a few pages, two score and a bit, after the elimination of the Cartesian self. It is evident that pride and humility, though directly contrary, have yet the same object. This object is self, or that succession of related ideas and impressions, of which we have an intimate memory and consciousness.156

This idea of the self that we do have and that we do use is a complex idea. It is, moreover, the idea of a structured bundle, where much of the structuring is done by the body. ...pride and humility have the qualities of our mind and body, that is, self, for their natural and more immediate causes.157

77 But to add further structure, there is also, Hume has already mentioned earlier, the natural teleology of the organism to provide a unity, as well as the teleology of the person him- or herself as he or she goes about the business of living in the world, sleeping, eating, making merrie, getting ready to go to London, or what not. There is, however, another artifice, by which we may induce the imagination to advance a step further; and that is, by producing a reference of the parts to each other, and a combination to some common end or purpose. A ship, of which a considerable part has been changed by frequent reparations, is still considered as the same; nor does the difference of the materials hinder us from ascribing an identity to it. The common end, in which the parts conspire, is the same under all their variations, and affords an easy transition of the imagination from one situation of the body to another.

The natural relation of the parts yields a sense of identity: But this is still more remarkable, when we add a sympathy of parts to their common end, and suppose that they bear to each other the reciprocal relation of cause and effect in all their actions and operations. This is the case with all animals and vegetables; where not only the several parts have a reference to some general purpose, but also a mutual dependence on, and connexion with, each other. The effect of so strong a relation is, that though every one must allow, that in a very few years both vegetables and animals endure a total change, yet we still attribute identity to them, while their form, size, and substance, are entirely altered. An oak that grows from a small plant to a large tree is still the same oak, though there be not one particle of matter or figure of its parts the same. An infant becomes a man, and is sometimes fat, sometimes lean, without any change in his identity.158

So the self and the mind is a bundle of appearances, just as a material object is a bundle of appearances. Back to Sellars. It is his account of mental events in which we are interested. It will be best to begin with those parts of the bundle that make up a mind that we call thoughts. These are the things that Descartes discovered as his attention turned inward. These are the entities the discovery of which he recorded when he said for the first time “cogito.” Note, by the way, that there is nothing wrong with Descartes saying/judging “cogito”: there, in saying that, he is simply recording his inner awareness. This is perfectly compatible with the empiricist’s PA: in fact, it is precisely what PA calls

78 for. It is Descates’ second step “ergo sum” wherein he does not discover merely his thinking but as a rationalist moves to the “sum”, that is, moves by way of an objective necessary connection to the metaphysical fact that he records as “sum res cogitans.” It is this second step that is incompatible with PA. But, to make the point again, the “cogito” alone is not wrong: in it he records his discovery of his own inner mental states of awareness – he is recording his acquaintance with these states. In the “cogito” he is now conscious of his inner states, states that were/are different – specifically and qualitatively different – from any of the sensible things and any of the material things which had previously been the objects of his attention. It is thought, these sorts of thought that Descartes discovers, that we wish here to consider, and more specifically consider what Sellars has to say about these mental events called thoughts. What is characteristic of thoughts is their intentionality. Descartes is doubting when he catches himself thinking; the doubting is a species of thinking. The doubting is a doubting that so-and-so; the doubting intends or has as its intention the state of affairs that so-and-so. This characteristic of the doubting is in fact what is characteristic of thought in general. We can say this: A thinking has a characteristic sort of property. It has a property which is such that the thinking is about, or, as Brentano taught us to say, intends other facts or states of affairs. Sellars’ account of mind turns in large part on his account of the intentionality of thoughts. Sellars argues that the proper account of intentionality, that distinctive feature that defines thought, is to be drawn in terms of the forms and functions of natural linguistic items. These functions are given by the world-word, word-word, and word-world connections and regularities and norms that constitute the meaning in the broadest sense of words and sentences. Sellars assumes, reasonably so, that one can give a behaviouristic account of this language. This assumption is what Sellars calls “verbal behaviorism”. According to VB [verbal behaviorism], thinking ‘that-p,’ where this means ‘having the thought occur to one that-p,’ has as its primary sense [an event of] saying ‘p’; and a secondary sense in which it stands for a short term proximate propensity [disposition] to say ‘p’.159

Thinking that p, judging that p, hoping that p, disapproving that p, wishing it were that p: all these have the form of “says that p”. Now, when something is said, a cognitive content is conveyed: when someone says that p

79 then the cognitive content of p is conveyed by that saying. So, like thoughts, sayings carry meaning, they too are intentional. Sellars is going to argue that the intentionality of thought is in fact parasitic on the intentionality of saying, or, rather, that thought is inner speech and saying is overt speech, and that the intentionality of both are essentially analogous and can be understood in the same way. Let us see. Sellars’ account of verbal behaviourism has its origin in his essay on “Empiricism and the Philosophy of Mind,” parts of which we have already examined. It is in this context we once again meet up with his mythical figure, Hypothetical Q. Jones. Jones lives in a society in which all language is behaviouristic. Where we would take some use of language as referring to inner states, there is only language that refers to the behavioural analogue of these, either overt behaviour or dispositions to overt behaviour. In this society everyone is said by Sellars to be “Rylean”: these are our Rylean ancestors. The “Rylean language” is behaviourist but for all that it is sophisticated. It includes, of course, proper names and predicates, including relational predicates. These have material objects, located in space, and the properties of these objects as their referents. It also contains the usual logical expressions and operators, as well as the apparatus necessary for the assertion of subjunctive conditionals. Consonant with Sellars’ account of linguistic meaning in terms of world-word, word-word, and word-world connections, that is, as a sort of functional classification, this (mythical) Rylean language also contains the fundamental resources of semantical discourse. This means that our Rylean ancestors can say of their friends’ or their children’s or, indeed, of even their own, utterances, that they mean thus and so, that they stand in various logical and other inferential relations to one another, that they are true or false, and so on. What this Rylean language lacks are any resources for speaking of inner episodes, thoughts or experiences. One can say “I am in pain”, one can even reflect on it and comment on it semantically. But there is no inner reality which it reports. Saying “I am in pain” has nothing to do with any inner state or episode. In fact, it represents nothing inner. Its meaning is given by the world-word connections that have to do with grimacing, grasping one’s finger which is bleeding, treating gingerly the bruised arm, and so on, and by the word-world connections that include bringing about from others such things as consoling behaviour and other sympathetic re-

80 sponses – “sympathetic” as we might say, but not meaning by it quite what our Rylean ancestors might mean by it, since for us but not for them it alludes to various inner episodes. There would be word-word connections to the effect that the pain is sharp or is nagging and that it is not pleasant. But these expressions and connections would find their cash value in behaviour, not in any experiences one might have. There might even be a philosopher who says “cogito” but what he or she would mean by that is that he or she is now disposed to say that he or she is saying or is disposed to say that such and such. Sellars now has a stranger appear, the genius Hypothetical Q. Jones (whom we have already met): [In] the attempt to account for the fact that his [or is it her?] fellow men [persons?] behave intelligently not only when their conduct is threaded on a string of overt verbal episodes ... but also when no detectable verbal output is present, Jones develops a theory according to which overt utterances are but the culmination of a process which begins with certain inner episodes .... [His] model for these episodes which initiate the events which culminate in overt verbal behavior is that of overt verbal behavior itself.160

Jones develops a theory, and in that theory postulates the existence of thought: this is how we, or our Rylean ancestors came to think of and about thoughts. In this theory that he develops, Jones introduces unobserved inner episodes as the hidden causes of various kinds of overt behaviour, in the way that the physicist postulates atoms as the causes of observable behaviour of, say, gases, where the motions of these unobservable entities are taken to explain the observable facts of pressure change. This theory, let us note again, postulates inner episodes as hidden causes. There are inner awarenesses of pains to explain why one might say, while grimacing, “I am in pain”, and there are inner awarenesses of thoughts to account for many of the never actualized short-term dispositions to say various things, e.g., “this apple is red” which one might think but never say. These inner thoughts are supposed by the theory to be cases, unobserved by the bystanders, in which one silently says to oneself what one is only disposed to say overtly. It is these postulated inner episodes of saying things to oneself that are thoughts. Sellars suggests that these inner, postulated thoughts have semantical properties akin to those semantical proper-

81 ties had by overt speech. He further argues that one can understand the intentionality of thought, the meanings carried by these inner episodes, in terms of the semantical properties which these episodes are postulated to have in analogy with the semantical properties of overt speech. Sellars use the myth of our Rylean ancestors as a tool to the formulation of his account of those inner episodes we call thoughts and of the intentionality of those thoughts: here is where we find the core of his analysis of that property, intentionality, which is distinctive of mind. But the myth makes clear that there is no essential difference between the meaning of overt language and the meaning of thought: if the inner episodes have intentionality, then so do episodes of overt behaviour or the short-term dispositions to such behaviour. The cognitive content of thinking that p and the cognitive content of saying that p are both given by the linguistic meaning which ‘p’ carries, its semantical or conceptual content. Although the primary use of semantical discourse remains the semantical characterization of the overt verbal episodes which are the items of our ancestors’ Rylean language, the Jonesean theory, we see, simply carries over the applicability of those semantical categories to its postulated unobserved inner episodes. i.e., to (occurrent) thoughts. The point of the Jonesean myth is to suggest that the epistemological status of thoughts (qua inner episodes) in explaining candid public verbal behaviour and short-term dispositions to such behaviour is to be understood logically and epistemologically as analogous to the logical and epistemological status of, e.g., molecules relative to the public observable behaviour of gases. Thus, we are told that [Thought] episodes are ‘in’ language-using animals as molecular impacts are ‘in’ gases, not as ‘ghosts’ are in ‘machines’.161

Atoms or molecules are posited in kinetic theory as conforming to causal regularities that are analogous to the regularities (the Newtonian laws of impact) governing ordinary objects such as billiard balls. The thought episodes posited by Jones are similarly taken to conform to regularities analogous to those describing the behaviour to be explained. These are the functional regularities that describe meaning. The intentionality of thought, that which makes it characteristically mental, is given by the cognitive content of the thought: the aboutness of thought is simply the aboutness of language – the cognitive content carried in the thought that p and the cognitive content carried in the saying that p are essentially the same, namely, the cognitive content of the sentence ‘p’.

82 Now, what is essential to cognitive content (Sellars holds) are the word-word connections. The world-word connections are irrelevant to that content. That means that the same cognitive content can be carried by it does not matter what. The concept of an occurrent thought is that of a causally-mediating cognitive role player, whose determinate character is left indeterminate. That therefore leaves some space open as it were for some sort of “identity theory”: identifying thoughts with brain states, is left as a real possibility. [The] fact that [thoughts] are not introduced as physiological entities does not preclude the possibility that at a later methodological stage they may, so to speak, ‘turn out’ to be such. Thus, there are many who would say that it is already reasonable to suppose that these thoughts are to be ‘identified’ with complex events in the cerebral cortex ...162

So, on Jones’ theory introduced to posit unobserved episodes to explain the verbal behaviour of our Rylean ancestors, we can all embrace materialism: the mental is the physiological; what makes it mental is the cognitive content that it carries, and the cognitive content that it carries, what it means, is given by the cognitive content of the sentence ‘p’ the tokening or utterance of which expresses the thought. Since, on Sellars’ account, the concept of a thought is fundamentally the concept of a functional kind, no ontological tensions would be generated within what Sellars calls the “scientific image” of the world and of the persons in it, by the identification of items belonging to that functional kind with, for instance, states and episodes of a person’s central nervous system, the person now thought of simply as yet another organism. Then there is what Sellars calls the “manifest image” of the world and the persons in it, our ordinary picture or image of persons as thinkers and feelers and doers. It is Sellars’ conclusion that this conception of thought that comes from the manifest image of persons as thinkers, can fuse smoothly with the scientific image’s conception of persons as complex material organisms having a determinate physiological and neurological structure. Sellars’ basic idea is that the intentionality of the mental is to be understood in terms of episodes that have the theoretical status of posited entities to which the theory assumes the semantic categories and rules of public language apply, where these are understood in terms of a functional classification of linguistic items. This has become the starting point of much recent work in the philosophy of mind. As one commentator, Daniel Dennett, has put it,

83 Thus was contemporary functionalism in the philosophy of mind born, and the varieties of functionalism we have subsequently seen are in one way or another enabled, and directly or indirectly inspired, by what was left open in Sellars’ initial proposal ...163

Later work defending versions of materialism, such as the efforts of Paul Churchland to effect the assimilation of thoughts and their intentionality to the functioning of brain states, largely find their inspiration in Selllars.164 They share his crucial premise that the cognitive content of concepts is given by the word-word inferential connections in which the concept is embedded.165 There remains one aspect of Sellars’ account of thought to be discussed. Sellars’ proposal is that we can understand the epistemological status of mental concepts by an appeal to the contrast between theoretical language and non-theoretical language. This proposal is inseparable from his comprehensive critique of the “myth of the given”. The philosophical framework that Sellars’ builds into the (mythical) given is taken to include the idea that empirical knowledge rests on a foundation, where this idea of a foundation is that developed also by Sellars. (We have seen that empiricists need not, and that some do not, commit themselves to there being such a foundation.) Another part of the assumed framework is the assumption that the privacy of the mental and one’s privileged access to one’s own mental states are fundamental features of experience, but that they are both only “privacy” and “privileged access”, and are not, as one (so Sellars thinks) usually takes for granted, both logically and epistemologically prior to all intersubjective concepts pertaining to inner mental episodes. Sellars argues, on the contrary, that what begins in the case of inner episodes as a language with a purely theoretical use can acquire a firstperson reporting role. The Jones’ story continues. This genius of language use, having introduced thoughts by positing them as theoretical entities, now introduces another feature into the expanded Rylean language by adding a dimension of first person discourse about these theoretical entities. As theoretical, the episodes remain unobservable. But the language has added to it a further role, new patterns of usage. One trains the language users by, as Sellars would have it, what is in essence a process of conditioning, to have what can be seen as a sort of “privileged access” to some of the inner mental episodes that have been posited as occurring in them. They are trained to respond directly and non-inferentially when the theory

84 asserts that a thought has occurred to them. That is, they are so trained that when a posited thought event occurs as it were within them, when he or she has one of the thoughts that are posited as being there within him or her, they then respond to that event with another thought which is about that first thought. This meta-thought is to the effect that one is thinking the first thought. It is seen to be a special virtue of this aspect of Sellars’ Jonesean story that it shows how the essential intersubjectivity of language can be reconciled with the “privacy” of inner episodes, i.e., ... that it helps us understand that concepts pertaining to such inner episodes as thoughts are primarily and essentially inter-subjective, as inter-subjective as the concept of a positron, and that the [first-person] reporting role of these concepts ... constitutes a dimension of [their] use ... which is built on and presupposes this inter-subjective status.166

The privacy is, strictly speaking, an illusion. It is simply an artifact of an essentially intersubjective language. One is given this special training only with regard to one’s own thoughts. It is in this sense167 that thoughts are private. Central to Sellars’ case against what is in his way of speaking the “Myth of the Given” is his recognition of the irreducibly normative character of epistemic discourse. We have of course seen that few empiricists would object to this point that Sellars makes and therefore that few indeed are caught up by the “Myth of the Given” in Sellars’ sense: his opponents are largely persons of straw. Be that as it may, Sellars’ point is that, in characterizing an episode or a state as that of knowing, we are not giving an empirical description of that episode or state, we are placing it in the logical space of reasons, of justifying and being able to justify what one says.168 Once one grants that the senses do involve differentially responding to a stimulus but that they as such, simply as senses, grasp no facts, in the epistemological sense of grasp, so that all knowledge that something is such-and-so (all subsumption of things under concepts) presupposes learning, concept formation, and linguistic or symbolic representation, then it follows that “... instead of coming to have a concept of something because we have noticed that sort of thing, to have the ability to notice a sort of thing is already to have the concept of that sort of thing, and cannot account for it.”169 This is no doubt correct: who would disagree? It is not this that is in fact central to Sellars’ attack on the “given”. Really what is central to Sellars’ attack on the “Myth of the Given”, what is in fact central but which centrality he fails to point out, is the holistic account of cognitive

85 meaning whereby the meaning of a concept is said to be the inferences in which it occurs and where these inferences are taken to hold ex vi terminorum. What this enables Sellars to do is ignore reference. The empiricist can accept most of what Sellars says about the role of norms in determining which differential responses to things sensed are to count as observations. But what he or she insists upon, is precisely what Sellars dismisses without so much as a goodbye – he simply takes it for granted that he can ignore it: what the empiricist insists upon is that it is reference and not the network that gives the cognitive meaning of a concept. In thinking that he can ignore reference, Sellars thinks that he can simply ignore the deliverances of experience. In thinking that he can ignore reference, Sellars thinks that he can ignore the plain fact that Descartes, and we, discover in the “cogito”: there are thoughts, we experience them, and they are experienced as characteristically different from other sorts of events: they are what they are and are not another thing – they are, and they are what they are, and in particular they are not brain states, they simply don’t have the properties had by things physiological. That is what we discover in the “cogito.” To think that mental states, thoughts, make their appearance in discourse as posited entities, is the height of absurdity. Atoms and molecules are posited, they are not observed: their existence is inferred, not witnessed. Thoughts, to the contrary, are observed, they are not posited. Has Sellars never caught himself thinking? has he never experienced what Descartes experienced when he exclaimed “cogito”? Of course he has! Of course he knows what thoughts are and how they differ from other things, e.g., such other things as physiological states of the brain. To be sure, he would likely deny this, the obvious testimony of his own inner awareness. But that is because he is committed to his theory and to its materialist implications. Materialism is indeed silly, as the experience of the “cogito” tells us. It is his commitment to his theory that allows him to ignore or at least explain away the clear testimony of experience. Let us be clear: IT IS THE TESTIMONY OF EXPERIENCE that there are thoughts; we experience them and experience them as they are, as the sorts of things they are; these entities are what they are and not other things, and in particular they are not brain states. Sellars can dismiss these entities, our thoughts, only because of his holism and what is an inadequate theory of meaning, inadequate because it is incompatible with the empiricist’s PA.

86

This deals with thoughts, one of the two aspects of the mental with which Sellars attempts to deal. The other issue remains. This issue is that of sensations or qualia, the sense impressions of the empiricist. It seems that Sellars’ account here has proved to be as uncritically accepted by many and as popular as his account of thoughts.170 Sellars correctly insists that the ‘of-ness” of sensations – e.g., a sensation’s being of a red triangle or of a sharp shooting pain – is not the intentional “of-ness” or “aboutness” of thoughts. The “rawness” of “raw feels”, as he misleadingly describes qualia and sense data, is there, it is non-intentional, simply given as their character, their character as sensible events as contrasted to the intentional nature of thoughts.171 Sellars’ views regarding sensory episodes regard them epistemologically as parallel to that of his account of occurrent thoughts: they are also posits in the Jones’ myth. But the resulting ontology of sensations differs from that of his functionalist account of thoughts. In a final episode of the Jonesean myth, sensations are introduced as elements of an explanatory account of the occurrence in various circumstances of perceptual cognitions, having determinate semantic contents – like when a Rylean says or is disposed to say “I see a red cat” when in fact there is nothing that is a cat and nothing red in the environment to which he or she could differentially responding – just like Macbeth and his dagger. Is this a dagger which I see before me, The handle toward my hand? Come, let me clutch thee. I have thee not, and yet I see thee still. Art thou not, fatal vision, sensible To feeling as to sight? or art thou but A dagger of the mind, a false creation, Proceeding from the heat-oppressed brain? I see thee yet, in form as palpable As this which now I draw. Thou marshall’st me the way that I was going; And such an instrument I was to use. Mine eyes are made the fools o’ the other senses, Or else worth all the rest; I see thee still, And on thy blade and dudgeon gouts of blood, Which was not so before. There's no such thing:

87 It is the bloody business which informs Thus to mine eyes. ...172

In circumstances such as these Jones proposes to explain the person’s saying that he or she perceives something by positing that there is privately before him or her an image, of a red cat, if that is what is said to be perceived, or a bloody dagger, if that is what is said to be perceived – where this posited image, to which the person is supposed to be perceptually responding, is a replica of an appropriate real entity, a real red cat or a real bloody dagger, as the case may be. But then Jones makes his theory more general and posits such inner replicas for every perceptual response, those where the object perceived is really there as well as those, like our red cat or Macbeth’s dagger, where it is not. More generally, then, Jones posits that there are sensations in any perceptual situation, inner sensory replicas of the outer and public sensible reality. The reality to which perceptual responses are normally made is a world of material objects, but now, Jones posits, there is an inner sensory replica of this outer reality. These inner replicas are the sensory impressions which Jones’ theory takes as constituting the sensory core of one’s perceptual response. They are the entities that are the counterparts of sense data for our Rylean ancestors. ... the hero [Jones] ... postulates a class of inner – theoretical – episodes which he calls, say, impressions, and which are the end results of the impingement of physical objects and processes on various parts of the body.173

Just as atoms are models of billiard balls, and just as (for Sellars) thoughts are models of sentences understood functionally, so are these Jonesean images, and, more generally, Jonsean sensations, models. But where the thoughts that are posited are modelled on utterances, sensations are taken to be posited as replicas of the sensible things and their qualities and relations. Sensations our genius Jones posits as “a domain of ‘inner replicas’ which, when brought about in standard conditions share the perceptible characteristics of their physical sources.”174 The core idea of the modelling of this posit is the occurrence as events “in” perceivers of “replicas.” In the model what is replicated is not perceivings of perceptible events (which would mistakenly inject into the account of impressions the intentionality of thought), but replicas of those perceptible events themselves. And, although the entities which are replicated are facts, things as qualified, the entities introduced by the theory are not simply further facts but rather states of a perceiving subject. Thus, although talk of the “aboutness” of

88 sensations, like that of the “aboutness” of thoughts is, on Sellars’ view, a matter of the correct classification, in the one case, that of thoughts, where the classification is based on a functional (logical, semantic) analogy with the language of overt speech, whereas in the case of sensations the entities posited are classified according to an analogy with perceptible things which is in the first instance extrinsic and causal, but which also attributes to the posited sensations a determinate intrinsic qualitative content. The specific point of the model is to insist that states of, e.g., sensing, say sensing red-ly (put adverbially to highlight the status of ‘sensation’ as a “verbal noun”, referring not straight off to a fact but to a state of the perceiver), are characteristically brought about in normal perceivers in standard conditions by the action of red objects on the eyes. So understood, their explanatory task in relation to cognitive perceptual takings (especially non-veridical perceptual judgments) occurs through their being conceived as resembling and differing from other sensory states – e.g., sensing green-ly, contrasted to sensing red-ly, or sensing square-ly, or sensing b-sharp-ly, or sensing sweet-to-the-taste-ly etc. They will resemble and differ from one another in a manner formally analogous to the way in which objects on which the “replicas” are modelled – e.g., red patches, green patches, square patches, and so on – resemble and differ from one another. Now, Jones’ theory does not introduce new facts but rather new characteristics, new properties in other words. Sensory contents, as posited in Jones’ theory, are not things or facts, but are states of perceivers. What for the empiricist are sense data or sense impressions that are before consciousness and independent of it, are on Jones’ theory, states of the perceiver. As ways of perceiving, sense impressions are construed by Jones’ theory as dependent on the perceiving of them. And Jones extends his theory to include all sensory awareness. So all sense impressions – everything that Sellars would classify as a “raw feels” –, even those which appear to be out there in space, are modes of perceiving. Since these sense impressions are thus dependent ontologically on the perceiving of them, Sellars’ account of sense impressions is essentially idealistic. If Sellars is a materialist, he is so only because he is also an idealist. At this point in the discussion two remarks need to be made. The first of these notes that there arises the same problem for Sellars’ adverbial analysis as arose for Aristotle. The latter, it will be recalled, made the mind of the knower identical with the thing known: the substantial form of the object known comes to be, literally and in itself, in the mind of the knower. So when one comes to know perceptually the oak tree

89 the substantial form of the oak comes to be in the mind of the knower. But, the objection runs, if the form in the material object is what causes it to be an oak, then why does the mind not become an oak? The answer offered by the mediaevals was that the form is in the oak tree substantially but is in the mind only intentionally. The form is the same form but exists in the mind as a property of the mind. Through it the mind was similar to the known, a replica of it, but (the objection now goes) how could a property be like a substance? A substance can resemble a substance – when they share a form or when they share a property –, and a property can resemble a property – when they share a second order property –, but how can a property resemble a substance? – that would require a property of a property to also be a property of a substance, it would require a second-order property to be a first-order property – something that is ontologically impossible. In Berkeley’s phrase, “only an idea can be like an idea.” In Sellars’ Jonsean world, there are perceivings and among perceivings are perceivings red-ly. On this view, the quality red seems to have the status of a property of a property. Yet in reality that upon which it is modeled is redness which is a property of a material object. How can a property of a property be like or model or resemble a property? All we are told is what the mediaevals said about forms in the mind: they exist as properties in the mind but they also exist intentionally. That is not an explanation; that is the labelling of an ontological problem only to ignore it. Similarly Jones: he has learned the games the mediaevals played. His positing may sound good, but it is ontologically incoherent, and to speak of replicas and positing and then to say nothing more is, to repeat, to label a problem without solving it. But that should not surprise us: it is evident that the problem is one in which there is incoherence and which therefore cannot be solved. The other problem with perceiving red-ly or perceiving square-ly is that the “raw feel” which is the replica is construed as a property of a perceiving. The knowing of the entity is construed as the relation between the person’s perceiving and a property of it. But the “relation” of knowing is not that of exemplification. Berkeley held that view, and so do many other idealists: to be (for a sense impression) is to be perceived. This is just the view that Moore so rightly criticized in his essay on “The Refutation of Idealism.”175 And Moore was right: the adverbial account of sense that Jones proposes simply won’t do. So we cannot in our Rylean world have the positings that he tries to have: the posited entities are ontologically impossible.

90 The astute will have noted that even with so many problems, the entities that Jones has posited are not quite the things that Sellars aims to find – recall his materialism: we are not yet quite at the point where sense impressions become material. Another step is needed if the Sellarsian project is to be completed. The Jonesean theory has it that the very colour qualia of which we are perceptually aware as existing in space are instead actually states of persons-qua-perceivers. This, remember, is in the “manifest image” of the world and persons in it. So already within the “manifest image,” within the image that takes persons as persons and not as organisms, the ontological status accorded to sensory “qualia” is incompatible with their existing in the objects in physical space that they appear to characterize. To this there is one further complication that needs to be added. This will give us the full picture of Sellars’ conception of reality and of our sensible responses to it. On Sellars’ view, the “manifest image” must in a sense be reduced to the “scientific image”: that is how he secures his materialism. Thus, the conception in the “manifest image” of sensory contents as states of perceivers must ultimately be “fused” with the “scientific image” – the former must in some way be reduced to the latter. They must be so related that the commitment of the “manifest image” to sensations must overcome the seeming barrier presented by the “scientific image” in its commitment to the idea that perceivers are complex systems of micro-physical particles. Sellars’ proposal is that sensory contents can be integrated into the “scientific image” only after both they in the “manifest image” and the currentlyfundamental micro-physical particulars of the “scientific image” are fit into some more adequate ontology. This ontology will be a monistic ontology the fundamental entities of which, Sellars speculates, are all “absolute processes”. Sensings qua absolute processes would then be material, at least so he proposes: [they would be material] ... not only in the weak sense of not being mental (i.e., conceptual), for they lack intentionality, but in the richer sense of playing a genuine causal role in the behavior of sentient organisms. ... Not being epiphenomenal, they would conform to a basic metaphysical intuition: to be is to make a difference.176

These speculations are, I suppose, fun. But in the end, like all versions of materialism, they are silly. In the end one comes down to experience: the red patch that I sense

91 when I perceive an apple, is simply not, in the world as we experience it, a property of a person. Macbeth’s dagger is sensibly there, before him, as bloody red as a real dagger, he experiences it that way and not as some property of his person. The deliverance of experience is clear here as it is on thoughts. Sellars can elaborate all he wishes, and make things complicated, but the convolutions cannot in the end serve to do anything but try unsuccessfully to disguise the reality of the world as we experience it. And in the world as we experience it, mind is simply not matter, and red patches are not physiological processes and neither are bloody daggers.177 Sellars’ proposal is that we can understand how we attribute mental concepts to things by an appeal to the contrast between theoretical language and non-theoretical language. This proposal is inseparable from his critique of the “myth of the given”. The latter involves a holism in which the cognitive content of our concepts is given not by their referents but by the inferences in which they occur, where these inferences are all true ex vi terminorum. It is the ignoring of reference that makes it possible to suppose thoughts and sensations are in the world because they are posited and not because they are experienced. But this account of meaning is unacceptable: it is incompatible with PA. We may conclude that Sellars has failed in his critique of empiricism and in his attempt to defend a form of materialism. Failures too are the other defences of materialism, such as that of Churchland, that take Sellars’ concept of meaning for granted. Macbeth’s experience is of a dagger, and the sensory core of this is a blood red patch having a dagger shape. When Macbeth experiences this blood red patch, he experiences it wholly, as it is. Similarly, when I experience a red patch, I experience it wholly, as it is. It does not follow that my experience is veridical, let alone incorrigible. Nor does it follow that Macbeth’s experience is veridical. But holism is wrong: what I experience, what is given to me in consciousness, certainly is what it is and is not another thing, not even partially another thing. Sellars notwithstanding, that’s the way the world is. PA simply tells us to take this world to be as it is.

92 Endnotes to Study One

1.The phrase is from G. Bergmann; see his “Strawson’s Ontology,” in his collection of essays, Logic and Reality (Madison, Wisc.: University of Wisconsin Press, 1964), p. 177. Compare also: Wittgenstein’s comment that “A philosophical problem has the form: ‘I don’t know my way about.’” (Philosophical Investigations, second edition [Oxford: Blackwell, 1953], sec. 123) and “My aim is: to teach you to pass from a piece of disguised nonsense to something that is patent nonsense” (Ibid., sec. 464). Then there is the comment of N. Malcolm, Ludwig Wittgenstein: A Memoir (London: Oxford University Press, 1958), p. 51, and John Wisdom, “Philosophical Perplexity,” Proceedings of the Aristotelian Society, 37 (1936-1937), pp. 71-88. 2.Though both the metaphilosophy we are considering and that of Wittgenstein begin from the distinction between philosophical statements and those of commonsense, different principles are used for delimiting commonsense. On the one side, there is the appeal to PA. On the other side, what is commonsense is what is sayable according to the rules of ordinary (i.e., non-philosophical) language. For Wittgenstein’s metaphilosophy, see O. K. Bouwsma, “The Blue Book,” in his Philosophical Essays (Lincoln, Neb.: University of Nebraska Press, 1965), pp. 175-202. 3.One must, of course, add that the sentences may contain defined terms, given that the definiens when fully expanded contains only terms which refer to entities of a kind with which one is acquainted. 4.Thus, let us carefully note, phenomenalism, the view that the only entities which exist are those with which we are in fact acquainted, is not entailed by, though it is clearly compatible with the metaphilosophy we are considering. But nothing in PA commits one to the view that the only sensibilia are those with which we are in fact acquainted. For, an entity may be in a category compatible with PA without itself being presented. For the connection between PA and entities with which one is not acquainted, see below, this volume, “Empiricism: Principles and Problems”: the crucial point is, of course, the logic, discovered by Russell, of definite descriptions. Some of the structural and historical connections between phenomenalism and the use of PA in ontology are explored below, in the present essay, especially in section IV. 5.Compare G. E. Moore’s comments on the position of G. F. Stout in “Are the Characteristics of Particular Things Universal or Particular?” in his Philosophical Papers (London: Allen and Unwin, 1959), pp. 17-31. See also his Lectures on Philosophy, ed., C. Lewy (London: Allen and Unwin, 1966), pp. 56-57. See also below, in the present volume, “Universals, Particulars, and Tropes: The Role of a Principle of Acquaintance in Ontology.”

93

6.It is trivial because it is more or less becomes true by definition that properties are universals. It is part of the grammar or axiomatics of ‘universal’ that a property will be a universal just in case that the same property can be in several things. Then ‘same’ is defined so that if two things have properties which are in and of themselves indistinguishable then they have the same property. So it becomes trivial that properties are universals. But, though trivial, it is nonetheless interesting because one can go on to argue that by so defining ‘same’ one can solve the ontological problem of sameness. We return to this point below, in the present volume, “Universals, Particulars, and Tropes: The Role of a Principle of Acquaintance in Ontology.” 7.For an idea of the sorts of diagnostic remarks which would be relevant, see the comments on Platonism, below. 8.It is of course impossible to argue for the existence of mental acts; one can only appeal to one’s human experience of one’s own self. Cf. G. Bergmann, “Professor Ayer’s Analysis of Knowing,” in his Metaphysics of Logical Positivism (New York: Longmans, Green, 1954), pp. 215-227. See also the Appendix to the present study, below. Let me make clear that I want to take ‘acquainted’ and ‘presented’ and similar terms in a commonsensical and philosophically non-problematic sense. Seeing, hearing, etc., are acts of acquaintance. We see trees, hear trains, look at rainbows, and so on. We also more narrowly see colours, hear sounds, and so on. This distinction connects up with the realism-phenomenalism controversy. As I here use ‘acquaintance’ and ‘presentation’, I do not restrict them to the narrow sense that seems at least to exclude realism, though for the most part what I have to say in this study is neutral between the broader and the narrower uses. There is, however, a contrast which is important for our purposes, that between these non-problematic cases of acquaintance and e.g., Plato’s “reminiscence” by which it was claimed that we have acquaintance with forms. Plato’s claimed acts of reminiscence simply do not exist: the claim that they do is radically false. And the reason for asserting this is that, unlike acts of seeing, hearing, sensing, etc., we are not acquainted with such acts. That we are not acquainted with such acts is shown by the fact that Plato introduces them only on the basis of an argument. On grounds of acquaintance one can say that seeing, unlike reminiscence, is commonsensical and philosophically non-problematic. Nor can one go beyond this appeal to acquaintance, e.g., by “analyzing” what acquaintance is, or by setting up criteria about what can and cannot be presented – or at least, one cannot go beyond this appeal to acquaintance if the metaphilosophy that we are examining is adequate. 9.Though, oddly enough, some of them consider themselves empiricists. 10.Cf. G. E. Moore’s discussion of this view in his Some Main Problems of Philosophy (London: Allen and Unwin, 1953), p. 252. See also Russell’s discussion of Meinong’s views in “Meinong’s Theory of Complexes and Assumptions,” Mind, ns 13

94 (1904), pp. 204-219, 336-354, 502-524. 11.Cf. J. N. Findlay, Meinong’s Theory of Objects and Values, second edition (London: Oxford University Press, 1963), p. 30ff; and R. Grossmann, The Structure of Mind (Madison, Wisc.: University of Wisconsin Press), pp. 122-23. 12.This point has been emphasized by Bergmann; see his “Two Criteria for an Ideal Language,” Philosophy of Science, 16 (1949), pp. 71-74. 13.Cf. J. L. Austin, “Are There A Priori Concepts?” in his Philosophical Papers (London: Oxford Unveristy Press, 1961), pp. 1-22. 14.It is perfectly clear that Plato himself begins from these premises; cf. Theaetetus, 184a-186e. 15.Cf. G. E. Moore, “Are the Characteristics of Particular Things Universal or Particular?” 16.Cf. A. Meinong, “Hume Studies (I)”, translated in K. Barber, Meinong’s Hume Studies: Translation and Commentary (Ann Arbor, Michigan: University Microfilms, 1966); W. Sellars, “Particulars,” in his Science, Perception and Reality (London: Routledge and Kegan Paul, 1963), pp. 282-297; G. F. Stout, “The Nature of Universals and Particulars,” Proceedings of the British Academy, 19 (1921-22), pp. 157-72. Compare J. Weinberg, “The Nominalism of Berkeley and Hume,” in his Abstraction, Relation and Induction (Madison, Wisc.: University of Wisconsin Press, 1965), and F. Wilson, “Weinberg’s Refutation of Nominalism,” Dialogue, 8 (l969), pp. 460-474 17.If similarity is construed as an external relation then one cannot avoid universals, as Russell, for example, pointed out; cf. his “On the Relations of Universals to Particulars,” in his Logic and Knowledge, ed. R. C. Marsh (London; Allen and Unwin, 1956), pp. 103-124. See also H. Hochberg, “Russell’s Proof of Realism Reproved,” in his Logic, Ontology, and Language (Munich: Philsophia Verlag, 1984), pp. 196-203. 18.Cf. G. E. Moore, Lectures on Philosophy, pp. 56-57; and “Are the Characteristics of Things Universal or Particular?” See also, “Universals, Particulars, and Tropes: The Role of a Principle of Acquaintance in Ontology,” in this volume, below. 19.That such a scepticism results is hinted at by Austin, in his “Are There A Priori Concepts?” 20.We exclude by means of PA certain categories; others we include on the same basis. Those we include we use to solve or dissolve the traditional ontological problems. There is an obvious danger (which could be and indeed has been at times raised to the level of an objection, e.g., against those who claim to be acquainted with bare particulars) that either the rejection of a category or an inclusion of a category is not the result of acquaintance but rather of the dialectical pressures the claims are designed to relieve. The appeal to acquaintance is prompted by the philosophical problems; that is

95 why PA is a metaphilosohical principle. The danger is clear, hidden bias might be introduced, and all appeals to acquaintance become suspect. (For discussion of this possibility with regard to so-called bare particulars, see “Effability, Ontology, and Method,” in the present volume, below.) Compare the metaphilosophy of the later Wittgenstein (cf. O. K. Bouwsma, “The Blue Book”). Philosophical uses are held to be problematic because they involve the misuse of the words of ordinary (philosophically non-problematic) language resulting from our being misled by hidden, that is, unnoticed, grammatical analogies. Wittgenstein’s job was to accurately describe usage so that the misleading features would come to be recognized for what they are. In turn, the philosophically problematic would disappear. Just as in the analogous case of the appeal to PA in the metaphilosophy that we are considering, the appeal to a description of ordinary language is prompted by the philosophical problems. Here, too, the question arises: Are the descriptions unbiased? Nothing guarantees it, in either case. The point is, of course, that it would be unreasonable to demand such a guarantee. The best that one can ever be expected to do is to guard against the possibility of bias. As Allaire once put it In this regard all philosophers are in the same boat. They all start from what they consider unproblematical or, as it is sometimes put, from what they hold to be commonsensical. With respect to their starting point they must always be vigilant. One cannot do more; one must not do less. (E. B. Allaire, “Bare Particulars,” Philosophical Studies, 14 [1963], p. 4) 21.E.g., A. J. Ayer, The Foundations of Empirical Knowledge (London: Macmillan, 1940), pp. 84-86; B. Russell, An Inquiry into Meaning and Truth (New York: Norton, 1940), p. 137ff. 22.Cf. his Foundations of Empirical Knowledge. 23.Cf. W. Sellars, “Empiricism and the Philosophy of Mind,” in his Science, Perception and Reality, pp. 127-196. There are some (in correspondence, Richard J. Bernstein has made this point) who are genuinely perplexed why I and others who are sympathetic with the metaphilosophy being discussed think that recent criticisms of PA really leave the principle unscathed. My basic reply to these persons is that I am genuinely perplexed why they think that these recent (attempted) criticisms are at all relevant to PA, why they believe that these attempts at all touch the principle. More accurately, I do not believe these criticisms touch the principle in its ontological use. Thus, Sellars’ well publicized criticisms suggest that there is a move from a commonsensical use of ‘acquaintance’ to a technical sense in which in which acts of acquaintance are held to be incorrigible. The point is, it is the commonsensical, and not the technical sense of ‘acquaintance’ which is relevant to ontology. Sellars may in fact successfully attack the technical

96 sense in which acts are incorrigible, but that has nothing to do with PA as an ontological principle. Or, at least, such is the thrust of the argument of the present study. It should be added, however, that Sellars’ attack is more complicated than the above mentioned comment might indicate. The other main aspect of it involves a de facto denial that there are any mental acts at all. Then there is his nominalism. The first of these themes will be discussed in Appendix Three of the present study, below; and both these themes will be examined, below, in subsequent studies. (See especially “Universal, Particulars, and Tropes: The Role of a Principle of Acquaintance in Ontology,” in the present volume, below.) But at this point we are concerned only with the argument based on the claim, implausible, we are arguing, that any use of PA requires it to appeal to incorrigible acts of acquaintance, and the claim, not implausible, that there are no such incorrigible acts. 24.Cf. Russell, Inquiry into Meaning and Truth, p. 139; and G. E. Moore, “Some Judgments of Perception,” in his Philosophical Studies (London: Allen and Unwin, 1922), pp. 226-377. 25.See also Appendices One and Three to this study, below. 26.Cf. D. Lewis, “Moore’s Realism,” in L. Addis and D. Lewis, Moore and Ryle: Two Ontologists (The Hague: Nijhoff, 1965), passim. 27.Cf. B. Blanshard, The Nature of Thought, 2 vols. (London: Allen and Unwim, 1939), vol. I, ch. xvi, p. 567ff; and W. Sellars, “Concepts as Involving Laws and Inconceivable without Them,” Philosophy of Science, 15 (1948), pp. 287-325; and many idealists when they discuss the “discursive” nature of thought, e.g., H. H. Joachim, Logical Studies (London: Oxford University Press, 1948). 28.Compare, among many others, W. Sellars, “Is There a Synthetic A Priori?” in his Science, Perception and Reality, pp. 298-320, and his “Comments on Maxwell’s ‘Meaning Postulates in Scientific Theories,” in H. Feigl and G. Maxwell, eds., Current Issues in the Philosophy of Science (New York: Holt, Reinhart and Winston, 1961), pp. 183-192; and B. Blanshard, The Nature of Thought, vol. I, ch, iv, p. 160ff. See also F. Wilson, “The Notion of Logical Necessity in Carnap’s Later Philosophy,” in A. Hausman and F. Wilson, Carnap and Goodman: Two Formalists (The Hague: Nijhoff, 1966), ch. 3; and below, in this volume, “Implicit Definition Once Again.” G. Frege makes the relevant point in his essay “On the Foundations of Geometry,” Philosophical Review, 69 (1960), pp. 3-17. 29.Thus, all thought about an object, since it does involve concepts, is inevitably discursive, in the sense of the idealists. 30.These points were noted already by Frege; see G. Frege, “On the Foundations of Geometry.” The points have also been made in F. Wilson, “The Notion of Logical Necessity in Carnap’s Later Thought,” ch. 3.

97 See also M. Brodbeck, “The Philosophy of John Dewey,” in E. B. Allaire et al., Essays in Ontology (The Hague: Nijhoff, 1963), pp. 188-215. 31.Cf. B. Russell, “The Monistic Theory of Truth,” in his Philosophical Essays, revised ed., (London: Allen and Unwin, 1966), p. 133, p. 137. 32.This rejects the position of W. Sellars, “Concepts as Involving Laws and Inconceivable without Them,” and F. H. Bradley, Essays on Truth and Reality (Oxford: Oxford University Press, 1914), p. 297, to mention only two cases. 33.Cf. Russell, “The Monistic Theory of Truth,” p. 146. 34.Cf. John Dewey, Logic: The Theory of Inquiry, p. 108ff, , p. 142, p. 184, pp. 496501; F. H. Bradley, “On Truth and Coherence” in his Essays on Truth and Reality, pp. 202-218; B. Blanshard, The Nature of Thought, vol. II, ch. xxv, and esp. p. 234f; and W. V. O. Quine, “Two Dogmas of Empiricism,” in his From a Logical Point of View (Cambridge, MA: Harvard University Press, 1953), pp. 20-46. 35.See, for example, G. Bergmann, Philosophy of Science, pp. 79-83. 36.This line of argument can be found in Quine and in Sellars. For more on this argument, see Appendix One to this study, below. 37.See below, in this volume, “Implicit Definition Once Again”. 38.Cf. F. Wilson, “The Notion of Logical Necessity in Carnap’s Later Philosophy,” Ch. 5. Also F. Wilson, “Barker on Geometry as A Priori.” Philosophical Studies, 20 (1969), pp. 49-53 39.See also Appendix One to this study, below. 40.See his Philosophy of Science, p. 81. 41.This lesson we ought all to have learned from J. L. Austin; cf. his Sense and Sensibilia (London: Oxford University Press, 1962). But Ayer was not convinced; see his “Has Austin Refuted the Sense-Datum Theory?” Synthese, 17 (1967), pp. 117-140. Ayer in turn has been criticized by L. Forguson, “Has Ayer Vindicated the Sense Datum Theory?” in K. T. Fann, ed., Symposium on J. L. Austin’s Philosophy (London: Oxford University Press, 1969). 42.Cf. Bergmann, “The Revolt against Logical Atomism,” in his Meaning and Existence (Madisson, Wisc.: University of Wisconsin Press, 1960), pp. 39-72. 43.But, to be fair, Bergmann came to abandon the position here discussed; see his “Realistic Postscript,” in his Logic and Reality (Madison, Wisc.: University of Wisconsin Press, 1965), pp. 302-340.

98

44.Cf. W. Sellars, “Is There a Synthetic A Priori?” 45.Cf. G. E. Moore, “External and Internal Relations,” in his Philosophical Studies, pp.. 276-309. 46.F. P. Ramsey puts it this way: He dismisses the view, which he attributes to Leibniz and Wittgenstein, that it is a logical truth that red and blue are incompatible; but in fact, “No one could say that the inference from ‘This is red’ to ‘This is not blue’ was formally guaranteed like the syllogism.” (“Facts and Propositions,” in his Foundations of Mathematics [London: Kegan Paul, Trench, Trubner & Co., 1931], p. 152). It was in his review of the Tractatus that Ramsey first made this clear. Wittgenstein, in the Tractatus, as Ramsey notes, did disagree about the necessity of statements of colour incompatibility. In his review of the Tractatus (reprinted in his Foundations of Mathematics), Ramsey noted Wittgenstein’s position: since induction has no logical basis, and inductive inferences therefore cannot be necessary, statements of colour incompatibility, which are synthetic, cannot be known incorrigibly to be true; but Wittgenstein also holds, it is necessary that a point in the visual field that is red is necessarily not blue (Ramsey is referring to Tractatus, 6.3751). “Hence,” Ramsey notes, “he [Wittgenstein] says that ‘This is both red and blue’ is a contradiction.” (Ramsey, Review of Wittgenstein’s Tractatus, p. 280) But the conclusion to be drawn is that what Wittgenstein shows in the Tractatus is that what we had previously thought was a necessary truth – necessary as a tautology is necessary –, the clear idea of logical necessity that Wittgenstein provides shows that he [Wittgenstien] is wrong, it is not necessary after all. Wittgenstein should give up his position that colour incompatibilities are necessary; we should simply recognize the fact, and get on with it: “...not all apparently necessary truths can be supposed ... to be tautologies.” (Ibid., p. 280.) For a good analysis of Wittgenstein’s position, see E. B. Allaire, “Tractatus 6.3751,” Analysis, 19 (1959) pp. 100-105. Allaire makes clear that a good number of misreadings of the Tractatus arise from a misreading of the passage to which Allaire (and Ramsey) refers, and a misunderstanding of the issues it raises. The misreadings derive from an uncritically accepted holism that takes for granted what Wittgenstein took for granted, that statements of colour incompatibility are logically necessary. 47.Ewing is an under-rated thinker, not much read nowadays by English philosophers. But then, if you thought that Wittgenstein was a god or Austin a genius you would have little time for Ewing. 48.A. C. Ewing, “Cause,” in E. H. Madden, ed., The Structure of Scientific Thought (Boston, MA: Houghton Mifflin, 1960), pp. 208-215. 49.“C” is a “pseudo-relation” rather than a relation because it relates, not things or properties of things, but facts or states of affairs. 50.Hume’s problems here are not concerned with any deep principles of empiricism

99 but with (a) the logic of relations, and (b) the logic of definite descriptions. Straighten these out, as one must if one is to be fair to Hume, then the problem of the missing shade disappears. Cf. F. Wilson, “Hume’s Fictional Continuant,” History of Philosophy Quarterly, 6 (1989), pp. 171-188. 51.Hume makes clear the challenge that faces the philosopher who rejects any definition of cause as constant conjunction: I know no other remedy, than that the persons who express this delicacy should substitute a juster definition in its place. But, for my part, I must own my incapacity for such an undertaking. When I examine, with the utmost accuracy, those objects which are commonly denominated causes and effects, I find, in considering a single instance, that the one object is precedent and contiguous to the other; and in enlarging my view to consider several instances, I find only that like objects are constantly placed in like relations of succession and contiguity. Again, when I consider the influence of this constant conjunction, I perceive that such a relation can never be an object of reasoning, and can never operate upon the mind but by means of custom, which determines the imagination to make a transition from the idea of one object to that of its usual attendant, and from the impression of one to a more lively idea of the other. However extraordinary these sentiments may appear, I think it fruitless to trouble myself with any further enquiry or reasoning upon the subject, but shall repose myself on them as on established maxims. (Treatise of Human Nature, ed. L.A. Selby-Bigge [Oxford: Oxford University Press, 1888] p. 187). 52.Ewing, “Cause,” p. 200, col. 1. 53.For discussion of such a position, see F. Wilson, “Wright’s Enquiry concerning Humean Understanding,” Critical Review of John Wright, The Sceptical Realism of David Hume, Dialogue, 25 (1986), pp. 747-752. 54.Ewing, “Cause,” p. 210, col. 1-2.. 55.Ewing, “Cause,” p. 209, col.2, and p. 212, col. 1-2. 56.Cf. G. Bergmann, “Purpose, Function, Scientific Explanation,” Acta Sociologica, 5 (1963), pp. 225-235; and A. Grünbaum, “Causality and the Science of Human Behavior,” in H. Feigl and M. Brodbeck, eds., Readings in the Philosophy of Science (New York: Appleton-Century-Crofts, 1953), pp. 766-778. 57.Cf. F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies (Toronto: University of Toronto Press, 1999), Study One. For further discussion of the Aristotelian account, see F. Wilson, “Explanation

100 in Aristotle, Newton and Toulmin,’ Philosophy of Science, 36 (1969), pp. 291-310 and pp. 400-428. The Aristotelian account is irremediably anthropomorphic; its nonHumean element is a result of blurring the intentional nexus with the Humean law. In other words, Ewing’s critique of Hume in his “Cause” is essentially an attempt to revive the Aristotelian patterns. 58.Cf. J. Weinberg, Abstraction, Relation, Induction, pp. 76-77. 59.Cf. L. Addis, “Ryle’s Ontology of Mind,” in L. Addis and D. Lewis, Moore and Ryle: Two Ontologists. 60.Cf. R. A. Watson, The Downfall of Cartesianism (The Hague: Nijhoff, 1966), pp. 42-43. 61.Ibid., p. 59, p. 62 62.Cf. D. Hume, Enquiry concerning Human Understanding, ed. L. A. Selby-Bigge (London: Oxford University Press, 1902), p. 65 and p. 69ff. 58.Cf. Ewing, “Cause,” p. 210, col. 2. 63.The term ‘gappy’ is J. L. Mackie’s; see his “Causes and Conditions,” in E. Sosa, ed., Causation and Conditionals (London: Oxford University Press, 1993). The term ‘imperfect’ is G. Bergmann’s; see his Philosophy of Science, ch. 2. See also M. Brodbeck, “Explanation, Prediction and ‘Imperfect’ Knowledge,” in H. Feigl and G. Maxwell, eds., Minnesota Studies in the Philosophy of Science, vol. III (Minneapolis, Minn.: University of Minnesota Press, 1962), pp. 231-272; and F. Wilson, Explanation, Causation and Deduction, (Dordrecht, Holland: D. Reidel, 1985). 64.This is the point of Hume’s two definitions of ‘cause’; see Enquiry concerning Human Understanding, pp. 76-77, and Treatise of Human Nature, p. 187. For discussion of Hume’s point, see F. Wilson, Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1992), Chapter One. 65.Thus, Hume’s first definition of ‘cause’ as regularity or constant conjunction: “...we define a cause to be an object precedent and contiguous to another, and where all the objects resembling the former are placed in a like relation of priority and contiguity to those objects that resemble the latter...” (Treatise, p. 187) But “accidental” generalities are also constant conjunctions. And so, ontologically, they do not differ in form. 66.Here is Hume’s second definition of ‘cause’: “A cause is an object precedent and contiguous to another, and so united with it that the idea of the one determines the mind to form the idea of the other, and the impression of the one to form a more lively idea of the other.” (Treatise, p. 187) Here we have, first, the psychological state (we have a causal generality when the mind is “determined” to use it in certain ways),

101 namely, second, to support the assertion of subjunctive conditions (“the idea of the one determines the mind to form the idea of the other”), and, third, to make predictions (“the impression of the one to form a more lively idea of the other”). We take up these Humean points again, later in this essay. Compare N. Goodman, Fact, Fiction and Forecast (London: University of London Press, 1954), for another version of the Humean position. 67.Cf. F. Wilson, “The Notion of Logical Necessity in Carnap’s Later Philosophy,” ch. 5. 68.Cf. G. H. von Wright, The Logical Problem of Induction, second ed. (Oxford; Oxford University Press, 1957), ch. xi. 69.Cf. Sextus Empiricus, Outlines of Pyrrhonism, trans. R. G. Bury (Cambridge MA: Harvard University Press, 1961), Book I, ch. vii and x, and Book II, ch. xii. 70.This is perhaps a little more controversial than the similar claim, we just saw was made, about Sextus.. (What? Hume not a sceptic? Nor Sextus? How plausible is such a claim?? The author has gone mad!) The interpretation of Hume that is here noted has been defended at length in F. Wilson, Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1997). R. H. Popkin would disagree; see his “David Hume: His Pyrrhonism and his Critique of Pyrrhonism”, Philosophical Quarterly, 1 (1951), pp. 385-407; Popkin’s reading is disputed in detail in F. Wilson “Is Hume a Sceptic about Reason?” Philosophy Research Archives, 10 (1984), pp. 275-320; “Is Hume a Sceptic with regard to the Senses?” Journal for the History of Philosophy, 27 (1989), pp. 49-73; and “Was Hume a Subjectivist?” Philosophy Research Archives, 14 (1989), pp. 247-82. The reading I have proposed in the essays mentioned in effect places Hume at the end of a line of thinkers in Britain who developed a mitigated scepticism or, as I would prefer to put it, a commonsensical philosophically non-problematic scepticism or, better yet, a falliblism, opposed to the dogmatism of transcendental claims, either of philosophy or of religion. Cf. H. G. van Leeuwen, The Problem of Certainty in English Thought (The Hague: Nijhoff, 1963). 71.Hume, Treatise of Human Nature, p. 187. Compare also Bayle’s remark B to the entry Pyrrho in his Dictionaire (Pierre Bayle, Historical and Critical Dictionary, Selections, trans. and ed., R. H. Popkin [Indianapolis IN: Bobbs-Merrill, 1965], p. 194f): Pyrrhonism is dangerous in relation to this divine science, but it hardly seems so with regard to the natural sciences or to the state. It does not matter much if one says that the mind of man is too limited to discover anything concerning natural truths ... concerning ... the causes producing heat, cold, the tides, and the like. It is enough for us that we employ ourselves in looking for probable hypotheses and collecting data. I am quite sure that there are very few good scientists of this century who are not

102 convinced that nature is an impenetrable abyss ... Thus, all these philosophers are Academics and Pyrrhonists in this regard. 72.John Locke, Essay concerning Human Understanding, ed. P. H. Nidditch (Oxford: Oxford University Press, 1975), Bk. II, ch. vii, secs. 19-21. 73.Cf. R. A. Watson, The Downfall of Cartesianism, p. 170ff. 74.Compare the full title of Berkeley’s Principles: A Treatise concerning Principles of Human Knowledge: wherein the chief causes of error and difficulty in the Sciences, with the Grounds of Scepticism, and Irreligion in the Sciences, are Inquired into. (See his Works, ed. A. A. Luce and T. E. Jessup [London and Edinbugh: Nelson, 1949], vol. II, p. 1). Again, the Three Dialogues are written, according to their full title, in opposition to Sceptics and Atheists; see his Works, vol. II, p. 147. 75.Cf. Berkeley, Principles, paras. 3-4, para. 8. Compare E. B. Allaire, “The Attack on Substance: Descartes to Hume,” Dialogue, 3 (1965), pp. 284-287. 76.Berkeley, Principles, paras. 29, 30, 33; also paras. 37, 38. 77.Principles, paras. 31, 65. 78.Principles, para. 29. 79.Berkeley’s problem is shaped by the fact that we have no ideas of mind or spirit, i.e., we are not acquainted with it. He does say that we must know it because we know how to use such words as ‘I’ (cf. Principles, paras. 27, 139); but that argument, as Oxford philosophers of a while ago were able to demonstrate, albeit unwittingly, is notoriously bad. Berkeley does go these Oxford philosophers one better, however, in that he does make a few remarks, not entirely pellucid, about knowing this self by way of notion; cf. Principles, paras. 27, 89, 135-140. See also the study below, in this volume, “On the Hausmans’ ‘New Approach to Berkeley’s Ideal Reality’.” 80.Principles, paras. 90, 91. 81.For further discussion of the inherence pattern, see below, in this volume, “On the Hausmans’ ‘New Approach to Berkeley’s Ideal Reality’.” 82.Principles, paras. 90-91. 83.Cf. E. B. Allaire, “Berkeley’s Idealism,” Theoria, 29 (1963), pp. 229-244; P. Cummins, “Perceptual Relativity and Ideas in the Mind,” Philosophy and Phenomenological Research, 24 (1963), pp. 201-214; and P. Cummins, “Berkeley’s Likeness Principle,” Journal of the History of Philosophy, 4 (1966), pp. 63-69. Also F. Wilson, “On the Hausmans’ ‘New Approach to Berkeley’s Ideal Reality’,” below, in the present volume.

103

84. See G. E. Moore, “The Refutation of Idealism,” in his Philosophical Studies, pp. 130; also below, in the present volume, “Moore’s Refutation of Idealism.” See also G. E. Moore,” The Subject Matter of Psychology,” Proceedings of the Aristotelian Society, 10 (1909-1910), pp. 32-62; and D. Lewis, “Moore’s Realism,” in L. Addis and D. Lewis, Moore and Ryle: Two Ontologists, ch. I. 85.Berkeley, Principles, para. 49. 86.Berkeley, Principles, paras. 90, 91. 87.See below, in this volume, “The Aboutness of Thought,” and also “Effability, Ontology and Method.” 88.Berkeley, Principles, paras. 29, 30, 33, and also paras. 37, 38. 89.Berkeley, Principles, paras. 31, 65. 90.Berkeley, Principles, para. 49. 91.Berkeley, Principles, paras. 33, 34. 92.Cf. Berkeley, Three Dialogues concerning Hylas and Philonous, in his Works. vol. II, p. 238. 93.Berkeley, Three Dialogues, p. 262. 94.Bracken, “Berkeley’s Realisms,” Philosophical Quarterly, 8 (1958), pp. 41-53, at pp. 45-47. See also R. H. Popkin, “The New Realism of Bishop Berkeley,” University of California Publications in Philosophy, 29 (1957), pp. 1-19. 95.Cf. Thomas Reid, Inquiry into the Human Mind, in his Works, ed. W. Hamilton (Edinburgh: Maclachlan and Stewart, 1852), ch. II, secs. v, vi, and ch. IV, sec. xx. 96.Nor should I be taken as suggesting that Bracken in no way recognizes this point; compare his “Berkeley on the Immortality of the Soul,” Modern Schoolman, 37 (1960), pp. 77-94; and “Berkeley and Malebranche on Ideas,” Modern Schoolman, 41 (1964), pp. 1-15. For more on Berkeley on notions, see below, in this volume, “On the Hausman’s ‘New Approach to Berkeley’s Ideal Reality’.” 97.For greater detail on these points, see F. Wilson, The External World and Our Knowledge of It: Hume’s Critical Realism, An Exposition and a Defence (Toronto: University of Toronto Press, forthcoming). 98.A. M. Quinton, “The Problem of Perception,” Mind, n.s. 64 (1955), pp. 28-51. 99.Quinton, p. 31.

104

100.Quinton, p. 33ff. 101.Quinton, p. 39. 102.Russell, The Problems of Philosophy (Oxford: Oxford University Press, 1912), p. 9. 103.Quinton, p. 29f. 104.R. Chisholm, “The Concept of Empirical Evidence: I. ‘Appear,’ ‘Take,’ and ‘Evident’,” Journal of Philosophy, 53 (1956), pp. 722-731, at p. 729. 105.Hume, Treatise of Human Nature, p. 187. 106.Ibid., p. 730. 107.R. Firth, “The Concept of Empirical Evidence: II. Ultimate Evidence,” Journal of Philosophy, 53 (1956), pp. 732-739, at p. 737f. 108.W. Sellars, “Empiricism and the Philosophy of Mind,” in Minnesota Studies in the Philosophy of Science, vol. I, ed. H. Feigl and M. Scriven (Minneapolis: University of Minnesota Press, 1956), pp. 253-329. All page references in parentheses which are otherwise unidentified are to this essay. 109.Misleading, because it confuses inner awareness, on the one hand, which we all have all the time, with, on the other hand, systematic introspection, which was the method used by nineteenth century introspective psychologists and which occurs only rarely. For more on the latter, see F. Wilson “Some Controversies about Method in Nineteenth- Century Psychology,” Studies in History and Philosophy of Biological and Biomedical Science, 30C (1999), pp. 91-142. 110.See our discussion, above, of colour incompatibilities. 111.G. Bergmann made these points already in his essay on “Logical Positivism,” in his The Metaphysics of Logical Positivism, pp. 1-16; and in his later essay on “The Revolt against Logical Atomism,” in his Meaning and Existence, pp.39-72, at pp. 4445. 112.Perhaps A. J. Ayer, Foundations of Empirical Knowledge (London: Macmillan, 1940), is one who comes close to Sellars’ model empiricist and foundationalist. Bergmann is perhaps another, or at least Bergmann at a certain point in his career. Certainly Bergmann has defended foundationalism, and defended it in a way that leaves it open to Sellars’ critique (we have discussed Bergmann’s position and criticized it, above, in part [II] of the main body of the present study). But not even Ayer and Bergmann are quite the persons of straw that Sellars uses as his foil. I know of no one who fits the image of the Sellarsian empiricist.

105 It cannot be over emphasized that in the convolutions of Sellars’ arguments and prose it is too often lost that much of what he says is perfectly compatible with ordinary empiricism. His students, it seems, are too caught up in those convolutions to think critically about the conclusions and how they were derived. Or there are those (and I think there are many of these) who think they like the Sellarsian conclusions, accept them for whatever reason, if any, that they have, and then refer to Sellars’ essays like “Empiricism and the Philosophy of Mind” without giving these essays a serious read let alone a serious analysis, or, in other words, simply not thinking to give Sellars’ and their own claims any significant criticism. 113.Robert Grimm, “A Note on ‘Empiricism and the Philosophy of Mind’,” Philosophical Studies, 10 (1959), pp. 49-52. 114.W. Sellars, “Inference and Meaning”, Mind, n.s. 62 (1953), pp. 313-338, at p. 336. 115.Compare his remark that “universals and laws are correlative; same universals, same laws, different universals, different laws.” (W. Sellars, “Concepts as Involving Laws and Inconceivable without Them,” in his Pure Pragmatic and Possible Worlds: The Early Essays of Wilfrid Selllars, ed. with Intro. by J. Sicha [Reseda, CA: Ridgeview,1980], pp. 95-124, at p. 104). The sameness of things, he is saying, is not objective, in things themselves and independently of us, but rather depends upon the theories that we adopt. 116.Here, once again, we find that the appeal is in effect to the doctrine of “implicit definition.” But this account of meaning will not stand up to scrutiny. See below, in this volume, “Implicit Definition Once Again.” 117. “Inference and Meaning,” pp. 336-337. 118. “Inference and Meaning,” p. 336. 119.In his Treatise of Human Nature. 120.Cf. Wilson, The Logic and Methodology of Science in Early Modern Thought, Study One. 121.Hume, Treatise, p.157ff. 122.Hume, Treatise, p.158. 123.N. Malebranche, De la Recherche de la Vérité, ed. G. Rodin-Lewis (Paris: J. Vrin, 1979), vol. III, p. 124. 124.Malebranche, Recherche, vol. II, p. 200, p. 203. 125.Hume, Treatise, p.159.

106

126.Hume, Treatise, p.161. 127.Hume, Treatise, p.172; italics in original. 128.Hume, Treatise, p.94, p. 170. 129.Hume, Treatise, p.177; cf. 155. 130.Hume, Treatise, p.172; italics in original. 131.Hume, Treatise, p.97, p. 170. 132.Hume, Treatise, p.167. I have defended this account of causation in detail in F. Wilson, Laws and Other Worlds; see also F. Wilson, Explanation, Causation and Deduction. 133.Hume, Treatise, p. 169. Cf. F. Wilson, “Hume’s Theory of Mental Activity”, in D. F. Norton et al., McGill Hume Studies (San Diego: Austin Hill Press, 1979), pp. 101120. 134.This is G. E. Moore’s phrase; see his essay on the “Proof of the External World,” in his Philosophical Papers (London: Allen and Unwin, 1959), pp. 127-150. 135.Roderick Firth, “Sense Data and the Percept Theory,” Mind, n.s. 58 (1949), pp. 35-56, 59 (1950), pp. 233-235. 136.John Wild, “The Concept of the Given in Contemporary Philosophy.” Philosophy and Phenomenological Research, 10 (1940), pp. 70-71. 137.Hans Reichenbach, Experience and Prediction (Chicago: University of Chicago Press, 1938), p. 164. 138.The Unity of Science (London: Kegan Paul, 1938), pp. 45-48. 139.William James, The Principles of Psychology, 2 vols. (New York: Holt, 1896) vol. 2, p. 80. James Mill certainly treated perception as a sum of distinct sensations, but his son had already rejected that view and had accepted the point James was making; cf. F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill (Toronto: University of Toronto Press, 1990). James likely learned it from Mill. 140.James, Principles, vol. 2, p. 81. 141.John Locke, An Essay Concerning Human Understanding, ed. A. C Fraser, 2 vols. (Oxford: Oxford University Press, 1894), vol. I, pp. 185-186. 142.Ibid., p. 186.

107

143.Berkeley, New Theory of Vision, in his Works, vol.I. 144.New Theory of Vision, p. 148. 145.Ibid., p. 150. 146.Hume, Treatise, p. 190. 147.C. I. Lewis, Mind and the World-Order: Outline of a Theory of Knowledge (New York: Scribner, 1929), p. 66. 148.H. H. Price, Perception (London: Methuen, 1964), p. 151. 149. C. D. Broad, The Mind and Its Place in Nature (New York: Harcourt Brace, 1929), p. 153. 150.G. Bergmann, “Acts,” in his Logic and Reality; and R. Grossmann, The Structure of Mind (Madison, Wisc.: University of Wisconsin Press, 1964). 151.Hume, Treatise, p. 251. 152.Ibid., p. 251. 153.Ibid., pp. 252-52. 154.Ibid., p. 252. 155.Ibid., p. 252. 156.Ibid., p. 277. 157.Ibid., p. 303. 158.Ibid., p. 257. 159.Sellars, “Meaning as Functional Classification”, Synthese, 27 (1974), pp. 417-37, at p. 419. 160.Sellars, “Empiricism and the Philosophy of Mind,” p. 186. 161.Ibid., p. 187. 162.Ibid., pp. 187-88. 163.Daniel Dennett, “Mid-Term Examination: Compare and Contrast”, in his The Intentional Stance (Cambridge, MA: Bradford Books, The MIT Press, 1987), pp. 33950, at p. 341.

108

164.Paul Churchland, “Eliminative Materialism and the Propositional Attitudes,” Journal of Philosophy, 78 (198l). 165.This does not surprise us. Paul Churchland and his wife, then Patricia Smith, were students of Sellars at the University of Pittsburgh. They simply took over Sellars’ account of meaning. That account is a crucial premise, both for his work and for theirs, but it is one that they haven’t questioned. That is unfortunate since that is the place where Sellars goes fundamentally wrong. 166.Sellars, “Empiricism and the Philosophy of Mind,” p. 189. 167.This is a Pickwickean sense, of course. 168.Sellars, “Empiricism and the Philosophy of Mind,” p. 169. 169.Ibid., p. 176. 170.Cf. Paul Churchland, “Reduction, Qualia, and the Direct Introspection of Brain States,” Journal of Philosophy, 82 (1985). 171.Sellars, “The Identity Approach to the Mind-Body Problem,” in his Philosophical Perspectives (Springfield, IL: 1967), pp. 370-88, at p. 376. 172.The Scottish Tragedy, Act 2, scene 1, lines 33ff. 173.Sellars, “Empiricism and the Philosophy of Mind,” p. 191. 174.Ibid., p. 191. 175.In Moore’s Philosophical Studies; see below, in the present volume, “Moore’s Refutation of Idealism.” 176.Wilfrid Sellars, “Foundations for a Metaphysics of Pure Process: The Carus Lectures for 1977-78,” Monist, 64 (1981), p. 126. It is worth noting that even epiphenomena “make a difference”; see Wilson, Psychological Analysis and the Philosophy of John Stuart Mill, ch. 8, and also L. Addis, The Logic of Society. “Making a difference” does not require that epiphenomena, whether sense impressions or thoughts, be made somehow material, contrary to what Sellars here wrongly supposes. 177.Cf. J. T. Stevenson, “Sensations and Brain Processes,” Philosophical Review, 69 (1960), pp. 505-510.

Two Hume and Derrida on Language and Meaning* ...Language itself is menaced in its very life, helpless, adrift in the Threat of limitlessness, brought back to its own finitude at the very moment when its limits seem to disappear, when it ceases to be self-assured, contained, ... and guaranteed by the infinite signified which seemed to exceed it.1

Is this true? But before that: What does it mean? Derrida is here making a contrast between two views of language. On the view that Derrida believes to be correct, the linguistic sign, whether spoken or written, acquired its meaning, its significance, from conventions. On this view, meaning is a matter of nomos or institution rather than physis of nature. Since meaning is a matter of convention rather than nature, the sign on this view is arbitrary And in this respect, there is no distinction between the linguistic or phonic sign and the written or graphic sign.2 The other view of language is rejected by Derrida. This view of language denies that all significance is a matter of nomos. There are, rather, at least some signs the significance of which is natural. This is a view of language that Derrida, quite correctly locates in Plato. It is there already, perhaps especially, in the Meno. Socrates distinguishes knowledge from true opinions: ... true opinions, as long as they remain, and all they do is good, but they are not willing to remain long, and they escape from a man’s mind, so they are not worth much until one ties them down by (giving) an account of the reason why. And that, Meno my friend, is recollection, as we provisionally agreed. After they are tied down, in the first place they become knowledge, and then they remain in place. That is why knowledge is prized higher than correct opinion in being tied down.3

*

Originally in: Hume Studies, 12 (1986), pp. 99-121.

110 Knowledge is thus certain or incorrigible. There is no knowledge that lacks a guarantee of its truth. This knowledge is obtained through a form of knowing that Socrates argues is akin to “recollection.” The details of what recollection is need not detain us: what is important is what recollection is knowledge of. There are in fact two cases with which the Meno deals. The first of these is the case of geometrical propositions.4 In the course of Socrates’ interrogation, Meno’s slave boy discovers that a certain geometrical proposition is true. Recollection is, thus, in the first place, a discovery of facts that make propositions true. That a proposition is true is, in general, not a matter convention: truth depends upon the facts that the proposition is about, and whether those facts do or do not obtain will in general not be dependent upon any social institutions or conventions. At the same time, it is compatible with this to hold that what a proposition means is a matter of convention. That this string of marks ‘Hume ist tot’ means in German that Hume is dead is a matter of convention, just as it is a matter of convention that this string of marks ‘Hume is not dead’ means in English that Hume is not dead; but it is not a matter of convention, but of non-linguistic fact, that the former is true and the latter is false. As for Plato’s dialogue, it is of central importance to note with respect to Socrates’ exchange with the slave boy that he (Socrates) first obtains from Meno an affirmative answer to the question “Does he speak Greek?” before he begins his examination of the boy. The boy can thus be taken to understand the conventions that determine the meaning of the proposition the factual truth of which he discovers. Grasping meaning in this way is a matter of grasping conventions; but to grasp the truth is not a matter of convention. However, this is not Plato’s view, as is made clear by the second example of the sort of thing of which recollection is supposed to yield knowledge. What Socrates seeks is a definition of the concept ‘virtue’; he even gives Meno a brief course on how to give good definitions.5 And this search is, for Socrates, analogous to the search for the truth of the slave boy’s geometrical proposition. As the slave boy searches for the truth with respect to a geometrical proposition, so Socrates searches for the truth with respect to the definition of ‘virtue.’ In each case, the search terminates in the same sort of thing, to wit, the knowledge of a form (eidos).6 As the recollection of a Form provides a non-conventional answer to the boy’s quest, so the recollection of a Form will provide a non-conventional answer to Socrates’ quest. The Form of virtue thus tells us non-conventionally what is the real definition of ‘virtue’, what is its true meaning. For Socrates, then, meaning is natural rather than conventional. Moreover, this natural

111 meaning is grasped through recollection, which is a form of knowledge. This knowledge, once acquired, is incorrigible. Not only truth but the very intelligibility of all discourse derives from the Forms; and genuine understanding requires the mind to penetrate beyond the conventions of language to that source of intelligibility which, once grasped, guarantees its own unchanging solidity and incorrigibility. But of course, this intelligibility is insight into the very being of things. For, as the Phaedo puts it, “if there is anything beautiful besides the Beautiful itself, it is beautiful for no other reason than that it shares in that beautiful,”7 and, more generally, as the speakers agreed, “each of the Forms existed, and ... other things acquire their names by having a share in them ....”8 The relation between the sign ‘F’ and an object to which it applies is that of naming; it is conventional. But that convention may or may not be correct. It is correct just in case that it signifies the Form F and a shares in that Form. Thus, a’s being, what a is correctly said to be, is constituted by its participation in the Forms. It is evident that, when the mind penetrates beyond linguistic conventions to the Forms that constitute the standard of true meaning, the mind has thereby penetrated to that which constitutes the being of things. The Forms are thus at once the ground of the being of things and the source of all intelligibility of discourse about those things. It is this Platonic view of language that Derrida rejects. On this view, reading and writing as patterns of signs are “preceded by a truth, or a meaning already constituted by and within the element of the logos.”9 As the Phaedrus insists,10 one must contrast the writing of truth in the soul by the Forms and ordinary writing. The latter is “a sign signifying a signifier itself signifying an eternal verity, eternally thought and spoken in the proximity of a present logos.”11 The natural, eternal and universal writing is contrasted to writing in its literal sense which is “thus thought on the side of culture, technique, and artifice,”12 i. e., conventional. But Derrida rejects such natural meanings: all meaning is conventional. Since all linguistic signs are equally conventional, equally arbitrary, a radical distinction between the linguistic and the graphic is forbidden.13 In terms of the distinction of the Phaedrus, all linguistic signs are to be reckoned on the side of writing in the literal sense, that is, the conventional; “writing thus comprehends language.”14 This puts it a bit paradoxically, but the point is clear enough. The Platonic tradition that Derrida disputes has, of course, a long history, though it is one with variations. One of these variations is Locke

112 whose “conceptualism” is, in a sense, Platonism with the Forms fallen into the mind. For Locke, objects known by sense are objectively similar or dissimilar in certain respects. The mind forms ideas from these, the objects given in sense experience (or inner awareness), by separating various aspects from things and uniting these into abstract ideas.15 Words are used by persons to “stand .... for the reality of things.”16 But this conventional connection by virtue of which we apply the same term to several things may be correct or incorrect; it is correct just in case the usage in which the words collect objects into classes corresponds to the way in which one’s ideas collect objects into classes. In this sense, the correct objective signification of words is determined by one’s ideas; words ought to express our ideas. Or, as Locke puts it, “... it is a perverting the use of words, and brings unavoidable obscurity and confusion into their signification, whenever we make them stand for any thing, but those ideas we have in our own minds.”17 Although Locke does not hold that the objective being of things is grounded in our abstract ideas; nonetheless, what things can correctly be said to be is grounded in our abstract ideas. Thus, where Locke is a conceptualist with regard to the reality of the Forms where Plato is a realist, the two agree that the intelligibility of all discourse derives from non-linguistic entities, and that genuine understanding requires the mind to grasp these pre-linguistic and non-conventional sources of true meaning. People introduce language, Locke tells us, in order to serve a certain purpose: “The comfort and advantage of society not being to be had without communication of thoughts, it was necessary that men should find some external sensible signs, whereof those invisible ideas, which his thoughts are made up for, might be known to others.”18 If this purpose is to be achieved, settled and shared conventions are required: “This is so necessary in the use of language, that in this respect the knowing and the ignorant, the learned and the unlearned, use the words they speak (with any meaning) all alike.”19 Pufendorf was to pursue this line of thought, and connect it to the notion of a social contract. Duties, for Pufendorf, are absolute or conditional20: the former derive directly from natural law,21 while the latter “presuppose an express or tacit agreement.”22 It is a general and absolute duty of natural law that one keep these agreements23; we enter into these agreements when advantage makes it reasonable to do so.24 There are three basic compacts upon which all others are conditional: “The rest presuppose either: some human institution, based upon a universal convention, ... and introduced among men, or else some particular form of government. Of such institutions we observe in particular three: language,

113 ownership and value, and human government.”25 It is the first of these three that concerns us here. ...that the use of language be not in vain, if each were to call a thing by any name he pleased, there must be among the users of the same language a tacit convention, to designate a certain thing by a certain word and no others. For unless there has been agreement upon a uniform application of words, it is impossible to gather from another’s speech the thoughts of his mind. Therefore by virtue of that compact every man is bound in common speech to employ words according as the established usage in that language prescribes.26

The conventions of language are just descriptions of the use of words; they are also binding, in the sense of being rules that have imperatival, and, indeed, moral force. In this claim, Pufendorf is surely correct: and in this respect he goes beyond Locke in attempting to locate a source for the obligation, namely, in a prior compact, if only a compact that is tacit and therefore not necessarily temporally prior but at least logically prior. Pufendorf’s account of the moral force or binding nature of semantic rules as rooted in a social compact is thus an account which clearly presupposes that there is thought that is prior to linguistic convention. For, if Pufendorf is correct then there must be a sort of thought or discourse which is prior to language, a form of thought in which reason can move, can recognize the utility of the linguistic conventions and of the compact to enforce those conventions, and can tacitly agree to such a compact by conforming one’s language to those conventions, thereby to enjoy the goods that such conformity generates. Thus, for Pufendorf as for Plato, thought precedes language, and, indeed, all social conventions, all social institutions. Rationality is prior to (all human) discourse, and the human being as the rational animal is prior to human being as the political animal. For Derrida, of course, all this is backwards. But in making such a claim, Derrida has been preceded by Hume. Hume objects to Locke’s account of ideas on grounds that are essentially those of Berkeley,27 that the claim that they can be formed by separating in thought aspects or properties of things presented in sense experience, is a claim that is inconsistent with two other Lockean claims, namely, first, the anti-Platonist thesis that the grounds which make the application of general terms true of things, i.e. observable properties, are ontologically not separate from things, and, second, the thesis, common to all philosophers of the age, that the thinkable is possible.28 For, if an abstract idea is formed by thought separating an aspect, and what is thinkable is possible,

114 then the aspect that thought separates turns out to be separable in reality, which is contrary to the anti-Platonist thesis that properties are not separable from things. However, Berkeley’s argument is negative: it rejects one account of how it is that we acquire the capacity to use general terms, but it does not provide a positive alternative account.29 We are aware in sense experience of entities that have various sensible properties. These entities, which are not subjective30 and which Hume calls “impressions,” are thus objectively similar and dissimilar. Awareness of these sensible entities causes31 ideas to form; these ideas are themselves sensory contents, and they are copies, generally fainter, of the impressions that are their causes.32 Ideas are therefore objectively similar and dissimilar to the things that are their causes and to the other sensible entities that are similar and dissimilar to the latter. Let us say that sensible entities, impressions and ideas, that are objectively similar in a certain respect form a resemblance class of things. The question, how do we acquire the capacity to use general terms? is therefore the question, how do we acquire the capacity to apply a general term to all the members of a resemblance class of entities? Hume answers this question in terms of a scientific psychological theory of learning. The basic principle of this theory is that if x’s are regularly present in experience with a’s, then the idea (image) of an x comes to be associated with a’s in one’s mind, both those a’s which are impressions and those a’s which are ideas derived from those impressions. To say that such an association obtains is to say that whenever an a is presented, either an a impression or an a image, then the idea of an x is evoked. Note that to say the idea of an x is associated with the idea of an a is to assert a regularity, but it is a conditioned regularity, one that holds, not of all persons at all times, but only of an individual and of that individual only consequent upon the obtaining of certain conditions, namely, those of learning, and in particular regular connection in experience. Hume’s theory of learning also has a principle of generalization, if an idea of an x is associated with a’s and a’s (and ideas of a’s) resemble b’s then the idea of an x comes to be associated with b’s. Now let x be a sign, either oral or written. If it is associated in experience with an object of a certain sort (e.g., others regularly so use it in one’s experience) then it (the idea of it) will come to be associated in thought with that object and will further come to be associated in thought with resembling objects and ideas.33 In this way one acquires the capacity to apply a word to all members of a resemblance class of entities.

115 Or, in other words, words become general through the operation of the associative mechanisms. Hume has thus succeeded in explaining, or at least sketching an explanation of how we acquire the capacity to use general terms without invoking any mysterious abstract ideas. Indeed, on Hume’s account an abstract idea simply is the acquired capacity to apply a general term to resembling things. It is not so much an entity but a disposition,34 and as such solves many of the problems that confront Locke’s account of how words become general through association with abstract ideas, for example, the problem of how the generic idea of a triangle which can – somehow – contain within it all the specific sorts of triangle, scalene, equilateral., etc., and the infinity of particular sensible triangles.35 The philosophical tradition from which Hume emerged had a settled doctrine in which thought proceeded by abstract ideas; reason consisted of arranging judgments into syllogisms. Locke challenged the view that the primary premises of syllogisms are known a priori36; Hume continued that critique.37 A general proposition consists of abstract ideas joined regularly in thought. Such a regular connection is the result of processes of association.38 The judgment is justified just in case that the association has taken place in conformity to the “rules by which to judge of causes and effects;”39 or so Hume argues.40 These scientific judgments include, of course, those of psychology, that is, the science the principles of which are used to explain the forming of such judgments.41 But this aside, Hume’s view of reasoning, that it consists in the deployment of ideas, is fairly traditional. The radical move is the construal of the ideas as the abstract ideas of his own account. Thus, the conjoining of ideas becomes the conjoining of capacities to use words; and the judgment will be occurrent rather than dispositional just in case that the dispositions that the abstract ideas are actualized, that is, just in case that signs one is disposed to employ actually are imaged or tokened. Thus, for Hume, as for Derrida, as the latter put it much later, “we think only in signs.”42 What happens in learning, of course, is that people come to conform to the settled convention for the use of a sign. Through learning, the convention is passed on from older to newer members of a linguistic community. This settled convention is precisely what Derrida refers to as the “instituted trace.”43 This trace is, as Hume says, a capacity; it is a disposition.. For Hume, and for Derrida, in contrast to Plato and to Locke, an idea is not a thing, not an entity, but a disposition or capacity. These dispositions are our abstract ideas. They not only make thought possible; thought consists

116 in the ordering of these dispositions and their exercise. The acquisition of these dispositions is a contingent matter. When as members of a linguistic community we acquire these dispositions, then we have the abstract ideas that make it possible to say what things are, that is, to make predications, correct or incorrect, of things, and which make it possible to ask questions for which these judgments are answers, such questions as, Are A’s also B’s? And, What sorts of things are these? or, more briefly, What is? All this is clear enough, and sober enough, but Derrida characteristically makes it sound more paradoxical than it is: “the trace is nothing, it is not an entity, it exceeds the question What is? And contingently makes it possible.”44 In order to develop fully Hume’s account of language, one would have to go on to say much more than can here be said about the status of conventions in Hume’s philosophy of human nature, and in particular, his scientific account of human being. One would have to spell out how the language of imperatives works, and specifically its connections with motivations. An imperative, through its connections to motivations, brings about, i.e. causes, the behaviour it describes. Thus, uttering the imperative, “Bring me that slab!” brings it about that the person a at whom it is directed so acts that the descriptive proposition “x brings me the slab” is true. This simple account of imperatives has to be generalized to the case of general imperatives or rules. These rules, considered descriptively, describe conventional behaviour patterns, and considered imperatively or as norms bring about, i.e., cause conformity to these patterns.45 Hume gives a discussion of such imperatives or norms in general terms but in a certain amount of detail. This discussion emerges in his account of the conventions of property, promising or contract, and allegiance or the norms of civil society. The discussion there, while concentrating on the norms defining civil society, shows how such norms are inseparable from the language of imperatives,46 and can be extended to cover the conventions of language themselves.47 They can also be extended to cover the conventions of rational thought, that is, the conventional thought patterns that are the “rules by which to judge of causes and effects.” In each case, the conventions are first learned more or less at one’s mother’s knee, by a process of trial and error. But they are, again in the first instance, maintained because of interest: each discovers that it is in the interest of him- or herself to maintain the conventions, to conform to those conventions him- or herself and to maintain the conformity of others to those norms But he or she not only discovers that it is in his or her interest

117 but in the interest of all. So the mechanism of sympathy comes into play, and the rules thereby acquire a moral force which is at once motivating and disinterested. This moral force not only motivates oneself, strengthens one’s resolve, if you wish, to conform to the conventions, but it also, since it is disinterested, moves us to institute means to secure the conformity of others to the conventional patterns. Without becoming too serious or too detailed about these things, it can be suggested that they might include, one, bringing it about, e.g., by teaching, but also by many other means, that newcomers to one’s society, including newborns, acquire the appropriate habits, conform to the relevant conventions. And it would include, two, bringing it about, e.g., by self-discipline, that one more consistently and regularly conforms to the conventional patterns,48 In this way the conventions of language and of rationality acquire and maintain a disinterested motivating and regulating power. There is nothing mysterious about it; it is, on the whole, and in outline at least, easily seen to be consequences of Hume’s scientific theory of learning and his theory of motivation, including in the latter not only the theory of the passions but also the mechanism of sympathy. The psychology may be, as these things go, relatively primitive, but there is little that could not, with minor adjustments, be fit into contemporary psychological theorizing. Once we recognize all this, we see that Derrida’s remark about convention, that “the instituted trace is ‘unmotivated’ but not capricious”,49 is not entirely just. These conventions of language are certainly not capricious, since they are learned; it is no accident that we conform to them. And insofar as there is no real meaning of the Platonic or Lockean sort to decide which conventions are correct, then they are indeed arbitrary and in that sense unmotivated.. Nonetheless, given the interest of human beings in communication, it is in one’s interest that there be some linguistic conventions or other, and, given that one is born into a community, it is in one’s own interest and that of the other members of the community to conform to the conventions have as a matter of fact – contingent fact – become instituted. As for the norms of rationality, conforming to these serves our interest in truth – our curiosity – which itself is both disinterested and pragmatic.50 Finally both conventions of language and of rationality acquire, as we have seen, a certain moral force – a semantic and rational imperative – which is another motivating factor. For these reasons Derrida’s comment that the conventions are unmotivated is misleading. Contrary to the Platonist, we may indeed choose conventions as we please, but only some among the many that we might choose actually do please us.

118 “We think in signs”: this describes Hume’s view as well as that of Derrida. What it means is that thought presupposes linguistic conventions. And, if Hume is correct, then we conform to these conventions as a matter of learning and, more specifically, learning in a social context. For Locke, for Pufendorf, for Plato, ideas are non-linguistic, and thought and rationality are prior to a human being’s participation in society. But for Hume, thought and rationality presuppose a person’s participation in society. The presupposition is a matter of contingent fact, the facts of learning theory, but for all that, the presupposition is there. Thus, for Hume, in total contrast to his predecessors, a person’s social being is prior to his rationality. Rationality undergoes a radical revision upon this view of human being and its place in the world. Reason is the capacity to grasp the causes of things .For Hume’s predeccesors, or for Plato and his successors at least, Forms are causes, they are the reasons for things. But once the Forms are eliminated, then, clearly they cannot explain and cannot be the reasons for things. Nor, since they don’t exist, can they be grasped by the mind: reason is no longer the mind grasping the Forms. Causation becomes regularity, conformity to a general pattern yields the reason for something occurring. Reason is now the capacity of the mind to grasp matter-of-fact regularities. Human beings do have a cognitive interest in causes, they are moved by the interest of curiosity. Reason is the strategy the mind adopts to best satisfy this interest.51 There is no absolute standard that justifies this interest; people just have it – when they have it – just as they have other, various motives. So far as we can discover, the rules of empirical science, the “rules by which to judge of causes and effects,” provide the best strategy for satisfying our curiosity, so far as such satisfaction can be attained. One can give similar considerations, in terms of other human interests, to prefer the strategies of thought of empirical science, to those of the metaphysician.52 How much less exciting is this scientific exploration of the mattersof-empirical-fact to the Platonic or Cartesian notion that rationality consists in the intuition of Forms or essences that transcend the world of space and time that we know by means of our sense experience. At the same time, though exciting, this world that transcends our ordinary world, it is inhuman, proposing to humankind cognitive aims that are, so Hume argues, unattainable.53 So, if the Humean view of reason is less exciting than that of Plato, it is also better suited to our being in the world in which we find ourselves existing, more adapted to those capacities we as a matter of fact have, and it is therefore the more human and the more humanistic. Hume lays out in detail this new concept of rationality, and defends

119 the norms of empirical science as those that can provide the correct strategy for satisfying our cognitive interests, or at any rate, those cognitive interests that admit of being fulfilled Derrida, too, proclaims the need for a new “rationality”. For Derrida this new rationality is, as one would expect, a search after truth. But the truth to be sought consists in the denial of any claim, Platonist or Lockean, that truth is to be discovered by locating a transcendent realm of non-empirical entities: the truth to be defended is that truth is not transcendent. This truth is hidden from us as it were by a veil of rhetoric invoked by those who are Platonists or Lockeans to defend the doctrine that truth is to be found in a transcendent realm of non-empirical ideas or forms. The new rationality that is required is a method for uncovering this truth, for piercing and then removing the veil of rhetoric that obscures truth that there is no transcendent truth. The new method that is required is a new rhetoric that can “deconstruct” that used by the defender of the claim that truth is transcendent, a new rhetorical style that can eliminate the Platonic view of language that has it that conventions of language are and ought to be grounded in real meanings.54 This new rhetoric developed by Derrida and, following him, other “deconstructionists,” which is the new rationality aiming to remove Platonic rationality, has been characterized, reasonably so, in this way: Since deconstruction treats any position, theme, origin, or end as a construction and analyzes the discursive forces that produce it, deconstructive writings will try to put into question anything that might seem a positive conclusion and will try to make their own stopping point distinctively divided, paradoxical, arbitrary or indeterminate.55

This requires careful qualification: as it stands, it denies all possibility of reason, given the demise of Platonic reason – which does not follow. Thus, first, it is said here, and elsewhere,56 that the point of deconstruction is that all thought is provisional – the incorrigible certainty of the Platonist does not exist – but, though all knowledge claims are provisional, it does not follow that they are therefore negative or that they must be negative. Second, concerning the alleged arbitrariness of all thought, something can be arbitrary relative to Platonic standards, as the deconstructionists maintain, without being unmotivated and pointless – as we have also seen, but which the deconstructionists wrongly infer and wrongly (and perhaps inconsistently) defend. And third, one can draw a reader’s attention to one’s rejection of absolutes and rationally intuited givens while also granting that

120 one’s language, shot through as it is with Platonism and the dross of transcendental metaphysics, may make one seem arbitrary or paradoxical or divided against oneself – all this may be granted without using the rhetorical ploys of paradox, etc, to highlight this: one can present in a nonparadoxical way the fact that one does, often, fall into paradox due to the pervasiveness of a residual Platonism in the language and discourse one uses to describe the world of everyday experience. To be sure, the rhetorical ploys of Derrida and the other deconstructionists may be fun, but they are certainly not necessary; and their indulgence by Derrida et al. testify more to an adolescent aim to shock than it does to a philosophical concern for the truth. It is argued that one requires paradoxical language as the rhetorical form necessary to recognize the genuine, this-worldly truth that there really is nothing to Platonic Truth. The paradoxes appear because it is necessary to use the very language of Platonism – the spurious metaphysical discourse of Platonism – one must use this language if one is to recognize the necessity of rejecting Platonism, and recognize that to reject it requires that one free oneself from the snares that Platonism and its notion of real definition place for us in the language of everyday discourse about the ordinary world of ordinary experience. Since ordinary language and ordinary accounts of language are pervaded by the Platonic view, one must use the very language of logos to argue against it. Hume, too, recognizes this problem. There are no absolutes, the Humean argues, there are no guarantees, yet the language of, e.g., knowledge, is full of conventions that imply that such a guarantee obtains, and Hume at times must use such language in the very act of denying that there are such guarantees. All that one can do is warn one’s readers, that one too is human can fall into precluded forms of discourse in the very act of criticizing them – all one can do is give such warnings – besides, of course, apologizing for the misleadingness of the language that, in the absence of any other, one must, perforce, use in uncovering new forms of rationality. And so Hume tells his readers that ’Tis easier to forbear all examination and enquiry in so natural a propensity, and guard against that assurance, which always arises from an exact and full survey of an object. On such an occasion we are apt not only to forget scepticism, but even our modesty too; and make use of such terms as these, ’tis evident, ’tis certain, ’tis undeniable; which a due deference to the public ought, perhaps, to prevent. I may have fallen into this fault after the example of others; but I here enter a caveat against any objections, which may be offer’d on that head; and

121 declare that such expressions were extort’d from me by the present view of the object, and imply no dogmatical spirit, nor conceited idea of my own judgment, which are sentiments that I am sensible can become no body, and a sceptic still less than any other.57

Clarity in language is necessary, and is something after which one strives, imperfectly no doubt, but to the best one can. But the real task, the deeper task for Hume is to actually develop his arguments. Hume develops his arguments as a case against the rationality of the Platonist. But the closest Derrida ever comes to argument is that the empiricist view of language is a matter of historical necessity; it is the direction in which history is moving: By a slow movement whose necessity is hardly perceptible, everything for at least some twenty centuries tended toward and finally succeeded in being gathered under the name of language is beginning to let itself be transferred to, or at least summarized under, the name of writing...58 Their movement was absolutely necessary with a necessity that cannot be judged by any other tribunal.59

But this is simply to make history the logos that justifies usage; the convention of treating all language as conventional is justified because it is the non-conventional outcome of history. But history can no more provide a non-contingent guarantee than can anything else. As Culler remarks, “History is not a privileged authority....”60 For, as Hume argued, as part of his empiricist programme, there are no objective necessary connections, no Forms or essences that could provide an objective ground of either certainty or value. The norm that directs one to proceed as if all language is conventional is justified, not by metaphysical necessity, not even when that is disguised as historical necessity, but by the fact that one’s arguments make it reasonable – but not absolutely certain as Plato required61 – that there are no Forms or essences or whatnot to provide objective necessities and real meanings. Moreover, having developed his argument against the Platonist (and the Lockean) – having, in other words, deconstructed them – Hume proceeds beyond the negativity of deconstruction and proposes, as a positive claim, a defence of reason, not the reason of the Platonist, but a new reason, the new reason of the empirical science the success of which in coming to know was manifest and obvious already in Hume’s day: Hume, then, proceeds to a positive defence of the new rationality encapsulated in his “rules by which to judge of causes and effects,” the rules of experimental

122 empirical science. And still further, he employs this rationality to develop a scientific theory, his psychological theory of learning – the associationist theory of learning that will satisfy our human curiosity about the causes of human behaviour, and in particular our curiosity about how it comes about that linguistic conventions are conformed to, and are passed on from generation to generation, and, indeed, about how they come to be changed – where, it must be noted, among the conventions thus explained are the conventions of both deconstructionist practices and empirical scientific rationality. (By the way, it is not at all paradoxical that deconstructive criticism does not stand outside the domain of learned convention, contrary to what J. Culler has suggested.62) In fact, Hume proposes a theory that enables us to understand, empirically, not only how the conventions of empirical science come to be, and came to be, conformed to, but also how, psychologically, as a matter of empirical fact, the illusions of the Platonist came to be. We can construct that sort of imperfect and gappy explanation sketch that Hume called a natural history and that Nietzsche called a genealogy.63 For the Humean, as for Derrida, of course, Platonism, contrary to its own intentions, is just one set of conventions among others and where for the Platonist the physis/nomos distinction is one of kinds of conventions – real and mere – for the Humean the two sides of this dichotomy turn out in reality to be equally conventional. The conventionality of physis, located by the Humean deconstruction, is disguised from the Platonist by the illusions of philosophy. One can put this paradoxically – the non-conventional of the Platonist is merely a(nother) convention – the sort of paradox that is beloved by Derrida. One must also say, also in a way beloved by Derrida, that for the Humean the traditional dichotomy disappears, and that the rhetorical ‘merely’ in ‘merely conventional’, for all its persuasive utility, in fact loses all force it might have in directing our attention to an objective contrast; its point, rather, is to puncture the illusions of the Platonist through an element of shock. But for Hume, unlike and in fact in contrast to Derrida, the crucial point is not the paradoxical formulation nor the shocking re-formulation in empiricist terms of the traditional discourse of the Platonist; it is rather that a theory can be developed by the empiricist to enable us to understand, empirically and scientifically, both human being at its best, as a rational animal, and human being at its less-than-best, as an animal that thinks but is gripped by the illusions of religion and of Platonist philosophy (and of the historicist illusions deriving from Hegelian and Marxist ideology).

123 There is none of this Humean sort of attempt to understand in Derrida. In this respect there is in Derrida as compared to Hume a real lack of humanity and of concern to understand one’s fellows. Hume is a genuine humanist, intent upon developing an account of rationality and of human understanding that will enable one to undertake with some confidence the task enjoined by the oracle, “Know thyself.” Derrida, in stark contrast, is content to be fascinated by the shallow question of how to develop a new rhetoric – shallow because, if Hume is correct then one could develop a new and adequate rhetoric only if one has acquired an empirical psychologic, of human being. Where Derrida asserts the errors of Plato, Hume argues that they are errors. Where Derrida proclaims a new rationality which is no more than a rhetoric, Hume develops an account of reason, as strategy and as empirical science, that can penetrate below rhetoric to truths that can satisfy our reasonable cognitive interests (we may have a cognitive interest in knowing Platonic Forms but such an interest would be unreasonable – unreasonable because it is incapable of being fulfilled – at least so the Humean analysis reveals). Where Derrida delights in the paradoxes of deconstruction, Hume proceeds to try to develop an empirically adequate theory of human nature. Derrida can be fun. So, too, can Hume. But Hume is also serious.

124

Endnotes to Study Two

1.J. Derrida, Of Grammatology (Baltimore: Johns Hopkins University Press, 1976), p. 6; his italics. 2.Ibid., p.44. 3.Meno, in Plato, Five Dialogues, trans. G. M. A. Grube (Indianapolis: Hackett, 2002), 98a. 4.Meno, 82b-85c. 5.Meno, 75b-76e. 6.Meno, 72c. 7.Phaedo, in Plato, Five Dialogues, 100c. 8.Phaedo, 102b. 9.Derrida, Of Grammatology, p. 15. 10.Phaedrus, in W. Hamilton, trans., Phaedrus and Letters VII and VIII (Harmondsworth, Middlesex: Penguin, 1973), 278a. 11.Derrida, Of Grammatology, p. 15. 12.Ibid. 13.Ibid., p. 44. 14.Ibid., p. 7; his italics; cf. p. 55. 15.See J. Locke, Essay concerning Human Understanding, ed. P. Nidditch (Oxford: Oxford U niversity Press, 1979), Bk. II, ch. xii, para. 1. 16.Locke, Essay, II, ii, 5. 17.Ibid. 18.Ibid, III, ii, 1. 19.Ibid., III, ii, 3.

125 20.S. von Pufendorf, De oficio hominis et civis juxta legem naturalem Libri duo (The Two Books on the Duty of Man and Citizen according to the Natural Law), 2 vols. [v. I. Introduction by W. Schücking. Photographic reproduction of the 1682 edition. Errata in the edition of 1682; v. II. Translation (by H.F. Wright) of the introduction by W. Schücking. Translation of the 1682 edition by F.G. Moore] (Oxford: Oxford University Press, 1929), Bk. I, ch, v, para. 1. 21.Ibid. 22.Ibid., I, ix, 1. 23.Ibid., I, ix, 3. 24.Ibid., I, ix, 2. 25.Ibid., I, ix, 22. 26.Ibid., I, x, 2. 27.Cf. J. Weinberg, “Abstraction,” in his Abstraction, Relation, Induction (Madison, Wisc.: University of Wisconsin Press, 1965). 28.D. Hume, Treatise of Human Nature, ed. L. A. Selby-Bigge, rev. P. Nidditch (Oxford: Oxford University Press, 1978), p. 17, pp. 19-20. 29.Cf. A. Meinong, “Hume Studies (I)”, in K. Barber, Meinong’s Hume Studies; Translation and Commentary (Ann Arbor, MI: University Microfilms, 1967). 30.Cf. D. Livingston, Hume’s Philosophy of Common Life (Chicago, IL: University of Chicago Press, 1984); and F. Wilson, “Is Hume a Sceptic with regard to the Senses?” Journal for the History of Philosophy, 27 (1989), pp. 49-73. 31.Hume, Treatise, p. 1. 32.Cf. F. Wilson, “Hume’s Theory of Mental Activity,” in D. F. Norton, N. Capaldi, and W. Robison (eds.), McGill Hume Studies (San Diego: Austin Hill Press, 1979), pp. 101-120; and for a more general account of introspective psychology, see F. Wilson, “Some Controversies about Method in Nineteenth Century Psychology,” Studies in History and Philosophy of Biological and Biomedical Science, 30C (1999), pp. 91142. Also F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill (Toronto: University of Toronto Press, 1990), Chapter Eight. 33.Hume, Treatise, p. 11. 34.Treatise, p. 22. 35.Treatise, pp. 20-21.

126

36.Cf. F. Wilson, “The Lockean Revolution in the Theory of Science,” in S. Tweyman and G. Moyal, eds., Early Modern Philosophy: Epistemology, Metaphysics and Politics (New York: Caravan Press, 1986), pp. 65-97; and The Logic and Methodology of Science in Early Modern Thought: Seven Studies (Toronto: University of Toronto Press, 1999). 37.Cf. F. Wilson, “Hume’s Defence of Science,” Dialogue, 25 (1986), pp. 611-628; and Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1997). 38.Cf. Wilson, “Hume’s Theory of Mental Activity.” 39.Hume, Treatise, I, III, xv. 40.Cf. F. Wilson, “Hume’s Defence of Science,” Dialgoue 22 (1983), pp. 661-694; and Hume’s Defence of Causal Inference. 41.Cf. Wilson, “Hume’s Theory of Mental Activity.” 42.Derrida, Of Grammatology, p. 50. Derrida is, of course, quoting Peirce when he uses this phrase. Idealism intervened between Hume and Peirce. In Britain it was Coleridge and Sir William Hamilton, and in America these ideas were propagated by Emerson and the New England transcendentalists. Peirce’s view of Hume and of empiricism was always distorted by what he had inherited from the idealist background. As a consequence he had to re-discover for himself much of what Hume had already done; but even so, much of what he arrives at is still obscured by the idealist dross that he was never able entirely to escape. There is a connection here to Derrida, who also shares some of this sort of idealist background and who has not been able to abandon some of the Hegelian notions he learned as a youth (see, for example, the passage cited by note 58, below). 43.Derrida, Of Grammatology, p. 46. 44.Ibid., p. 75. 45.For discussion of this pattern, see F. Wilson, “Marras on Sellars on Thought and Language,” below, in the present volume; and also F. Wilson, “John Stuart Mill on Psychology and the Moral Sciences,” in J. M. A. Skorupski, ed., Cambridge Companion to John Stuart Mill (Cambridge: Cambridge University Press), pp. 203-254. 46.Hume, Treatise, pp. 551-52. Cf. P. Árdal, “Convention and Value,”in his Passions, Promises and Punishment (University of Iceland Press: Reykjavik, 1998), pp. 22-237; and F. Wilson, “Árdal’s Contribution to Philosophy,” intro. to Páll Árdal, Passions, Promises and Punishment, pp. 7-40.

127

47.Cf. Livingston, Hume’s Philosophy of Common Life, pp. 65-72, and also F. Wilson, “Association, Ideas and Images in Hume,” in P. Cummins and G. Zoeller, eds., Minds, Ideas and Objects (Atascadero, Calif.: Ridgeview Publishing Co., 1992), pp. 255-74. 48.Cf. Wilson, “Hume’s Theory of Mental Activity..” 49.Derrida, Of Grammatology, p. 47. 50.Cf. Wilson, “Hume’s Defence of Science.” 51.Cf. Wilson, “Is Hume a Sceptic with regard to Reason?” Philosophy Research Archives, 10 (1984), pp. 275-320. See also F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies (Toronto: University of Toronto Press, 1999). 52.Cf. Wilson, “Hume’s Defence of Science.” 53.Cf. Wilson, “Hume’s Cognitive Stoicism,” Hume Studies, 1985 Supplement, pp. 521-68. 54.This understanding of the new rationality as a new rhetoric is brought out nicely by J. Culler, On Deconstruction (Ithaca, NY: Cornell University Press, 1982), pp. 81-82, p. 259. 55.J. Culler, On Deconstruction, p. 259. 56.Ibid., p. 225. 57.Hume, Treatise. Pp. 273-4. 58.Derrida, Of Grammatology, p. 6. 59.Ibid., p. 7; cf. p. 14. 60.Culler, On Deconstruction, p. 130. 61.Cf. Wilson, “The Lockean Revolution in the Theory of Science”; and Hume’s Defence of Causal Inference. 62.See J. Culler, On Deconstruction, pp. 87-88, pp. 183-84, pp. 325. 63.Cf. Wilson, “Hume’s Defence of Science.”

Three Empiricism: Principles and Problems* Empiricism is defined by its basic principle, the Principle of Acquaintance (PA), which asserts that no entity is to be admitted into one’s ontology unless one is acquainted with that entity, or entities like it, in ordinary awareness of the world, either sensory awareness or inner awareness of our own conscious states. As William James once put it, this metaphysical first principle states that “everything real must be experienceable somewhere, and every kind of thing experienced must somewhere be real.”1 James was merely re-stating what others had earlier stated. David Hume was making essentially the same point when he insisted that all our ideas derive from impressions. As he argues, “... all our simple ideas in their first appearance, are derived from simple impressions, which are correspondent to them, and which they exactly represent.”2 When Hume says that all our ideas derive from impressions, what he is maintaining is that there is no kind of thought, no form of cognition, which gives us access as it were to a world that is somehow beyond the world given to us in sense experience or inner awareness. All our ideas or concepts are ideas or concepts of things in the world of sense experience, either things given to us in sense experience or inner awareness or things that could be so experienced. When Hume states that all our ideas or concepts derive from impressions, he means that they derive from entities as they exist in the world presented to us in ordinary experience. They either, as names of individuals, refer to particular things presented in experience, or, as predicates, refer to the kinds of things, where things of these kinds have been presented in experience. There can of course be kinds of kinds as well as kinds of individual things, that is, genera as well as species. * Appeared originally as “Empiricism: Principles and Problems,” in W. Sweet, ed., Approaches to Metaphysics (Dordrecht, The Netherlands: Kluwer, 2004), pp. 265302.

130 As Hume indicates, we can, using these basic concepts form other concepts. “I observe,” he tells us, “that many of our complex ideas never had impressions that corresponded to them, and that many of our complex impressions never are exactly copied in ideas. I can imagine to myself such a city as the New Jerusalem, whose pavement is gold, and walls are rubies, though I never saw any such. I have seen Paris; but shall I affirm I can form such an idea of that city, as will perfectly represent all its streets and houses in their real and just proportions?”3 If we take as basic or undefined the concepts that apply directly to things and kinds of things as these are presented in experience, then we can use these to introduce complex or defined concepts. These defined concepts permit us to think of kinds of things that are not themselves given in sense experience. We can, for example, form the concept of a unicorn. This is a kind that is not itself presented to us in sense experience, but it is a kind that is experienceable because the concept is defined in terms of kinds which have been presented in ordinary sensible experience. There are also what Hume refers to as “relative ideas,”4 for example, the idea of “the father of Caius”. Concepts of this latter sort came later to be called “definite descriptions.” The concept of a unicorn refers a kind of thing which does not exist: there are no unicorns. Things of that sort are, however, possible. Definite descriptions also enable us to refer to things that do not exist, for example, the present king of France. Again, while there is no present king of France, such a thing is possible. In enabling us to conceive of things which are possible but not actual, these concepts are to be contrasted to supposed concepts of things, for example, Platonic forms, which rationalist philosophers argue exist but are not given in sensory experience. Given the empiricist principle PA, Platonic forms will not be admitted into an empiricist ontology, nor will one be able to form a genuine concept of such entities: we lack any form of thought or cognition in which such entities are presented to us. We can think of or conceive things that do not exist. Defined concepts such as that of a unicorn and definite descriptions such as the present king of France enable us to think about what does not exist. But these concepts are all defined on the basis of concepts that do refer to things or sorts of things which are presented in sensible experience. Since all concepts which are legitimate are defined on the basis of the latter, if the concept of a thing cannot be defined on that basis, then such things are simply not possible. Thus, the basic concepts that refer to things and kinds of things that are given in ordinary sensible experience establish the limits of the

131 possible. Or to put it another way, for the empiricist the possible is contained in the actual. It was John Locke who introduced the essential empiricist themes into modern philosophy.5 The philosophers against whom Locke was arguing were the Aristotelians and the rationalists. These philosophers shared the view that causal relations have an objective necessity which, once grasped, gives the knower knowledge of causal connections that is certain, not only beyond all reasonable doubt but beyond all possibility of doubt. It was, moreover, agreed by these thinkers that knowledge of these objective necessary connections does not derive from sense experience. The Aristotelians and the rationalists argue that causally relations are objectively necessary. Thus, suppose when a thing is F then its being F causes it to be G. Then the position is that there is a connection, label it “–>”, between F and G such that there is a guarantee with regard to anything that is F that it is also G. We have, in other words, the connection (*) F –> G This connection is necessary, and guarantees that an event being F implies that there is an event that is G. If we take Fat1 to represent the event that a is F at t1, then, using the symbols of formal logic, we may represent the regularity by (**) (x)(Fxti e Gxti + 1) Since, on this position, (*) is to guarantee the truth of (**), it is also a necessary truth that (***) ( F –> G) e (x)(Fxti e Gxti + 1) Suppose that the individual a which is F is presented to us in ordinary experience. We can represent this by “a is F”, or, in the symbols of logic, by (#) Fat1 Since we are presented with the property F, and since (*) holds of F of necessity, we can discern the fact (*) in the situation (#). And since (***), too, is a necessary truth, we can infer from what we discern in (#) not only that (##) Gat2 but also the general fact or regularity (**): (x)(Fxti e Gxti + 1) We discern in a’s being F not only the connection among properties (*) but also the fact that a is determined by being F to also be G, and moreover the general fact that anything which is F will be G.6

132 On this account there will be a distinction between casual regularities, which are objectively necessary, and “mere” or accidental regularities. Both regularities have the form (**). But the causal regularities will be supported by inference from facts like (*) and (***). Accidental regularities, in contrast, have no such support in a necessary connection holding between the properties. Because of the objective necessity that ties the properties F and G to one another, it is not possible for something to be F and yet not to be G. The properties are such that if (#) Fat1 obtains then it is not possible that the negation of (##), namely, (###) ~Gat2 be true. The truth of (#) excludes the possibility of (###). What this means is that the objective necessary tie represented by (*) guarantees that one will never discover a counterexample to the generalization (**). That, of course, is but another way of saying that the causal regularity is objectively necessary. In contrast, if there is no fact of necessary connection (*), nor, therefore, a fact like (***), then the regularity – the “mere” regularity – (**) will be such that there is no guarantee against a counterexample. In other words, where there is no objective necessary connection, a regular connection will always be contingent. Since the event (#) Fat1 guarantees the existence of the event (##) Gat2 these two events are inseparable: if the one does not exist, then neither does the other exist. Where one has a regularity which is a mere regularity, then it is the separability of the properties that are regularly connected – “merely” regularly connected – that permits the possibility of counterexamples. Locke developed an empiricist case against the objective necessary connections of the Aristotelians and the rationalists, the connections that we have represented by ‘–>’.His appeal was to PA: we are acquainted in ordinary experience with the properties of things but with no connection that ties them necessarily to other properties. In terms of our schemata, we are acquainted with properties such as F and G but not with a connection –> that would make (*) true. Here is his argument: ’Tis evident that the bulk, figure, and motion of several Bodies about us, produce in us several Sensations, as of Colours, Sounds, Tastes, Smells, Pleasure

133 and Pain, etc. These mechanical Affections of Bodies, having no affinity at all with those Ideas, they produce in us, (there being no conceivable connexion between any impulse of any sort of Body, and any perception of a Colour, or Smell, which we find in our Minds) we can have no distinct knowledge of such Operations beyond our Experience; and can reason no otherwise about them, than as effects produced by the appointment of an infinitely Wise Agent, which perfectly surpasses our Comprehensions....7

Properties are perceived to be just as they are, in themselves; to know them as they are we need not know any of the relations in which they stand to other entities. ... the immediate perception of the agreement or disagreement of identity being founded in the mind's having distinct ideas ... affords us as many self-evident propositions, as we have distinct ideas. Every one that has any knowledge at all, has as the foundation of it, various and distinct ideas: And it is the first act of the mind (without which it can never be capable of any knowledge) to know every one of its ideas by itself, and distinguish it from others. Every one finds in himself, that he knows the ideas he has; that he knows also, when any one is in his understanding, and what it is; and that when more than one are there, he knows them distinctly and unconfusedly one from another.8

Locke’s appeal to an empiricist’s Principle of Acquaintance is clear.9 Given that the empiricist principle PA is a principle in ontology, this is an ontological argument: from the fact that the necessary connections are not presented in ordinary experience, the appeal to PA means that that sort of connection does not exist: it is not to be admitted into any ontology that pretends to be empiricist. It follows that properties are all ontologically separable, and therefore logically separable. And in the absence of the excluded necessary connections, it follows that events such as (#) and (##) are logical atoms: they have no ontological, and therefore no logical, connections to each other or to other events. Or, to put it another way, since the events have no necessary connections to other events, they are ontologically or logically self-contained. Since the events are logically separable, the truth of (#) is compatible with both the truth of (##) and the falsity of (##). Though (#) be true, there is no guarantee that (##) be true. Even if the regularity (**) holds as a matter of fact, it is only contingently true. To put this point another way, since there is no objective necessary connection that makes (*) true, neither can (***) be a truth. The distinction of the Aristotelians and the rationalists be-

134 tween causal regularities, which are objectively necessary, and accidental generalities, which are not, disappears. Furthermore, it is no longer possible to find in a single instance the guarantee that a causal regularity is true. Since the regularity is true of a population while all we ever observe is sample, and nothing in the sample guarantees the truth of the regularity, it follows that the evidence on the basis of which the regularity is believed to hold must always be partial. It can, therefore, never be asserted with complete certainty: there will always be a logical gap, as it were, between evidence and assertion. Thus, on the empiricist ontology based on PA, casual judgments will always be tentative. These features of an empiricist ontology were clearly recognized and defended by Hume. He argued in particular that all causation is regularity. He developed the argument against objective necessary connections in two ways. Both arguments are based on PA. The first argument is to the effect that we are not acquainted with any entity that might reasonably be said to be a power the exercise of which establishes a necessary connection between events. As Hume puts the point, “All our ideas are derived from and represent impressions. We never have any impression that contains any power or efficacy. We never therefore have any idea of power.”9 The second argument also appeals to PA. This argument is of a piece with Locke’s argument, that properties are presented as not standing in necessary connections to one another. Hume has two versions of this argument. The first version of this argument turns on the logical separability of events. If we take the idea of some cause and the idea of its effect, then there is no contradiction in supposing the former to exist and the latter not: “the actual separation of these objects is so far possible, that it implies no contradiction nor absurdity; and is therefore incapable of being refuted by any reasoning from mere ideas; without which ’tis impossible to demonstrate the necessity of a cause.”10 There is no contradiction in separating the ideas because these ideas derive from perceptions and in perception there is no necessary connection that is presented to us: “as all our ideas are deriv’d from impressions, or some precedent perceptions, ’tis impossible we can have any idea of power and efficacy, unless some instances can be produc’d, wherein this power is perceiv’d to exert itself.... [But] these instances can never be discover’d in any body...”.11 This argument is simply an application of Locke’s argument that properties given in sense experience are separable in the sense of being logically self-contained. Sup-

135 pose the event (#) Fat1 causes the event (##) Gat2 On this argument, the properties F and G are logically separable, and therefore the event (#) is logically compatible not only with (##) but also with (###) ~Gat2 There is therefore no guarantee that a’s being F will bring about a’s being G. Hume’s second version of this argument turns on a more stringent account of separability, call it structural separability. On this notion of separability, the event (#) Fat1 is separable from the event (##) Gat2 just in case that a which is F at t1 would exist unchanged even if a at t2 were to not exist, nor, therefore even if the event of a being G at t2 were not to exist. Hume holds that events given in sense experience are separable in this sense also. He puts the point this way, ... as all distinct ideas are separable from each other, and as the ideas of cause and effect are evidently distinct, it will be easy for us to conceive any object to be non-existent this moment, and existent the next, without conjoining to it the distinct idea of a cause or productive principle. The separation therefore of the idea of a cause from that of a beginning of existence, is plainly possible for the imagination; and consequently the actual separation of these objects is so far possible, that it implies no contradiction nor absurdity...12

Or, as he says elsewhere, ... upon the whole, there appears not, throughout all nature, any one instance of connexion, which is conceivable by us. All events seem entirely loose and separate. One event follows another; but we never can observe any tye between them. They seem conjoined, but never connected. And as we can have no idea of any thing, which never appeared to our outward sense or inward sentiment, the necessary conclusion seems to be, that we have no idea of connexion or power at all...13

The claim that events are separable in this second sense has implications for metaphysics that are not made by the claim that events are separa-

136 ble in the first sense. Suppose that there is a relation R that holds between the two events: (&) R(at1, at2) In that case, the property (&&) R(xt1, at2) holds of a at t1. But (#) is structurally separable from (##). That means that, even if a at t2 were not to exist a which is F at t1 would exist unchanged. But if a at t2 were to cease to exist, so would the fact (&). And if (&) ceases to exist then the property (&&) ceases to hold of a at t1. Hence, given the structural separability of the two events, there can be no property (&&) which a at t1 ceases to have. And that in turns means that (&) cannot exist. The point is general: if the two events are structurally separable then there can be no genuine relations that hold between the two events. If a relation R holds between the two events, then it must be reducible to non-relational properties, called the foundations of the relation. That is, (&) must be logically equivalent to (&&&) r’(at1) & r”(at2) Hence, if events are structurally separable, there can be no genuine relations; all relations must be reducible to non-relational foundations. Hume’s second argument for the non-existence of objective necessary connections thus has the metaphysical implication that there are no genuine relations. Hume accepts this conclusion: he tells us “that all our distinct perceptions are distinct existences, and that the mind never perceives any real connexion among distinct existences.”14 This is not an implication of the first argument, based on the weaker notion of logical separability. On this notion the two events (#) and (##) are logically separable because one can both assert (#) and deny (##), i.e., assert (#) and assert (###) which is the negation of (##). But it is logically consistent to assert both (#) and (&) and to deny (##). So the logical separability of (#) and (##) does not have the implication that a relational fact like (&) cannot exist, or, if it does, that it must be reducible to nonrelational foundations. The Aristotelians and the rationalists had in fact agreed in part with these empiricist theses and arguments with regard to the entities given in sense experience. They argued that in fact Locke was correct: the objective necessary connections are not given in sense experience. But, they also argued, they are given in another form of experience, in what can be called a “rational intuition” – ‘rational’ because it discerns the reasons for things and ‘intuition’ because it grasps the connections like (*) among properties.

137 But the Aristotelians and the rationalists disagreed as to how we come to have such a rational intuition. The Aristotelians argued that we come to have such an intuition through a process of “abstraction” in which the active intellect as it were lifted from the events given in sense the formal facts like (*).15 The rationalists argued that such a process of abstraction was simply not possible. Descartes made the case using the example of wax. During a causal process of melting, wax goes through a series of changes. He first notes the changes in the sensible qualities such as colour, taste and smell. These all change while the wax remains the same. What is it then in this bit of wax that we recognize with so much distinctness? Certainly it cannot be anything that I observed by means of the senses, since everything in the field of taste, smell, sight, touch, and bearing has changed, and since the same wax nevertheless remains.16

It follows that there are no necessary connections that link these to the wax: one can, as Descartes says, “reject” these while affirming the wax. He proposes to do just this in order to discover just what, essentially, the wax is. Let us consider it attentively and, rejecting everything that does not belong to the wax, see what remains.17

What remains the same throughout is that the wax is extended and capable of change in shape. Certainly nothing is left but something extended, flexible, and moveable.18

But this extended thing during the process of melting undergoes a series of changes in shape. This series is a continuous series. Since this series is continuous it contains an infinite number of shapes. The formal facts that provide the links between the parts of the process are therefore infinite in number. But sense is finite, and can discern only a finite number of shapes. The formal facts which link the parts of the process could never all be lifted from the events that are presented to us in sense experience. No process of abstraction could provide the knowledge of the necessary connections that link the parts of the process. But what is meant by flexible and moveable? Does it consist in my picturing

138 that this wax, being round, is capable of becoming square and of passing from the square into a triangular shape? Certainly not; it is not that, since I conceive it capable of undergoing an infinity of similar changes, and I could not compass this infinity in my imagination. Consequently this conception that I have of the wax is not achieved by the faculty of imagination.19

The argument from the continuity of processes in which the parts are linked by necessary connections is a strong one. It turns on the mind not being able through sense experience to grasp the structure of such an infinite continuous series. Descartes therefore concluded that the knowledge of the objective necessary connections that link events is innate. The rational intuitions of the connections of things are therefore always present in the mind, and learning is a matter not so much of discovery but of bringing to the level of conscious or explicit awareness what is already there in the mind implicitly. Since the Aristotelians and the rationalists agreed that objective necessary connections are not given in sense experience, it follows that the claim by empiricists such as Locke and Hume that these are not so presented is one with which they would agree. Where they would disagree is with regard to the limitation of ontology to the entities that are given in sense experience. There are, they argue, further entities, such as necessary connections, that are not given in sense experience (or inner awareness), though they are, they also insist, otherwise given in experience, namely, by rational intuition. What Locke and Hume insist upon is that there are no such rational intuitions. They agree with Descartes that the Aristotelian account of how we acquire knowledge by abstraction is not acceptable. As for innate ideas, Locke provides a systematic sceptical argument that there simply are no innate ideas of necessary forms such as (*) nor innate knowledge of principles such as (***). Hume takes for granted this argument against innate ideas, taking for granted that without innate ideas the appeal to PA for the absence of necessary connections is conclusive. ...’tis impossible we can have any idea of power and efficacy, unless some instances can be produc’d, wherein this power is perceiv’d to exert itself. Now as these instances can never be discover’d in body, the Cartesians, proceeding upon their principle of innate ideas, have had recourse to a supreme spirit or deity, whom they consider as the only active being in the universe, and as the immediate cause of every alteration in matter.20

139 Later thinkers objected to the empiricist ontology that Hume was defending. Specifically, they objected to the consequence of Hume’s empiricist arguments for the separability of things that there are no relations. To the contrary, it was argued, the world as we experience it is a structured world. In other words, relations are presented to us. Here is what the idealist F. H. Bradley says: We must get rid of the idea that our mind is a train of perishing existences, that so long as they exist have separable being, and, so to speak, are coupled up by another sort of things which we call relations.21

What we are presented with is not a whole consisting of separable parts but a whole in which the parts are so related that they are not separable. If we turn to what is given this [a train of perishable existences] is not what we find, but rather a continuous mass of presentation in which the separation of a single element from all context is never observed, and where, if I may use the expression, no one ever saw a carriage, and still less a coupling, divided from its train.22

The empiricists’ ontology of relations that is part of Hume’s metaphysics makes it impossible to recognize this fact, that what we are presented with in sensation and thought are wholes in which the parts are stand in relations to one another. But if we are presented with relations, then they must, on empiricist grounds, be admitted to one’s ontology. The empiricists’ Principle of Acquaintance requires the inclusion of relations in one’s ontology. However, the arguments developed by the empiricists in criticism of objective necessary connections require the non-existence of genuine relations. It follows that the empiricists’ Principle of Acquaintance imposes an ontology which is inconsistent with the empiricists’ arguments against objective necessary connections. The idealists concluded that one had therefore to admit into one’s ontology objective necessary connections, for otherwise relations would be excluded. Bradley proposes that genuine relations are incompatible with the independence or separability that is a consequence of view of relations consequent upon Hume’s arguments against necessary connections. He tells us that “...a mode of togetherness such as we can verify in feeling destroys the independence of our reals.”23 Bradley therefore proposes an ontology that acknowledges the reality of relations. Specifically, he argues that “Relations are unmeaning except within and on the basis of a substan-

140 tial whole, and related terms, if made absolute, are forthwith destroyed.”24 On this account, unlike that implied by Hume’s argument regarding structural separability, there are relations that are genuine in the sense that their relata are not independent, or, equivalently, in the sense that the being of one relatum is not separable from the being of the other relatum. If it [a relation] is to be real, it must be so at the expense of the terms, or, at least, must be something which appears in them or to which they belong. A relation between A and B implies really a substantial foundation within them. This foundation, if we say that A is like to B, is the identity X which holds these differences together. And so with space and time – everywhere there must be a whole embracing what is related, or there would be no differences and no relation. It seems as if a reality possessed differences A and B, incompatible with one another and also with itself. And so in order, without contradiction, to retain its various properties, this whole consents to wear the form of relations between them.25

Bradley concludes against the account implied by the claim that things are structurally separable that “there must be a whole embracing what is related.” From this he infers his own account of the nature of this whole. The relata a and b are different things within a whole (a, b). This whole then “consents to wear the form of a relation”; thus, if a and b stand in the relation R, then the correct representation of this fact consists in attributing a property corresponding to R, say r, to (a, b). Thus, according to Bradley's account, the correct way to represent the fact reported by (@) a is R to b is given by (+) (a, b) is r or, r(a, b) Like the empiricist account of relations, this account reduces statements such as (@) that apparently have two subjects to statements which have only one subject. The traditional empiricist account, however, does this in a way that makes the subjects of predication the two individuals a and b; the result is two facts in which these individuals are separable. In contrast, the account that Bradley gives makes the subject of predication an individual or particular thing, only now it is a whole, a single thing of which a and b, the apparent subjects of (@), are but aspects and not the real subjects of predication. On Bradley’s view, then, the whole (a, b) is itself a particular thing,26 of which the two terms a and b are but aspects, and where the arrangement

141 r characterizes this whole: it is the form which this whole takes. But this whole consists of the relata as parts. Thus, the relation holds of the relata, not separately as in the account implied by the structural separability argument, but jointly27: “where the whole, relaxing its unity, takes the form of an arrangement, there is co-existence with concord.”28 In another sense, however, the relational form r on this account is itself part of the experienced whole, that is, the whole which we experience when we experience a relational state of affairs of the sort that we express in ordinary terms as (@). Bradley makes this point in his late essay on “Relations”: Certainly every content and aspect of the relational situation as an experienced fact may and must be taken as qualifying in some sense the situation as a whole; and, without so much as this, we cannot have a relation at all. But you cannot take the particular terms as thus qualifying the relation, even if you could take them, so far as they are particular or individual, as qualifying the whole. In short, to experience a relational situation as one whole and one fact, you must take it so that, as relational, the whole is not, and cannot be, qualified by its aspects or parts. The relation, as soon and so far as the whole situation has become relational, has become no more than one of the parts. And to regard this part as itself the entire whole is an obvious absurdity.29

If we take our experience of relational facts as given, then our experience of them is as wholes. That is, the experience that we ordinarily express by statements like (@) is an experience of a whole. But within this whole it is possible to distinguish parts or aspects. On the one hand, there are the entities that are related, a and b. And on the other hand, there is the way in which they are connected, what in (@) is represented by R. These three things are within the experienced unity or whole. They are not qualities separable from the experienced whole. Nor, as Bradley emphasizes in the last sentence, can the structuring relation be regarded as the whole of what is experienced: one cannot ignore the things that it structures. But these parts are all inseparable parts of what is taken ontologically to be a seamless whole. This has been characterized as an “absolutist” view of things. It is clear that Bradley in an important way accepts Hume’s conclusion, based on the structural separability argument, that things which are real do not stand in genuine relations to other things, that is, to other distinct things. William James thus quite correctly notes that “Taken thus in all its generality, the absolutist contention seems to use as its major premise Hume’s notion ‘that all our distinct perceptions are distinct existences, and that the mind never perceives any real connexion among distinct exis-

142 tences’.”30 What Bradley does with regard to these experienced wholes or unities is argue for a certain ontological account of them. This account consists in his rendering of the experienced fact represented by (@) as being more perspicuously represented by (+), that is, his account of relations as the form of the whole which has a and b as its aspects. But at the same time he also argues that substantial whole (a, b) is not given in ordinary experience. In experiencing, say, a being to the left of b, we experience two particular things, a and b. We also experience the unity of a being to the left of b, the sort of unity represented by (@). But Bradley asserts that there is a further entity, a further particular thing, present in the situation. That is the whole (a, b) that is the subject of the arrangement. This further whole is not given in ordinary experience. In ordinary experience we are presented with two individual things, a and b, and are not presented with what is as it were a third particular thing of which these two are but parts or aspects. Bradley accepts this point. He therefore argues that this whole, this thing of which the two presented things are but aspects, is, rather, something discovered by the faculty that Bradley calls “thought.” The object of thought is not something given in sense experience: “That it [the object of thought] is not mere sense-experience should be a commonplace.”31 Rather, “judgement, on our view, transcends and must transcend that immediate unity of feeling upon which it cannot cease to depend.”32 Thought “grows from, and still it consists in, processes not dependent on itself. And the result may be summed up thus; certainly all relations are ideal, and as certainly not all relations are the product of thinking.”33 And since relations are in effect not known by sense experience our knowledge of them is a priori; and since they are all ideal, transcending the entities of sense experience, they hold necessarily. Bradley thus lines up with the rationalists such as Descartes in holding that objective necessary connections that we know are known by a cognitive means other than sense experience (or inner awareness). He appeals to our experience of things – we experience them as related or structured. This appeal is used against the claim made by empiricists such as Hume that there are no relations – this is implication of the claim that things are structurally separable. But then it is argued that this experience is not of the sort which the empiricists would allow. The idea is that, since sense experience (and inner awareness) is of separable particulars, the structure of which we have experience cannot be given in ordinary sensible

143 experience (or ordinary inner awareness). The experience in which it is given must therefore be some non-sensible form of experience. This form of experience is what Bradley calls “thought.” There were two responses to Bradley’s critique of empiricism and to his alternative ontology. These two responses came from William James and Bertrand Russell. Both involve rejecting Hume’s claim that things given in experience are structurally separable. That permits these philosophers to accept the argument deriving from Locke and Hume that things given in experience are logically separable while rejecting the implication of the structural separability argument that there are no genuine relations. And because they reject the principle that only those things are real that are logically separable they also reject the absolutist position of the idealist critics of empiricism such as Bradley. James simply insisted, with Bradley, that we are presented with things as related. He accepts the empiricist’s Principle of Acquaintance but insists, contrary to Hume, that relations are among the entities that are presented to one in ordinary experience. James refers to this position as “radical empiricism.” He makes clear that such a position accepts PA: To be radical, an empiricism must neither admit into its constructions any element that is not directly experienced, nor exclude from them any element that is directly experienced.

His point in emphasizing the latter is that the empiricism of Hume fails to admit relations even though they are indeed presented to one. James continues: For such a philosophy, the relations that connect experiences must themselves be experienced relations, and any kind of relation experienced must be accounted as ‘real’ as anything else in the system.34

On this point he agrees with the idealist critics of Hume. But contrary to these critics he also emphasizes that things as related are given in experience as distinct things. Contrary to the implication of the absolutist account of relations, James insists that this separateness does not disappear upon the recognition that the distinct things are after all related: “...whatever separateness is actually experienced is not overcome, it stays and counts as separateness to the end.”35 James elsewhere uses a perceptive metaphor. The stream of consciousness, he tells us, “Like a bird’s life, it seems to be made an alterna-

144 tion of flights and perchings.” The latter are the “substantive parts” of the world as we experience it, the former are the “transitive parts.”36 The flights are the relations, the properties of things are perchings. In these terms, the upshot of the argument from structural separability is that one must treat flights as perchings. But the idealists such as Bradley, in their account of relations, also treat flights as perchings. Only, instead of their being perchings on particulars in the stream of things as experienced, they are perchings on substantial wholes into which the particulars as experienced are absorbed. James argument is that a consistent empiricism must admit both perchings and flights into its ontology, and that the error into which Hume’s empiricism falls is that it omits flights, or, rather, tries to transform flights into perchings. And in their own way the idealists, while recognizing the need to admit structure, relations, into one’s ontology, do so only if you wish confusedly because they also insist that there are no genuine flights, only perchings. Like James, Russell accepts the reality of genuine relations, and accepts that objects that are related maintain their distinctness. Russell argument is that unless one accepts such an account of relations one comes into conflict with facts about relations that likely should be taken to be obvious but which were long neglected by philosophers and logicians. He develops this argument against both the Humeans and the idealists: in both cases, the account of relations simply will not do. On Bradley’s idealist or absolutist view, relational states of affairs consist of a relational property r being predicated of a complex individual whole (a, b). The relation of a to the whole (a, b) is the same as the relation of b to that whole. That is, the role of a in that whole is symmetrical with the role of b. Thus, Bradley’s schema (+) r(, b) represents indifferently both the fact that (^) a is R to b and its converse, the fact that b is R to a (^c) Where R is a symmetrical relation, one for which we have (x)(y)[Rxy e Ryx] then we have no problem: if (^) obtains so does (^c), and (+) can represent the two indifferently. But the same does not hold for asymmetrical relations.37 Where R is asymmetrical, we have (s) (x)(y)[Rxy e ~Ryx] In this case (^) obtains while (^c) does not. In the case of an asymmetrical

145 relation, there is a difference – an ontological difference – between a relational fact and its converse that is not captured in any account, like Bradley’s, that requires both facts to be represented indifferently by the same notation. But the monistic or absolutist account was introduced to solve the problem of relations. Since it cannot do that, it must be rejected.38 To this objection, Bradley replies,39 that the incompatibility between a relation and its converse, that is, the law (s), if it is to be more than a matter of chance, must be the expression of a real relation that obtains between a relational fact constituted by R and the converse of that fact. It will, therefore, be a necessary fact about R, part of the meaning of R, that its obtaining excludes its converse obtaining. Now, this may well be the case, given the monistic or absolutist account of relations. It does not, however, adequately reply to Russell. For, even if it is somehow a necessary truth that the obtaining of a relation excludes the obtaining of its converse, that is, if it is somehow a necessary truth that a relation is asymmetrical, it still does not follow that one has provided an account of relations that adequately captures the difference between a relation and its converse. The point remains that (+) represents both (^) and (^c) indifferently, this because a and b occur symmetrically in (a, b). There is therefore nothing that accounts for the difference between (^) and (^c). Bradley looks at the contrariety between (^) and (^c) rather than the difference between them that is presupposed by the contrariety.40 The same sort of argument can be made against the account of relations that is implied by the Humean claim that things are structurally separable. On this account, a relational state of affairs (%) Rab must be reducible to a pair of non-relational states of affairs: (%%) r’(a) & r”(b) where r’ and r” are the foundations of the relations. Suppose that R is the relation of being a father of. Then the relative product of R with itself (R|R)xy = (›z)(Rxz & Rzy) is the relation of being a grandfather. Suppose that a is the grandfather of b: (R|R)ab Then we have (›z)(Raz & Rzb) If we now reduce R to its foundations we have (›z)[r’(a) & r”(z) & r’(z) & r”(b)] Notice that this implies

146 r’(a) & r”(b) so that not only is a the grandfather of b but also the father of b: my grandfather turns out to be necessarily also my father. Hardly what mother nature intended! Moreover, whatever is the z is that is said to exist as the son of a and the father of b, it is implied that r’(z) & r”(z) i.e., that this individual is its own father: every son who has a son is also his own son. Again, not what nature intended! That there are relations which are relative products of other relations is a simple fact. But the account of relations implied by the structural separability of things requires that the grandfather of a boy is the father of that boy and that a son who has a son is also his own father. So much the worse for that account of relations. Like the idealist account of relations it simply cannot account for simple facts about relations. As Russell sees it, both the account of relations implied by Hume’s claim that things are structurally separable and the alternative account of relations proposed by the idealists assume that predication always involves only one term. This is the “common opinion ... that all propositions, ultimately, consist of a subject and a predicate.”41 Russell rejects this common assumption. In rejecting this assumption, Russell is making essentially the same point as James when the latter insisted that there are both perches, that is, non-relational predications, and flights, that is, relational predications, and that where a relation relates two distinct things those things retain their distinctness in the relational state of affairs. Russell's account of relations takes the grammatical form of (@) to perspicuously represent its logical form. The objective fact represented by (@) does not dissolve into a pair of facts about individuals – (&&) – as on the account implied by the structural separability argument. But this unity is not a whole of which the relation is predicated, as on Bradley’s account. Rather, the relation is predicated of the terms jointly. It is a and b being related that is the unified whole, rather than a and b being constituted into a whole of which the relation is then predicated. Notice, however, that on the James-Russell account of relations, as on Bradley’s idealist account, when we have a relational state of affairs (@) a is R to b or Rab the two things a and b are located in a genuine unity in the sense that, if

147 one of a or b were not to exist, that unity could not exist. Thus, if b were to cease to exist the thing a would cease to have Rxb predicated of it. And so, if b were to cease to exist, a would change in the sense that what could be predicated of it when b exists cannot be predicated of it when b ceases to exist. Hence, upon the James-Russell account of relations, as upon Bradley’s account, things that are related are structurally inseparable. It follows that if the empiricist, or, in James’ term, the radical empiricist accepts the reality of genuine relations, then one must reject Hume’s claim that things are structurally separable. And if one rejects this point, one must reject the argument based upon it that Hume uses to deny the reality of objective necessary connections. Now, Bradley and the other idealists similarly rejected the structural separability of things and the argument against objective necessary connections that was based upon that notion. They saw correctly, as James and Russell were also to see, that if one allows into one’s ontology genuine relations, then one must accept that things are structurally inseparable. The idealists went on to develop an account of relations which allowed for things to be structurally inseparable. But upon their account, objective necessary connections returned to the ontology, along with a non-sensible form of experience, what Bradley called “thought,” as a way of knowing these connections. James and Russell, in contrast, locate the relations which they admit into their empiricist ontologies in the world of ordinary experience. There is nothing in the account of relations which they proposed that requires them to hold that these entities and the states of affairs into which they enter are outside the world of sensible experience and can be known only by some form of rational intuition or non-sensible form of experience. What it is important to notice is that this acceptance of relations and of the notion that things as related are structurally inseparable does not require us to reintroduce objective necessary connections. The first of Hume’s arguments, already given as we saw by Locke, based on PA, that properties are logically separable, yields the empiricist conclusion that there are no objective necessary connections. This argument is independent of the argument based on the claim that things are structurally separable and one can consistently accept this argument while rejecting the notion that things are structurally separable. The introduction of genuine relations into an empiricist ontology by

148 James and Russell had important consequences. Thus, for example, it became possible to describe how the world as actually experienced contains within itself and points towards world that lies outside and beyond itself. As James has put it, “Mainly ... we live on our speculative investments, or on our prospects.”42 It is hard to see how this is possible either on Hume’s empiricism, given its account of relations, or on Bradley’s idealism, given its account of relations. On the former, there is, in James’ phrase, no “hanging-together.”43 Hume does allow that we have “relative ideas,” and in terms of these he suggests that we can think of things that are not themselves presented but that are related to what is presented. The suggestion is a good one. But given the account of relations implied by his claim that things are structurally separable, the experienceable but unexperienced entities thought of by means of such concepts are in fact wholly unconnected to the things of the world as experienced. There is an ontological gap between the world we experience and the world beyond what we experience. As for the idealist account of relations, here we do have genuine connections. But they do not point to a beyond that is coordinate with what is given in ordinary experience. The relation points not to another particular thing like the thing that we are presented with in ordinary experience; rather, it points to a whole which is not given in ordinary experience. There is a beyond but this beyond is an entity that is outside not only the world as experienced, that is, sense experience, but the world as experienceable. Both these views can be contrasted with what we have upon the James-Russell account of relations. On this view, “... we at every moment can continue to believe in an existing beyond. It is only in special cases that our confident rush forward gets rebuked.” This beyond is one that is of a piece with the world as we experience it. “The beyond,” James argues, “must ... always in our philosophy [radical empiricism] be itself of an experiential nature.”44 To see how this goes, consider a person a, who, let us assume, we are acquainted with in our ordinary experience. a has a father and this father has a father and so on: for every person there is another person who is his or her father. If R is the relations of being a father of then we can form the relative product of R with itself R|R = R2 to give the relation of grandfather, and the relative product of this with R (R|R)|R = R3 to give the relation of great grandfather, and so on to the concept (v) Rn Using this concept we can form with regard to a the idea that there is a per-

149 son who is his or her great, great, ... great grandfather: (vv) (›y)[Py & Rnya] Furthermore, we can infer this is in fact true. Since we know it to be a fact about the relation R that for every person there is another person who is his or her father, i. e., that (vvv) (x)[Px e (›y)[Py & x … y & Ryx]] Given the general fact (vvv) that is confirmed in our experience of things in the ordinary world we can form the idea (v) and with that not only think of things outside the world of experience but can have a reasonable expectation that those things exist. The things with which we are acquainted thus point towards a beyond, towards things which are not experienced but which we can reasonably expect, in the right circumstances, to come to experience. In this way things in this world point to things that are not presented, and the latter are firmly connected with the things that are presented to us in ordinary experience. There is a beyond that is experienceable but not experienced and this beyond is solidly linked with that which is experienced. This linkage is provided by the relations such as R which connect things, and our knowledge, confirmed in the world as experienced, of general facts such as (vvv) about these relations.45 A second important consequence of the introduction of genuine relations into the empiricist ontology has to do with causation. Again the problem for the empiricist ontology lies in the implications of the claim that things are structurally separable. Consider one of the standard examples of causation, “the communication of motion, which I see result at present from the shock of two billiard balls...”.46 We have here a case of cause and effect: the first billiard ball strikes the second, and the motion of the first is communicated to the second. Now, Hume argues that cause and effect are distinct. This is not merely a matter of logical separability but also structural separability. Thus, he tells us that ... as all distinct ideas are separable from each other, and as the ideas of cause and effect are evidently distinct, it will be easy for us to conceive any object to be non-existent this moment, and existent the next, without conjoining to it the distinct idea of a cause or productive principle. The separation therefore of the idea of a cause from that of a beginning of existence, is plainly possible for the imagination; and consequently the actual separation of these objects is so far possible, that it implies no contradiction nor absurdity...47

150 The cause and the effect are related to each other; they are, Hume argues, “contiguous in time and place, and that the object we call cause precedes the other we call effect.”48 Contiguity in time and contiguity in place are relations. But if things are structurally separable, as Hume asserts, then these relations have to be understood as reduced to their foundations; otherwise the one thing cannot remain unchanged if the other were to cease to exist. So there is a problem here. But there is worse. Hume’s account implies that, while the cause and effect are contiguous, the cause immediately precedes the effect: that is the point of speaking of contiguity. But events which are contiguous have no other events between them. The problem is that this will not do for the case of transfer of motion from one billiard ball to another. This latter is a continuous process. A process of this sort involves a relational structure in which the events related are compact, in the sense that between any two events there is always another. There is therefore, contrary to what Hume’s account of cause requires, no immediate predecessor for the event that consists of the second billiard ball acquiring the motion of the first. Furthermore, the very moment when the one billiard ball loses the motion is the moment when the second gains it: the “transfer” is instantaneous. Hume in fact argues that such instantaneous transfer is not possible. For it would imply that the effect is contemporaneous with the cause, and “if one cause were contemporary with its effect, and this effect with its effect, and so on, it is plain there would be no such thing as succession, and all objects must be coexistent.”49 However, to repeat, if we understand the process in the way required by Newtonian physics, then the process is continuous, and the transfer occurs instantaneously. In order to deal with the problem, it is necessary to hold that the series of events in which the one billiard ball strikes the second are related by the relation of successor, where this relation has the properties of compactness and continuity. (The latter implies the former.) Thus, if S is the successor relation, then we will have the law that if one event is a successor of a second then there is a third event which succeeds the first and is succeeded by the second: (w) (x)(y)[(x … y & Sxy) e (›z)(z … x & z … y & Sxz & Szy)] There are further axioms that define compact and continuous relations. (w) and the other properties of compact and continuous relations are analogous to the property (vvv) of the relation R that we noted above. They state a regularity about succession. It is a generality that implies the existence of events in a certain order. It is confirmed in experience, though, of course, since it is a regularity we have not confirmed all its instances in experi-

151 ence. In other words, like (vvv) and R, (w) and the other axioms imply that there are events in the series which are not themselves given in experience. In particular, ever so many of the events implied by compactness are below the threshold at which it is possible for us to distinguish. But that of course simply means that they are part of the beyond that is implied by the relational structures such as (vvv) or (w). It is important to note that to say that a relation presented in sense experience has a property such as (vvv) or (w) is to state a matter of fact. It is not a matter of necessity, and certainly not a matter of logical necessity. Note, moreover, that Descartes’ criticism of Aristotelian theories does not apply. Descartes objected to the Aristotelian account of knowledge that it could not account for our knowledge of the changes in the sample of wax as it melted and transformed itself through an infinity of shapes. The imagination, he argued, could not encompass an infinity. Since there was in fact an infinity to be known there must therefore be some form of knowledge other than that of sense experience. This objection applies to Hume’s account of causation. The transfer of motion from one billiard ball to another is a continuous process, and therefore involves an infinity of events. That is the implication of the compactness axiom (w). It would seem that Descartes’ objection to Aristotelian theories of knowledge apply equally well to Hume’s account of causation. There is, Descartes can argue, a continuous process; this process involves an infinity of events; but the imagination can form ideas of only a finite number of things; our empirical intellect can therefore not grasp the infinity of things; there must therefore be some other way of knowing causation – so at least it could be argued. It is Russell’s account of relations that provides the reply. What we need in order to conceptualize the relevant notion of infinity is not an actual infinite of images of the series of events in the process. One needs a concept that enables one to think of them as it were simultaneously. This we can do by means of a law such as (w). A generalization enables one to think of a population without having to think of each member of the population. All that the empiricist requires is that the non-logical terms that appear in the statement of the generalization are empirical concepts, that is, concepts that refer to what is given in ordinary experience or can be defined on the basis of such concepts. In a generalization like (w) the relational concept “S” refers to something that is given in ordinary experience. That is the force of James’ radical empiricism that insists on the basis of the empiricist’s PA that relations be admitted into one’s ontology. The generalization (w) is therefore an empirical generalization. Because it is a

152 generalization it enables us to think of the entire population of intermediate events, the infinity of events that the compactness implies exist. So, contrary to Descartes, the empiricist can hold that an infinity is thinkable on the empiricist account of things. Once one has a relational of succession that is compact and continuous, the problems for causation created by the billiard ball example are solved. But the simplistic notion that causes are separable but contiguous must go. Not so, however, the notion of law or regularity: this remains, and so does the notion that there is no objective necessity to such regularities, that is, the notion that they are simply matter of fact generalizations. These regularities will be such that there is no contradiction in supposing that they are false. This is the substance of Hume’s claim, based on PA, that cause and effect are logically separable. What has disappeared with the new notion of relations is the notion that things are structurally separable, not the notion that they are logically separable. And so, as Hume claimed, it is logically possible that the second billiard ball not move off when struck by the first. Thus, the substance of Hume’s view that causation is regularity can still be defended against its rationalist and Aristotelian critics.50 One of the main theses of empiricism has been the notion that causation amounts to matter-of-fact regularity. However, so long as one holds on to the notion that things are structurally separable, there can be no genuine relations. But without genuine relations the empiricist account of causation becomes problematic. It simply cannot be made to fit standard examples of causation, e.g., the transfer of motion consequent upon the impact of one billiard ball on another. What James did with his radical empiricism was argue that genuine relations are in fact compatible with PA, and that they ought therefore to be admitted into the empiricist ontology. It was Russell’s contribution to work out in detail the logic of relations, and to show how to fit such properties as compactness and continuity into a world that admitted genuine relations. In this way the contributions of Russell and James amounted to showing how to solve problems implicit in the claim that things are logically separable.51 Essentially, what they argued was that this notion simply had to be abandoned. Once it is, nothing essential to empiricism is lost – causal relations remain matter-of-fact generalizations – , while at the same time the problems with that view disappear. The new view of causation did not go unchallenged, however. Henri Bergson argued that one has in the experience of activity in the case of the will a phenomenon which no empiricist could consistently admit into his or

153 her ontology: we experience it, he argued, but cannot capture it in empiricist conceptual scheme. Locke had already argued that it is from the will that we obtain our idea of a necessary connection, or, as he puts, our idea of an active power. He proposes “to consider here by the way, whether the mind doth not receive its idea of active power clearer from reflection on its own operations, than it doth from any external sensation.”52 He argues that it is not through sensation that we obtain the idea of active power: here events are, as he argues, separable, and all that one can obtain is regularity, the passing on of motion (in the case of billiard balls) rather than the initiation of motion. when by impulse [one billiard ball] sets another ball in motion that lay in its way, it only communicates the motion it had received from another, and loses in itself so much as the other received: Which gives us but a very obscure idea of an active power of moving in body, whilst we observe it only to transfer, but not produce any motion.53

It is from inner awareness that we obtain our idea of an active power. Specifically, we obtain it from our experience of the action of the will in volitions that cause bodily action. The idea of the beginning of motion we have only from reflection on what passes in ourselves, where we find by experience, that barely by willing it, barely by a thought of the mind, we can move the parts of our bodies, which were before at rest.54

Hume was later to argue, however, that the case of the will is no different from the case of the billiard balls: in both, the effect is separable from the cause. Some have asserted that we feel an energy or power in our own mind; and that, having in this manner acquired the idea of power, we transfer that quality to matter, where we are not able immediately to discover it. The motions of our body, and the thoughts and sentiments of our mind (say they) obey the will; nor do we seek any further to acquire a just notion of force or power. But to convince us how fallacious this reasoning is, we need only consider, that the will being here considered as a cause has no more a discoverable connexion with its effects than any material cause has with its proper effect....The effect is there distinguishable and separable from the cause, and could be foreseen without the experience of their constant conjunction. We have command over our mind to a certain degree, but beyond that lose all empire over it: and it is evidently impossible to fix any precise bounds to our authority, where we consult not ex-

154 perience. In short, the actions of the mind are, in this respect, the same with those of matter. We perceive only their constant conjunction; nor can we ever reason beyond it. No internal impression has an apparent energy, more than external objects have. Since, therefore, matter is confessed by philosophers to operate by an unknown force, we should in vain hope to attain an idea of force by consulting our own minds.55

This argument goes through whether the notion of separability is that of logical separability or that of structural separability. If the latter, then of course there is no continuity between the volition and the action. Indeed, taking the volition or activity as an event that takes a finite amount of time, then it is an event that has temporal parts. These too will be structurally separable. So there will not even be continuity within the volitional activity itself. But, in contrast, if the argument is made in terms of logical separability, then there is no reason to deny that there is continuity within the volition itself and between the volition and the bodily action. It was Bergson’s argument that, although we experience, and therefore are aware of, activity, nonetheless an empiricist cannot consistently admit it into his or her ontology. His argument is that the empiricist account of concepts provides for ideas of how a thing is but not how it is becoming, that is, how it is changing or moving. But in activity we are aware of a change, a becoming, a movement. The correct philosophy is one of “dynamism,” where “Dynamism starts from the idea of voluntary activity, given by consciousness...”.56 Contrary to Hume, this dynamic feature of reality involves a continuity that cannot be captured in empiricist concepts. Since there is more to the world than the empiricist account of concepts allows, the empiricist ontology is inadequate. It is inadequate, ironically enough, because it does not allow for something that we experience. Since we do experience activity, and since the empiricist cannot account for that, it follows that there is a kind of experiencing, a kind of intuition, which is beyond the empiricist intellect, its object uncapturable in empiricist concepts. “...[I]n default of knowledge properly so called, reserved to pure intelligence, intuition may enable us to grasp what it is that intelligence fails to give us...”.57 Bergson argues that movement is a passage from something being somehow or somewhere to its being somehow or somewhere else. Thing a is F and it moves from being F to being G. When the thing a is, it is at rest. The movement consists of the passage from being F to being G, from being at rest in one state to being at rest in another state. The movement itself involves a continuity: it is a passing from one state of rest to another. “Every

155 movement, inasmuch as it is a passage from rest to rest, is absolutely indivisible.”58 In this dynamic feature of the universe, there is continuity; the parts are not genuine parts, they are inseparable. He considers moving one’s hand from A to B. “My consciousness gives me the inward feeling of a single fact, for in A was rest, in B there is again rest, and between A and B is placed an indivisible or least an undivided act, the passage from rest to rest, which is movement itself.”59 Our ideas, however, represent the ways in which things are similar; they represent properties that things have in common. But these similarities are themselves, as the empiricist insist, distinct, or separable. There may well be contiguity between ideas considered as occurrences in the mind. As Bergson puts, “...between any two ideas chosen at random there is always a resemblance, and always, even, contiguity...”.60 But contiguity impies separateness, not continuity. The similarities are of the way things are, and when they are they are at rest. The ideas do not capture the continuity of movement: “...we must not confound the data of the senses, which perceive the movement, with the artifice of the mind, which recomposes it.”61 Our ideas, the intellect of the empiricists, represent motion as a series of states at rest, a series of stills, as it were, which, however rapid, and however close the contiguity, are still a series of separable images, not the genuine continuity given to us in sense experience. The senses, left to themselves, present to us the real movement, between two real halts, as a solid and undivided whole. The division is the work of our imagination, of which indeed the office is to fix the moving images of our ordinary experience, like the instantaneous flash which illuminates a stormy landscape by night.62

The real movement is not the passage of contiguous parts as represented by empiricist concepts. It is rather a continuity, in which the end of one part is the beginning of the next.63 Thus, Bergson contrasts “[t]he simultaneties of physical phenomena, absolutely distinct in the sense that the one has ceased to be when the other takes place, cut up into portion, which are also distinct and external to one another...,” with “an inner life in which succession implies interpenetration...”.64 Our inner life involves “succession without externality,” a succession where the parts are “interpenetrating.”65 This is one part of Bergson’s argument. There is a second part to which we shall return shortly. But first we have to look at the reply by Russell to this first part of Bergson’s argument.66 Bergson’s argument is very much of a piece with Descartes’ criti-

156 cism of the Aristotelian doctrine of abstraction, that the infinity of motion cannot be represented by the finite concepts derived from sense. Bergson makes essentially the same point. Descartes argued that we must rely upon a rational intuition to give us insight into the infinity and continuity of motion. Bergson makes essentially the same point: in order to know activity we need a form of knowledge beyond that which is available to empiricists. But where Descartes relies on a “rational intuition,” Bergson relies upon a sensory intuition. Bergson’s view is also akin to that of Bradley, but where Bradley has a special form of experience which he calls “thought”, akin to the “rational intuition” of Descartes, Bergson again has sensory intuition. This, as James once stated, places Bergson closer to the empiricist tradition than are Descartes and Bradley; “[a]s one who calls himself a radical empiricist,” he says, “I can find no possible cause for not inclining to Bergson’s side.”67 Bergson finds volitional activity in our ordinary experience and correctly makes the same sort of appeal as the empiricist – the radical empiricist – that this must therefore be included in one’s ontology. The issue is whether what is thus admitted is of a sort that eludes ideas as traditionally understood by the empiricist, whether the empiricist has an account of the concepts of things that allows this entity to be thought. Bergson argues that the volitional activity cannot be captured in empiricist concepts. Russell disagrees. Bergson’s argument has two aspects. On the one side, there is the argument that activity is inconsistent with a picture of the universe in which a thing only is and is never besides a thing that becomes. Our concepts are always concepts of the way that a thing is. We therefore need to suppose that there is another way of knowing, a form of intuition, through which we become aware of the becoming of things. Only through this other way of knowing do we become aware of the “absolute” movement of the thing where one, according to Bergson, is “attributing to the moving object an interior and, so to speak, states of mind...”.68 On the other side, there is the argument that our concepts are all of distinguishable aspects of things, which are therefore separable. We can therefore never think the continuity of becoming of which we become aware in our intuition of activity. Moreover, the imagination through which we form our concepts of the qualities of things is finite, and can therefore never grasp the infinity that exists in the continuity of inseparable parts of activity. The activity of things is a simple indivisible thing in itself but at the same time insofar as it is changing it has an indefinite multiplicity of states: it is truly something

157 infinite, for “that which lends itself at the same time both to an indivisible apprehension and to an inexhaustible enumeration” and is therefore “by the very definition of the word, an infinite.”69 Russell replies to the first aspect by pointing out that it simply assumes that, when a thing is changing, there must be a state of change, that is, that “[t]he thing must, at each instant, be intrinsically different from what it would be if it were not changing.”70 The reply to this is simply that there is no need to suppose that there is a state of change; a change consists of nothing more than a transition from what a thing is at one time to what a thing is at another time. To this Bergson replies that it “implies the absurd proposition that movement is made of immobilities.”71 This, however, simply begs the question: it is simply to assert that motion is not a process in which a thing is in different ways successively. Of course, if the motion is continuous then to speak of “successive” states is perhaps misleading. ‘Successive’ suggests that the state that succeeds is contiguous with the one that precedes. But where the motion is continuous, the successive states are not contiguous. For, as we have noted, any continuous motion is compact, where any two (separate) successive states are such there is a third distinct state between them. As for the second aspect of Bergson’s argument, the reply consists in pointing out that where one has a continuous series one must have a relation, and that there is nothing inconsistent with the empiricist account of concepts and of ontology with the admission of relations: that is the point of “radical empiricism.” As Russell says, “...a motion is made out of what is moving, but not out of motions. It expresses the fact that a thing may be in different places at different times, and that the places may still be different however near together the times may be.”72 William James accepted Russell’s argument. He granted the point that activity as we experience it is a continuity without distinguishable parts. In this respect he rejected the notion that activity consists in separable parts. In other words, James agreed with Bergson that our inner life involves “succession without externality,” a succession where the parts are “interpenetrating.”73 But he disagrees with Bergson with respect to the claim that the continuity of such inner activity cannot be adequately represented by concepts that conform to empiricist principles. It can be so represented once one admits relations, as the radical empiricist will do, rejecting the notion that things are structurally separable. The continuity we experience in our inner activity can be conceptually decomposed into the infinity of parts that is required by the notion of continuity. As he put it, “[t]he in-

158 finite character we find in it is woven into it by our later conception indefinitely repeating the act of subdividing any given amount supposed.”74 James thus rejects the first part of Bergson’s argument.75 At the same time, however, he did accept the second part of Bergson’s argument. It is to this second part that we now turn.76 Bergson argues that causation on the empiricist account of causation, causation is regularity and that involves as a basic principle the rule that “...the same causes produce the same effects.”77 There are, however, exceptions to this rule: “the principle of causality,” he argues, “admits of an incomprehensible exception.”78 These exceptions are to be found in our inner mental activity. This activity therefore shows the inadequacy of the empiricist account of causation. Bergson argues that “[t]o say that the same inner causes will reproduce the same effects is to assume that the same cause can appear a second time on the stage of consciousness.”79 However, “...the same feeling, by the mere fact of being repeated, is a new feeling.”80 The crucial fact is memory: when on the next occasion the feeling is called forth, the memory of the earlier occurrence affects the feeling and gives it a new shape. “Our past, ... as a whole, is made manifest to us in its impulse...”.81 For this reason, the same cause in our inner life never produces the same effect. “From this survival of the past it follows that consciousness cannot go through the same state twice. The circumstance may still be the same, but they will act no longer on the same person, since they find him at a new moment of his history.”82 There is therefore something to causality which is more than regular connection. To be sure, with regular connection the idea of the effect is implied by the idea of the cause. But regular connection is not enough. “It seems ... that, if the idea of the second phenomenon is already implied in that of the first, the second phenomenon itself must exist objectively, in some way or other, within the first phenomenon.”83 We find this connection in way in which the end of a mental activity is prefigured in its beginning, with the former flowing continuously out of the latter. “We go, in fact, through successive states of consciousness, and although the later was not contained in the earlier, we had before us at the time a more or less confused idea of it.”84 Intuition thus gives us a notion of causation which is more than regularity, where there is a real objective connection between cause and effect. The source of this concept of necessary connection is not the rational intuition of Descartes. It is our sensible intuition. The Cartesian notion implies that the phenomena can be put into a mathematical formula. But this is simply a rationalistic version of the empiricist principle of

159 “same cause / same effect,” and fails to take account of the fact of our having a history: “That under the influence of the same external conditions I do not behave to-day as I behaved yesterday is not at all surprising, because I change, because I endure.”85 The continuity is not that of the mathematical formula but rather the special sort of continuity that we discover in the inner awareness of our own mental activities, in which the future grows out of the past that prefigures it. William James accepts this argument. We discover novelty as part of our mental activities: “... the perpetual flux is the authentic stuff of each of our biographies, and yields a perfect effervescence of novelty all the time.”86 It is here that we find the truth about causality. “...[R]eal effectual causation as an ultimate nature, ... is just what we feel it to be, just that kind of conjunction which our own activity-series reveal.”87 The necessary connection that is there is our activity, the connection to which Locke directed our attention; it is, however, not the rationally pellucid connection of the rationalists. “Even so our will-acts may reveal the nature of causation, but just where the facts of causation are located may be further problem.”88 And so “...the [empiricist] attempt to treat ‘cause,’ for conceptual purposes, as a separable link, has failed historically...”.89 What is crucial is the fact of novelty. Whether we accept Hume’s account of causation or that of Descartes, “... no real growth and no real novelty could effect an entrance into life.” This negation of real novelty seems to be upshot of the conceptualist philosophy of causation.90

It is the fact of novelty that provides the basis for the argument that regularity theories will not do. ... the concrete perceptual flux, taken just as it comes, offers in our own activity-situations perfectly comprehensible instances of causal agency. The transitive causation in them does not, it is true, stick out as a separate piece of fact for conception to fix upon. Rather does a whole subsequent field grows continuously out of a while antecedent field because it seems to yield new being of the nature called for, while the feeling of causality-at-work flavors the entire concrete sequence as salt flavors the water in which it is dissolved.91

James has already granted that the feature of relatedness can be captured by an empiricist ontology; that, after all, was the point of “radical empiricism.” He has also already granted that the empiricist can allow that there

160 is continuity in the process. For James, as for the second part of Bergson’s argument, what is crucial is the fact of novelty; it is not so much continuity as growth that is central. However, is it really the case that novelty and growth cannot be captured in the empiricist ontology of causation? It is certainly true that if one thinks in simple-minded “stimulusresponse” terms, then Bergson is correct: that is not the way that people work. We cannot simply say that (") R = f(S) As Bergson points out, the same external stimulus will evoke a different response the second time it occurs, that is, is presented or occurs within our experience. Our second reading of a poem will yield a different response than our first. What we need, rather, is something like ($) R = f(S, H) where “H” represents the history of the individual. Once we see this, however, it is evident that one can perfectly well fit the fact that history is relevant to determining the response to a stimulus into the regularity view of causation. It is just that the regularities are of the form ($) rather than of the form ("). Regularities of the sort ($) can be fit into mathematical form. The relevant form is that of integral-differential equations first explored by Volterra.92 Psychologists involved in the study of behaviour have long been familiar with the fact that there is an “historical” dimension to human being. In order to predict what a person is going to do – or a white rat, for that matter – one must know the schedule of reinforcement.93 In fact, the laws of association that the empiricists defended are historical in precisely this sense: the strength of the association depends upon the past history of observation. There is, however, the idea implicit in Bergson and James that somehow the novel occurrences cannot be predicted. The mere fact of historical background will not, contrary to Bergson’s views on memory, establish this: as ($) makes clear, there can be regularities of a perfectly good empiricist sort that relate past history to present response. Nonetheless, there are other possibilities with regard to novelty. Processes can produce new configurations of things. But this is hardly an interesting sense of novelty, since there is no problem with prediction in such a case. A. O. Lovejoy has suggested two cases of a more interesting sort. There is novelty in the sense of “[n]ew qualities ... attachable to entities already present, though without those accidents in [the an-

161 tecedent phenomena].” And there is novelty in the sense of “[p]articular entities not possessing all the essential attributes characteristic of those found in [the antecedent phenomena], and having distinctive types of attributes (not merely configurational) of their own.”94 But contrary to what Bergson and James seem to suppose, there is no reason why there should not be regularities which relate the antecedent phenomena to the novel qualities or entities.95 In the first place, something could be novel in either of these senses without being novel in a temporal sense. Thus, the properties of water may be emergent relative to the properties of hydrogen and oxygen, without however being temporally subsequent to hydrogen and oxygen. In the second place, even if the property or entity is novel in the temporal sense, so that we have not observed things of that kind, it does not follow that there is no regularity connecting those properties to the antecedent phenomena. All that follows is that we cannot know those regularities prior to observing things of the novel kind. But an incapacity to know is compatible with causation. Of course, in the absence of knowledge we cannot predict, but again the incapacity to make a justified prediction is consistent with causation in the sense of regularity. So the notion on which Bergson and James base their argument, that novelty is inconsistent with regularity, is simply mistaken.96 The other point that perhaps ought to be made is that when we experience mental activity, we are not at the same time making a causal judgement about it. When I am writing a philosophy essay, the various stages emerge one after the other, in a process that is at once piecemeal and yet united not only by a continuous stream of thought but also by the intention of writing an essay. The intention does not of course include from the outset the details of the words that emerge, but it does serve to organize the whole process. The process is causally united, and it is experienced as a united process. But in so experiencing it, I am not at the same time judging that it is causally united. And certainly, while I am intending the outcome, and while that intending causally brings about the essay as an effect, I am certainly not predicting that outcome. It is only in subsequent reflection that I become aware of the causal structure of the process; only later after its completion can I reflect upon it and recognize it as an instance of a regularity. In that sense, Bergson and James are correct: the experience of causation in activity is prior to any knowledge of such activity as an instance of a regularity. It follows that the second part of Bergson’s argument, and James’ version of it too, fails in its aim to introduce into philosophy a category of

162 causation that is inconsistent with Hume’s empiricist ontology: the regularity view can stand. Nor is there any need to introduce some non-ordinary form of knowing through which we are supposed to be able to know or experience the entities that elude the empiricists’ categories or violate the Principle of Acquaintance.

163

Endnotes to Study Three

1.William James, “The Experience of Activity,” in his Essays in Radical Empiricism, intro. Ellen Kappy Suukiel (Lincoln: University of Nebraska Press, 1996), pp. 155189, at p. 160. 2.David Hume, Treatise of Human Nature, ed. L. A. Sleby-Bigge, rev. P. H. Nidditch (Oxford: Oxford University Press, 1976), p. 4. 3.David Hume, Treatise, p. 3. 4.Cf. Hume, Treatise, p. 68. Compare D. Flage, “Hume’s Relative Ideas,” Hume Studies, 7 (1981), pp. 5373; and “Locke’s Relative Ideas,” Theoria, 47 (1981), pp. 142-159. 5.John Locke, An Essay concerning Human Understanding, ed. P. H. Nidditch (Oxford: Oxford University Press, 1979). 6.On these issues, see F. Wilson, Logic and the Methodology of Science in Early Modern Thought: Seven Studies (Toronto: University of Toronto Press, 1999), Study Two. 7.Locke, Essay, IV, iii, 28, pp. 558-9; see also IV, vi, 10, pp. 584-5. 8.Locke, Essay, IV, viii, 2. 8.Cf. F. Wilson, “Acquaintance, Ontology and Knowledge,” in this volume, above. Also “Moore’s Refutation of Idealism,” also in this volume, below; and “Perceptual Ideality and the Ground of Inference: Comments on Ferreira’s Defence,” Bradley Studies, 1 (1995), pp. 139-152. 9.Hume, Treatise, p. 160. 10.Hume, Treatise, p. 80. 11.Hume, Treatise, p. 160; Hume’s italics. 12.Hume, Treatise, p. 70. 13.David Hume, Enquiry concerning Human Understanding, ed. L. A. Selby-Bigge, 2nd ed. (Oxford: Oxford University Press, 1992), p. 61. 14.Hume, Treatise, p. 636.

164

15.Cf. Wilson, The Logic and the Methodology of Science in Early Modern Thought: Seven Studies, Study One. 16.R. Descartes, Meditations on First Philosophy, in Philosophical Essays, trans L. J. Lafleur (New York: Macmillan, 1964), p. 87. 17.Descartes, Meditations, p. 88. 18.Descartes, Meditations, p. 88. 19.Descartes, Meditations, p. 88. 20.Hume, Treatise, p. 160; Hume’s italics. 21.F. H. Bradley, “Association and Thought,”Mind, o.s. 12 (1887), pp. 354-381, p. 357. 22.Bradley, “Association and Thought,” p. 357. 23.F. H. Bradley, Appearance and Reality, Second Edition (Oxford: Oxford University Press, 1897), p. 125. 24.Bradley, Appearance and Reality, p. 125. 25.Bradley, Appearance and Reality, p. 18. 26.Cf. F. H. Bradley, “Relations,” in his Collected Essays, 2 vols. (London: Oxford University Press, 1935), vol. I, pp. 635-6. 27.Bradley, “Relations,” p. 636. 28.Bradley, Appearance and Reality, p. 19. 29.Bradley, Appearance and Reality, p. 636. 30.William James, “The Thing and Its Relations,” in his Essays in Radical Empiricism, p. 103, referring to Hume, Treatise, p. 636. 31.F. H. Bradley, “Association and Thought,” p. 357. 32.F. H. Bradley, Essays on Truth and Reality (London: Oxford University Press), p. 231. 33.Bradley, Appearance and Reality, p. 426. 34.William James, “A World of Pure Experience,” in his Essays in Radical Empiricism, p. 42.

165

35.James, “A World of Pure Experience,” p. 87. 36.William James, Principles of Psychology, 2 vols. (New York: Dover, 1950), vol. I, p. 343. 37.B. Russell, Principles of Mathematics, second edition (London: Allen and Unwin, 1937), p. 221. 38.For a more recent discussion of the problem of order, see E. B. Allaire, “Relations: Recreational Remarks,” Philosophical Studies, 34 (1978), pp. 81-89. 39.Bradley, “Relations,” p. 672. 40.In fact the difference is there also in the case of symmetrical relations, and is also equally overlooked in such cases. But in those cases it is open to the monists to dismiss the difference as ontologically irrelevant. In the case of asymmetrical relations, such dismissal is not possible. 41.Bertrand Russell, Principles of Mathematics, p. 221. 42.James, “A World of Pure Experience,” p. 88. 43.James, “The Thing and Its Relations,” p. 107. 44.James, “A World of Pure Experience,” p. 88. 45.On these points, compare G. Bergmann, “Remarks on Realism,” in his Metaphysics of Logical Positivism (New York: Longmans, Green, 1954), pp. 153-175. It is clear that, contrary to what P. Ferreira suggests, empiricists as well as Bradley can accept that there is a “beyond.” Cf. P. Ferreira, Bradley and the Structure of Knowledge (Albany, NY: State University of New York Press, 1999), p. 159ff. 46.Hume, Treatise, p. 164. 47.Hume, Treatise, p. 82. 48.Hume, Treatise, p. 155. 49.Hume, Treatise, p. 75. 50.Cf. B. Russell, “On the Notion of Cause, with Applications to the Free-Will Problem,” in H. Feigl and M. Brodbeck, eds., Readings in the Philosophy of Science (New York: Appleton-Century-Crofts, 1953), pp. 387-407. 51.Russell first argued these points in his Principles of Mathematics, in the first edition of 1903; see (second edition) p. 221.

166

52.Locke, Essay, Bk. II, Ch. 21, sec. 4. 53.Ibid. 54.Ibid. 55.Hume, Treatise, p. 162. 56.Henri Bergson, Time and Free Will , trans. F. L. Pogson (Lodon: Allen and Unwin, 1916), p. 141. 57.Henri Bergson, Creative Evolution, trans. Arthur Mitchell, forward by Edwin Edman (New York” Modern Library, Ransom House, 1944), p. 195. 58.Henri Bergson, Matter and Memory, trans. Nancy Margaret Paul and W. Scott Palmer (London: Allen and Unwin, 1911), p. 246. 59.Bergson, Matter and Memory, p. 246. 60.Bergson, Matter and Memory, p. 213. 61.Bergson, Matter and Memory, p. 247. 62.Bergson, Matter and Memory, pp. 247-248. 63.Compare Aristotle’s definition of continuity: “That which, being successive, touches, is contiguous. Since all change is between opposites, and these are either contraries or contradictories, and there is no middle term for contradictories, clearly that which is between is between contraries. The continuous is a species of the contiguous; two things are called continuous when the limits of each, with which they touch and are kept together, become one and the same, so that plainly the continuous is found in the things out of which a unity naturally arises in virtue of their contact.” Metaphysics 1069a5 ff, in J. Barnes, ed., The Complete Works of Aristotle (Princeton: Princeton University Press). 64.Bergson, Time and Free Will, p. 228. 65.Bergson, Time and Free Will, p. 109. 66. Bertrand Russell, “The Philosophy of Bergson.” 67.William James, “Bradley or Bergson?” Journal of Philosophy, 7 (1910), pp. 29-22, at p. 33. 68.Henri Bergson, An Introduction to Metaphysics, trans. T. E. Hulme (London: Macmillan 1913), p. 2.

167

69.Bergson, An Introduction to Metaphysics, p. 6. 70.Russell, “The Philosophy of Bergson,” p. 332. 71.Bergson, Creative Evolution, p. 325. 72.Russell, “The Philosophy of Bergson,” p. 333. 73.Bergson, Time and Free Will, p. 109. 74.William James, Some Problems of Philosophy, Intro. by Ellen Kappy Suckiel (Lincoln: University of Nebraska Press, 1996), p. 184. 75.For a discussion of some of these points, see Ivar Segelberg, “Zeno’s Paradoxes,” in Three Essays in Phenomenology and Ontology, trans. H. Hochberg and S. R. Hochberg (Stockholm: Thales, 1999). 76.For a discussion of a number of these points, see P. Ferreira, Bradley and the Structure of Knowledge (Albany: State University of New York Press, 1999), Ch. 10. 77.Bergson, Time and Free Will, p. 199. 78.Bergson, Time and Free Will, p. 199. 79.Bergson, Time and Free Will, p. 199. 80.Bergson, Time and Free Will, p. 200. 81.Bergson, Creative Evolution, p. 8. 82.Bergson, Creative Evolution, p. 8. 83.Bergson, Time and Free Will, p. 203. 84.Bergson, Time and Free Will, p. 211. 85.Bergson, Time and Free Will, p. 209. 86.James, Some Problems of Philosophy, p. 151. 87.James, “The Experience of Activity,” p. 185. 88.James, Some Problems of Philosophy, p. 216. 89.James, Some Problems of Philosophy, p. 217. 90.James, Some Problems of Philosophy, pp. 204-205.

168

91.James, Some Problems of Philosophy, p. 218. 92.V. Volterra, Theory of Functionals and of Integral and Integro-differential Equations, ed. Luigi Fantappiè, trans. M. Long (London: Blackie, 1930), p. 147ff. 93.Cf. G. Bergmann and K. W. Spence, “The Logic of Psychophysical Measurement,” in M. H.. Marx, ed., Psychological Theory: Contemporary Readings (New York: Macmillan, 1951), pp. 256-276. Bergmann and Spence point out (p. 272) that the best version of “R = f(S)” would be R = F(S, T, D, I) where “S” is the stimulus, “D” is the motivation, “T” is the training or previous experience, and “I” is the individual differences. 94.A. O. Lovejoy, “The Meanings of ‘Emergence’ and Its Modes,” Proceedings of the Sixth International Congress of Philosophy, ed. E. S. Brightman (New York: Longmans Green, 1927), pp. 26-27. 95.For a good discussion of the problem of novelty, see G. Bergmann, Philosophy of Science (Madison, WI: University of Wisconsin Press, 1957), Ch. III. 96.Popper advanced a variation on Bergson’s argument in his Poverty of Historicism (Boston: Beacon Press, 1957), Preface (though he does acknowledge his debt to Bergson). He restricts the novelty to new knowledge, arguing, correctly, that one cannot predict new knowledge, and inferring, incorrectly, that therefore there can be no laws governing the development of social processes. All that follows here, as in the case of Bergson and James also, is that we cannot know the laws. But, to repeat, this does not mean that there are no laws. For an insightful discussion of Popper’s views on this point, see L. Addis, The Logic of Society (Minneapolis: University of Minnesota Press, 1975), pp. 106-111.

Four On the Hausmans’ “New Approach to Berkeley’s Ideal Reality”* No doubt most will now (unlike the olden days, when idealism flourished) – most will now agree with the Hausmans1 that Berkeley must be mistaken because idealism is false. The problem then becomes that of finding out wherein lies the error: where does the argument of the mitred empiricist go wrong? There is, alas, no agreement upon precisely where that error is. The inherence account of his idealism – the theory that his idealistic claim that ideas must be in minds derives from his construing ideas as properties and reasoning in conformity with the traditional dictum that properties cannot exist apart from substances − is one such account − making the error that leads to idealism a matter of confusing the claim that properties cannot exist apart from the entities of which they are predicated with the claim that properties cannot exist apart from substances, noting that for Berkeley ideas (sensible qualities) are predicated of the wholes of which they are parts and not the minds to which they are present. This account of the error Berkeley’s idealism was first propounded by Allaire, and was subsequently defended by Cummins and Watson.2 George Pappas offers another account.3 And now the Hausmans offer a third. But if we cannot agree on precisely where Berkeley goes wrong, then that calls into question our right to claim that he does indeed go wrong. Perhaps, after all, his arguments are correct, and the reason we cannot locate the error is that there is no error. We tend to forget that not so long ago the assumption was that Berkeley was correct, that idealism of some form was true, and that materialism and empiricism were vulgar views easily disposed of in the first few lectures of a philosophy course. If this earlier evaluation is correct, then it is no wonder that we can’t find where Berkeley goes wrong. Now, I do happen to agree that Berkeley goes wrong. But I also think * Originally appeared in R. Muehlmann, ed., Berekeley’s Metaphysics: Structural, Interpretative and Critical Essays (University Park, PA: Pennsylvania State University Press, 1995), pp. 67-87.

170 that, to get a focus on his argument, it may well help to turn to the earlier view. Looking at Berkeley from that perspective may throw into relief various features that we neglect if we continue to look at him as an empiricist who somehow goes wrong, perhaps by not going far enough, as far at least as Hume went. We all know the standard history: Philosophers before Locke had two kinds of substance. The empiricists abolished substance. Locke started this empiricist critique of substance. Berkeley carried it a step further by getting rid of material substance, though he retained mental or spiritual substance. Hume completed the inevitable process by getting rid of mental substance too. The result was the empiricist world consisting of a bunch of sensory atoms, which, when grouped together in one way constitute bodies and when grouped in another way constitute minds. In my view this traditional history is as mistaken as it is common. In particular, it is mistaken with regard to Berkeley. The discussion since the re-construction of his Commonplace Book sees Berkeley as struggling ambiguously to make the move from Locke to Hume. He did succeed in getting rid of material substance. Mental substance did remain, but though he relies upon this to do a lot of work for him at various points he nonetheless felt the empiricist drive to eliminate it. For, after all, at one point in the Commonplace Book Berkeley did state the Humean view of mind as a bundle of perceptions. It is just that the traditional view of substance had such a hold upon him that in the end he could not give it up. In particular, he could not give up the view that properties must be in substances. Or so at least the inherence account. As a consequence he landed in the idealism that we all agree is false. The problem is that this account makes Berkeleyan substances ontologically unmotivated: they are there, and lead to idealism, simply because Berkeley cannot find it within himself to abandon the substance tradition. It would be better if we could find a motive for substances. I shall propose that there is indeed such a motive. Moreover, I shall suggest, once that motive is clear we also see why Berkeley is committed to the position that properties must be in mental substances, that is, why Berkeley is committed to the central thesis upon which the inherence account of his idealism depends. Thus, the account that I shall suggest does justice to Berkeley’s arguments and commitments and makes clear why the inherence account of Berkeley's idealism really does make sense of what he has to say. These points will begin to become clear once we reflect upon the inadequacies of the empiricist account of Berkeley as adumbrated by the

171 Hausmans. 1. The Empiricist Berkeley The Hausmans begin, not with the philosophers that lay behind Berkeley, and whom he had read, but whigishly at the other end of the history of philosophy, with D. C. Dennett’s account of early modern philosophy and how it came to its end.4 On Dennett’s view, the early modern philosophers were trying to do what he, Dennett, is trying to do, and they should be read, he thinks, in that light; it is just that he is doing it a bit better. What he, and they, are/were trying to do is explain what it is to understand. People, of course, are understanders; so might be other entities. An entity is an understander just in case that it grasps the meanings of things. As Dennett puts it, ...nothing is intrinsically a representation of anything; something is a representation only for or to someone; any representation or system of representations thus requires at least one user or interpreter of the representation who is external to it. Any such interpreter must have a variety of psychological or intentional traits...: it must be capable of a variety of comprehensions, and must have beliefs and goals (so it can use the representation to inform itself and thus assist it in achieving its goals)...5

The things that are meaningful include behaviour, specifically rational behaviour, and still more specifically verbal behaviour, but also ordinary objects. Thus, for example, a tree is, or becomes, meaningful to one when one perceives it. The meaning to one of an object or behaviour is the way in which one responds to the object or behaviour. The sounds I hear, or the marks I see, are meaningful when I respond to them as carriers of linguistic meaning. The object before one is meaningful to one as a tree when one responds to it in ways that are appropriate to trees, that is, discriminates it from the background, identifies it as a tree, and further treats it as a tree. The meaning may vary significantly given only slight variations in the behaviour. Consider the different responses to ‘Your son was killed’ and ‘Our son was killed.’6 Further, an object or behaviour may have effects which are not part of its meaning. To grasp the meaning of an object or behaviour is to select those effects that constitute the meaning and to respond appropriately. But no piece of behaviour is intrinsically meaningful; its meaning is a matter of certain, but not all, its effects. How do we select those effects that constitute mean-

172 ing? Here is a first stab: Meaning is that which understanders select as meaningful. But this is clearly circular, unless, that is, we can explain what it is to select. This is what the philosophers of the early modern period were trying to do, according to Dennett. These philosophers, quite correctly on his view, took ideas to be representations that are internal in “the sense of being immediately present to consciousness.” As Dennett puts it, “One understands an object when one has ideas that represent that object.”7 One understands a piece of behaviour, say an uttered sentence, if one has the ideas that give meaning to that sentence. These ideas represent what the sentence means; and more specifically, the sentence means what is represented by those ideas which, by established linguistic convention, the sentence is associated. The object or behaviour evokes the response. The response is the presence in a conscious state of the idea. The object or behaviour evokes the response because that idea represents or refers to that object or behaviour. We have here an empirical pattern or regularity linking the object or behaviour on the one side and the idea on the other. This regularity is an instance of the sort of pattern-governed responses to objects and behaviour that one finds in systems, such as persons, that understand. These patterns are not simply patterns, however; they are also rule-governed.8 The latter is a point that we shall take up directly. For the present, what is important is the pattern. The patterns we have just noted can be called “system-entry transitions.” There are also “intra-system” transitions, inferences, deductions, etc., and “system-exit transitions,” actions, rational actions, speech acts, etc. The system-entry transitions are part of the broader pattern of system transitions that constitute the fact that the ideas refer to or designate certain things or behaviour; they are part of the broader pattern of system transitions that constitute the semantics of ideas. There is no problem here if ideas wear their semantics on their faces, that is, if that semantics or meaning is intrinsic. In that case, it would be transparent to consciousness what the reference is of the ideas that are present to it when it perceives objects or understands behaviour. Ideas, however, according to the early modern philosophy, and quite correctly, are sensations, and, like ordinary objects and items of behaviour taken as such or sentences taken as a set of marks or noises, they have no intrinsic meaning. Indeed, nothing has intrinsic meaning. Here we may recall Dennett’s point that “...nothing is intrinsically a representation of anything; something is a representation only for or to someone...”9 So, if we need an understander to grasp the meanings of things, and this meaning is not intrin-

173 sic, that is, these meanings are carried by ideas whose meanings are, like the meanings of ordinary things, external and not intrinsic, then, in order to have the meanings of those ideas understood, then we need another understander within the person who will select those effects of the ideas that constitute their meaning. The consciousness to which the ideas are present must therefore not be mere consciousness but must be, beyond that, an understander. The consciousness to which ideas are present thus becomes the inner homunculus – as Dennett makes the inference, “...any representation of system of representations...requires at least one user or interpreter of the representation who is external to it....Such an interpreter is...a sort of homunculus...”10 – and this homunculus is that substantial self that we find in Descartes and Berkeley. Every real person thus comes to have, according to Dennett’s account of early modern philosophy, a wee inner person. Beyond the real person who is conscious of inner ideas or representations, there must be a further inner person to whom ideas are present as external objects are present to the real person, and whose role is to understand the ideas, grasp their meanings, as the real person understands the meanings of perceived objects and behaviour. Dennett suggests, naturally enough, that this does not solve the problem: in fact, a scientific psychology that aims to give knowledge of meaningful behaviour, is impossible: “...psychology without homunculi is impossible. But psychology with homunculi is doomed to circularity or infinite regress, so psychology is impossible.”11 The case Dennett makes is clear enough. In the first place, it is either circular or leads to a vicious infinite regress. If one attempts to explain the understanding that a person exemplifies by the understanding exemplified by the homunculus, then it is circular, explaining understanding in terms of understanding. And, if one insists that understanding at each level is so to be explained, then the little homunculus will require a still smaller homunculus inside it, which will require a further, even smaller, homunculus, and so on, to infinity, and viciously so since one never reaches a point where understanding has been explained. Moreover, in the second place, from the viewpoint of empiricism, the wee inner self remains forever elusive: we seem never to be acquainted with it in our ordinary experience. As Dennett puts it, “One is conscious only of the products of the producer, which one then consciously tests and chooses.”12 Upon Dennett’s view the early modern philosophers attempted to solve the problem of understanding by means of a metaphysics of a substantial mind, a wee inner homunculus. In the end the problems became

174 evident, and philosophers had to reject the homunculus. It was Hume who saw the problems: “Hume’s internal representations were impressions and ideas, and he wisely shunned the notion of an inner self that would intelligently manipulate them.”13 So Hume argued consistently on empiricist grounds that we should eliminate the substantial mind.14 The result was an inner self that was no more than a bundle of impressions and ideas. Unfortunately this new inner self lacked the resources to account for understanding: impressions and ideas are without any intrinsic meaning, and there is nothing else in the Humean apparatus to explain understanding. On the one hand, Hume is wise, in Dennett’s view: Hume wisely shunned the notion of an inner self that would intelligently manipulate the ideas and impressions, but this left him with the necessity of getting the ideas to ‘think for themselves’.

But also not so wise: His associationistic couplings of ideas and impressions, his pseudo-chemical bonding of each idea to its predecessor and successor, is a notorious nonsolution to the problem.15

Abolishing the self that could intelligently manipulate ideas …left [Hume] with the necessity of the ideas and impression to ‘think for themselves’. The result was his theory of the self as a ‘bundle’ of (nothing but) impressions and ideas. He attempted to set these impression and ideas into dynamic interaction by positing various associationist links, so that each succeeding idea in the stream of consciousness dragged its successor onto the stage according to one or another principle, all without the benefit of intelligent supervision. It didn’t work, of course. It couldn’t conceivably work, and Hume’s failure is plausibly viewed as the harbinger of doom for any remotely analogous enterprise.16

So, although Hume sees the problems of the metaphysics adopted by the early modern philosophers, he has nothing adequate with which to replace that metaphysics, and the problem of understanding remains unresolved. That is, unresolved until the twentieth century and Dennett comes along. The Hausmans are in substantial agreement with Dennett’s version of the history of empiricism in the early modern period. They argue that Dennett’s model can provide us with insight into Berkeley’s idealism. They take for granted that Dennett’s picture is a fair one. The ideas or inner representations that explain perception and other forms of understanding

175 have, in analogy to language, (relevant) sorts of meaning: there is their syntax and their semantics. The syntax has to do with their complexity; the syntactical structure of a complex idea, say a proposition, is given by the relations which structure the simple components into that complex. The semantics has to do with the reference of ideas to the objects external to themselves which they represent. For Descartes, Malebranche, and Locke, each in his own way, this external reference is to something other than an idea. This creates a problem for the semantics of ideas: how does the homunculus interpret ideas, that is, understand their meanings, if that which they mean is neither an idea nor like an idea? One perceives an object. This perceiving consists in selecting the object by means of an idea. This idea means the object. But the idea is simply a sensation, without intrinsic meaning. Thus, first, the homunculus must grasp this meaning. Second, given the principles of this metaphysics of understanding, only ideas are present to the homunculus. And third, the meaning of an idea consists of a semantic relation between the idea and the object which is external to, and unlike it. These three propositions together entail that the homunculus cannot grasp the meanings, the semantics, of ideas. Hence the scepticism that the earliest critics had already seen to be implicit in Descartes, Malebranche and Locke. The Hausmans’ suggestion is that Berkeley, along with other early modern philosophers, accepts the first two propositions, but, in order to avoid scepticism, he rejects the third. When he does this, he eliminates the entities that constitute the external referents of ideas. Since now there are no external objects, only ideas, it follows that there are only two categories of entity that exist, to wit the ideas and the homunculus that grasps their meaning. But ideas, on the metaphysics of understanding developed by early modern philosophers, are precisely those entities that are present to the wee inner homunculus. For, the latter is introduced precisely in order to account for the understanding, not of external objects, but of ideas. Hence, the existence of all entities besides the homunculus consists of their being present to the homunculus: their esse is percipi. In other words, on the Hausmans’ proposed reading of Berkeley, his idealism follows immediately from his attack upon scepticism, given, that is, Berkeley’s commitment to the metaphysics of understanding that Dennett ascribes to the early modern philosophers. Other aspects of Berkeley’s thought also fall into place. Thus, according to the Hausmans’ proposal, Berkeley rejects the claim of Descartes, Malebranche and Locke that ideas represent, or mean, objects that

176 are unlike themselves. This rejection is embodied in the Berkeleyan formula that only an idea can be like an idea. By making the idea itself the object that is to be known, Berkeley eliminates the semantic problem that is the root of the scepticism: there is no semantic problem because there is no external reference. The meaning of the idea is, as it were, the idea itself. We now understand the weight that Berkeley gives to the principle that only an idea can be like an idea when he argues for his idealism. Moreover, the metaphysics of understanding developed by the early modern philosophers required the homunculus to grasp the meaning of the ideas with which it is presented it must select that meaning. This it will do by means of yet another idea, an act of awareness through which the mind, that is, the wee homunculus, grasps the meaning or comes to understand the idea that is before it. When Berkeley eliminates the external reference of ideas, and makes the idea mean nothing beyond itself, he also eliminates the need for such an act to be present to the homunculus when it grasps the meaning of the idea. This accounts for Berkeley’s collapse of the act/object distinction when he presents his argument against the philosophies of Descartes, Malebranche and Locke. This account of Berkeley’s thought has the virtue of avoiding the problem of the inherence reading of Berkeley’s idealism. The latter argues that Berkeley's idealism arises from his falling into the thesis that ideas inhere in mental substances. This runs into the objection that if ideas inhere in mental substances, then the ideas will be predicable of the mind, something which Berkeley explicitly denies: sensations are in the mind by way of idea and not by way of mode.17 The Hausmans’ suggestion does not require the relation between the idea and the homunculus to be that of inherence. The problems of the inherence thesis are thus avoided. Moreover, we discover the way in which to avoid Berkeley’s idealism. The problem is not so much specific to his own metaphysics as it is one that lies more deeply within the general metaphysics of understanding of the early modern philosophy that Berkeley himself accepts. The problem is the homunculus. This wee inner person has only ideas present to it. This creates the problem of how it can grasp the semantics or meaning of the ideas that are present to it, that is, recognize the external referents of the ideas. Once the homunculus is eliminated we can have both ideas as internal representations and external objects that are meant by these ideas. Thus, when one eliminates the homunculus one eliminates the problems that are created by the metaphysics of understanding when it introduces an understander whose awareness is restricted to ideas alone: get rid of the

177 homunculus and both the idealism and the scepticism that motivates that idealism disappear. But the homunculus should on empiricist grounds be eliminated, as Hume saw. The problem is not so much the homunculus as what we should replace it with if we are to account for how persons understand meanings not only of external objects but of inner representations. 2. Problems with Berkeley as Empiricist Ingenious as this reading that the Hausmans give to Berkeley may be, it does not, it seems to me, do justice to what Berkeley was trying to do, nor, indeed, to what early modern philosophy was trying to do. In the first place, Berkeley’s argument against material objects does, contrary to what the Hausmans’ assert, not solve what they call the semantic problem. In the second place, Dennett’s reading of early modern philosophy is far too narrow to do justice either to the problems that early modern philosophy was addressing or to the solutions that they offered. For the moment take for granted that Berkeley eliminates material substances. That transforms ordinary objects, as he well understands, into collections of sensible events. (These events are, of course, ideas, given the argument for idealism.) These collections are not merely momentary but are spread out through time: an ordinary thing is a lawfully, temporally, and spatially ordered sequence of sensible events. Where previously things were substances, they are now analyzed as processes consisting of many parts, all sensible, and, therefore, all separable. To perceive an object is to be aware of a momentary sensible event – an idea if you wish – which is part of the larger sequence that is the object. We perceive a tree which is a whole, but are aware through our senses of only the part of the tree that is presented to us, and not, for example, the far side. The momentary sense impression, or, in the early modern terminology that Berkeley adopts, the idea of which we are directly aware, it must be insisted, represents the object which we are said to perceive but which is not, as such, presented to us in sense experience.18 We are in fact aware in sense of only part of the whole that we perceive, and that part is the representative in consciousness of the whole. There is, therefore, a semantic relation between the idea of which we are directly aware and the object that we perceive. Berkeley does eliminate any semantic relation between the idea and an object no part of which we ever do, or even can, be aware, but he does not eliminate the semantic relation between the presented idea and something not presented but which that idea represents, namely, the unperceived parts of the object

178 perceived.19 The Hausmans’ notion that Berkeley solves a semantic problem by eliminating the external reference of ideas is, therefore, mistaken: for, whatever Berkeley does, he does not eliminate the need for ideas to stand in semantic relations to things other than themselves, things which are collections of ideas, most of which are not given in sense experience, and of which, by virtue of those semantic relations, they are the representations in consciousness. The problem for early modern philosophers is not with the notion that ideas represent but that they represent objects that these philosophers argued had to be of a sort, which, as it turns out, creates particular difficulties. According to the traditional view deriving from Aristotle, and still defended by such Berkeleyan contemporaries as John Sergeant,20 “like knows like”: the object known comes to be such by virtue of a causal process through which its properties come to be present in the mind. But the substance philosophy of the early modern philosophy creates problems about knowing that go beyond the traditional Aristotelian account of knowledge. Whatever is the specific account one gives for the ontology of the knowing situation, one cannot avoid the fact that knowledge does involve a semantic relationship between what is present to consciousness and the object or objects known. Whatever else this means it means that there is a semantic pattern connecting the object known to the conscious state that is the knowing of it. By virtue of this pattern, the object known is able in a system-entry transition to evoke the conscious state that constitutes the knowing of it. The obtaining of the semantic relation thus requires a causal interaction between the known and the knower. The crucial point is that the metaphysics of substances adopted by the early modern philosophers renders any such causal interaction impossible. Here two points are relevant.21 In the first place, the early modern philosophers argued that there were two sorts of object, namely minds and body or bodies. The essential property of mental substance was thought, the essential properties of body were either extension alone or extension and solidity. According to Descartes and Malebranche – Locke wavered on this in the correspondence with Stillingfleet – extension and thought are contraries: if characteristics of one sort are present then characteristics of the other sort are excluded. It therefore follows that properties of body cannot be in the mind, and the former can, therefore, not be known by the latter. In the second place, mind and body are separate substances, independent of each other. To say this is to say that the one could exist and remain identical with itself even if the other ceased to exist. It follows that the substances can exemplify only non-

179 relational properties.22 That being so, it follows in particular that there can be no causal transaction between them.23 Transeunt causation, though not immanent causation, is impossible on the account of substances adopted by the early modern philosophers. There is, in short, no possibility that a causal process could bring it about that properties of body could come to be present in the mind, and, more generally, no possibility that there could be any causal process of the sort that must obtain if there is to be a semantic relation between the object known and the state of consciousness which is the knowing of it. The sceptical conclusions are inevitable, as Berkeley clearly saw. Berkeley removes this looming scepticism by giving up the ontology that is its source: he rejects the substance analysis of things. This in turn requires him to re-think the relation between the representative in consciousness of the thing and the thing it represents. In place of the ontology of substances Berkeley offers his analysis of things as wholes consisting of sensible parts. Some of these parts we actually sense; these sensible parts that are present in consciousness are what Berkeley, following the tradition, calls ideas. The relation between the thing represented by the sensed sensible part and that part is not that of an external substance causing the part to come to be in a second substance. Rather, the relation between the sensed part and the whole which is the thing it represents is that of part to whole, and, more specifically, the pattern or regularity among the parts of the whole that defines the sort of whole that it is, e.g., a tree in a quad. This avoids the problem of substantial interaction. Furthermore, the representative is a sign of the whole, that is, of all the other parts that constitute the whole. These other parts are also sensible, having precisely the same sorts of characteristics that the idea has. There is therefore no problem of one kind of thing making a toto caelo different kind of thing known. So Berkeley's radical innovation in ontology solves, or rather dissolves, the sceptical crisis that was implicit in early modern philosophy. This of course assumes that the representatives in consciousness of ordinary things are objective parts of those things. The brown of which we are conscious is precisely the property brown that characterizes the tree in the quad. Indeed, as Berkeley insists,24 this is precisely what common sense would seem to maintain. Moreover, there is no reason a priori to think that this should not be so. It is simply dogma to hold that the representation in consciousness is an entity which is not part of the thing known. It is of course a dogma that receives support from the traditional substance account of knowing, so it is not surprising to find Descartes, Malebranche

180 and Locke accepting it. There is less excuse for Dennett. To be sure, some representations are not parts of ordinary things – think of after images or cases of perceptual error, Macbeth’s dagger for example, cases which Berkeley himself recognized. But from the fact that some sensible contents are not parts of things it cannot be concluded that none are. The trick is to distinguish those which are and those which are not. On Berkeley’s view, there is not much of a problem here: it is simply a matter of coming to learn the regularities or patterns among sensible contents. Through experience we become aware that some entities that we sense are parts of patterns that define ordinary things, while others do not fit into such patterns. There are no more problems about this than there are problems about learning that smoke means fire. At least, there are no problems of principle. And equally, there is no problem of principle in maintaining that the ideas that represent things are parts of those things. Dennett, in fact, simply takes for granted that representations are not parts of the thing known. Dennett is also wrong about the homunculus. Ideas or sense impressions are of course intrinsically meaningless. To say that they have meaning, including a semantics, is to presuppose that there is a person who understands them. But there is no reason why this person needs to be a wee inner person; it can just as well be taken to be the ordinary person – just as the known can be taken to be the ordinary tree. Hume saw this clearly. A person is intimately related to his or her own body, and is capable of monitoring and controlling, within limits at least, both its own behaviour and its own mental representations. Thus, the patterns that define the system transitions are themselves rule governed. That is, if we have the pattern (*) all A are B then the rulish thought that A’s ought to be B, or, in symbols, (**) O (all A are B) – ‘O’ indicating the operator “it is obligatory that” – brings it about that, generally at least, (*) is true.25 These rulish thoughts bring it about that behaviour and thought – reality – conforms to their content. Rules of this sort were explored in detail by Hume, both in the context of social norms such as property and promise keeping – his discussion of the latter is of particularly important26 – and in the context of cognition – see his "Rules by which to judge of causes and effects."27 The set of such standards and rules forms a passably coherent set that constitutes a standard to be lived up to. Each person represents him- or herself as an “I” having both a body and a mental life with a past and a future, an “I” which has both an actual life and

181 an ideal standard for that life, an “I” which is aware of its own behaviour and mental life and attempts, often at least, to bring that behaviour into line with the standard, the “ego ideal” if you wish, that it sets before itself. This idea of oneself is of course complex. When Hume rejects the notion of a self in Book I of the Treatise it is the idea of the self as a simple substance. But in Book II he goes on to point out that for purposes of discussing the passions and our system of moral values and rules we do need to refer to our idea of the self. This is a complex idea – the self is “that succession of related ideas and impressions of which we have an intimate memory and consciousness” (Treatise, p. 277), “that individual person of whose actions and sentiments each of us is intimately conscious” (p. 286); it includes an awareness of the body – he speaks of "the qualities of our mind and body, that is, self” (p. 303); and it is, Hume says, an idea of which we are constantly aware – “the idea, or rather impression, of ourselves is always intimately present with us, and our consciousness gives us so lively a conception of our own person that ’tis not possible to imagine that anything can in this particular go beyond it” (p. 317). Hume had earlier pointed out that the identity of a person is not that of a simple substance but rather teleological, analogous to the sort of unity that we attribute to plants and animals (p. 254); the imagination leads us to mistake the succession of related objects for an identical object: we hide the interruption by feigning that the person is a simple substance, “something unknown and mysterious connecting the parts beside their relation” (p. 254).28 Such a substantial self violates the empiricist Principle of Acquaintance, and must therefore be exorcised from ontology. But this does not eliminate the self as a complex entity that unites the series of ideas and impressions into a teleological unity. For us the important point is that this concept of the self is not that of an inner homunculus. It is not the concept of a person that lies inside and controls the real person. It is simply the concept of a person who controls and monitors his or her own behaviour and, in many cases, the representations that are part of the person as a cognitive system. It is the person who understands, not the inner person or homunculus. Dennett suggests that the only unity that Hume provides to the self is the glue provided by association, and that this is inadequate to account for the capacity of the person to understand. However, this is to ignore the role that Hume assigns to the idea of the self in controlling our mental life and behaviour. Dennett seems not to have read in the Treatise beyond the end of Book I. If he had, he would have discovered that Hume developed certain themes for which Dennett is now claiming originality.

182

3. Berkeley's Ontology of Substances Berkeley also recognizes, I would like to argue, the Humean point that the self is a complex entity that unites the series of ideas and impressions into a teleological unity. In his Commonplace Book, the so-called Philosophical Commentaries, he tells us that “By Soul is meant onely a Complex idea made up of existence, will & perception in a large sense. therefore its is known & it may be defin’d” (PC, 154). This concept of the soul as a complex entity is the concept of a person. Berkeley came later to distinguish this concept from that of the soul as the substantial mind that, on his view, supports ideas: “The Concrete of the Will & understanding I must call Mind not person, lest offence be given, there being but one volition acknowledged to be God. Mem: Carefully to omit Defining of Person, or making much mention of it” (PC, 713). The person as a complex entity can provide the teleological glue that structures our mental life and behaviour. There is no need for Berkeley to introduce the substantial mind as an entity that can account for the fact that persons understand. Dennett’s scheme thus in no way helps us to understand the role of the substantial mind in Berkeley, nor in fact in early modern philosophy in general. In particular, contrary to the Hausmans, it does not help us to understand the Berkeleyan move to idealism. Why, then, does Berkeley have a substantial mind? The concept of a person as a complex but structured bundle can do the sorts of things that Dennett, Hume, Berkeley and the Hausmans all need, namely, account for the capacity of persons to understand. What it cannot do, according to Berkeley, is play a certain metaphysical role. To be sure, he apparently begins, as the Philosophical Commentaries make clear, with the thought that the concept of the person as complex can do this job. But he changes his mind, again as we have seen the Commentaries make clear. Berkeley makes evident in his last work what he thinks it is that the substantial mind does. In Siris29 he tells us that It is a doctrine among other speculations contained in the Hermaic writings that all things are One....If we suppose that one and the Mind is the universal principle of order and harmony throughout the world, containing and connecting all its parts, and giving unity to the system, there seems to be nothing atheistical or impious in this supposition (287).

And a little later he adds:

183

...although such phantoms as corporeal forces, absolute motions, and real spaces do pass in physics for causes and principles..., yet are they in truth but hypotheses, nor can they be the objects of real science. They pass nevertheless in physics, conversant about things of sense, and confined to experiments and mechanics. But when we enter the province of the philosophia prima, we discover another order of beings, mind and its acts, permanent being, not dependent on corporeal things, nor resulting, nor connected, nor contained; but containing, connecting, enlivening the whole frame, and imparting those motions, forms, qualities, and that order and symmetry, to all those transient phenomena which we term the Course of Nature (293).

Still later we read: ...the mind contains all, and acts all, and is to all created beings the source of unity and identity, harmony and order, existence and stability ( 295).

We see two things: first, that mind or self-consciousness is the source of order and unity among the separable entities given in sense experience; and second, that this unity derives from the unifying activity of an entity that is itself a unity, one, without separable parts. This notion, that the unity of the universe derives from the unity of self-consciousness, goes all the way back to one at least of Berkeley’s sources, Plotinus.30 The materialists, that is, the Stoics, argued that the world consists of matter in motion. The soul is the origin of the motion of the body, but it too is material. Plotinus argued that this was impossible. Matter, as lifeless, cannot move itself; in fact it could not even stay together in a particular configuration: “...body in itself could not exist in any form if soul-power did not; body passes; dissolution is in its very nature; all would disappear in a twinkling if all were body.” Indeed, “Matter itself could not exist [without soul]: the totality of things in the sphere is dissolved if it be made to depend upon the coherence of a body which, though elevated to the nominal rank of ‘soul’, remains air, fleeting breath..., whose very unity is not drawn from itself” (Enn, IV, 7, 3). Soul, as cause, contains within itself the forms or reasons of the things the temporally ordered diversity of which it produces. This produced order reflects in time the timeless order implicit in the cause: “If the leading principle of the universe does not know the future which it is of itself to produce, it cannot produce with knowledge or to purpose; it will produce just what happens to come, that is to say by haphazard. As this cannot be, it must create by some stable principle; its creations, therefore, will be shaped in the model stored up in

184 itself; there can be no varying...” (Enn, IV, 4, 12). Moreover, our consciousness is in the first instance, as indeed the materialists hold, a series of events in time. However, each of these events is related to the others in the series; and they are related moreover to each other as modifications of a single consciousness, a consciousness which is a consciousness of each of them and all of them. “There can be no perception without a unitary percipient whose identity enables it to grasp an object in its entirety” (Enn, IV, 7, 6). The conscious self of which the events are the modifications cannot be the series as a whole, since within the series the events are successive; but at any moment the consciousness of those events does not involve a succession: “...prior and past are in the things it [soul] produces; in itself nothing is past; all...is one simultaneous grouping of Reason-Principles” (Enn, IV, 4, 16). That is, while the consciousness of the series is a consciousness of a before and after, within that self-consciousness there is no before and after; within the consciousness there is no such relation between the components as there is between the events of which it is the consciousness. The consciousness which a self has of itself must lie outside the temporal sequence of events of which it is conscious. Self-consciousness must therefore be an entity outside the temporal changes of the self, an eternal entity to which the events in the temporal sequence are related as modifications. This position, which is essentially that of Berkeley, continued to be defended long after Berkeley. It was held by T. H. Green,31 for example, that a whole consisting of separable parts could have no unity; no part of such a whole could provide that unity. Where such a whole did have unity, the latter derived from a timeless self-consciousness that comprehended the whole. Green resisted Mill’s notion with respect to the self – shared with Dennett and Hume –, that a part could represent the whole, and thereby provide the unity. "To be conscious of it [one’s personal history],” Green tells us, “we must unite its several stages as related to each other in the way of succession; and to do that we must ourselves be, and distinguish ourselves as being, out of that succession.” It is only through our holding ourselves aloof, so to speak, from the manifold affections of sense, as constant throughout their variety, that they can be presented to us as a connected series, and thus move to seek the conditions of connection between them (p. 92).

But Green’s argument for this is not so much to deny Mill’s point as it is to insist that the relational structure of the self presupposes a self that is out-

185 side time. According to Green, both selves and material objects have as their basic constituents entities given in sense experience, or, as he calls them, feelings. On this point he is in agreement will Mill and with Hume. But, ...feelings are facts; but they are facts only so far as determined by relations, which exist only for a thinking consciousness and otherwise could not exist (p. 53).

Structure can be provided only by a thinking, conscious, self-conscious, unity or substance, in which the whole as such exists. It is this metaphysics of structure and not Dennett’s metaphysics of understanding to which we should relate Berkeley if we are to appreciate his idealism. In fact the problem of understanding is subsumed by Berkeley within the broader concept of structure. In the Philosophical Commentaries Berkeley returns again and again to the issue of unity and structure. Thus, in ¶ 71 we learn that “By immateriality is solv'd the cohesion of bodies, or rather the dispute ceases.” Space is a structure relating bodies, and so empty space is a meaningless notion: “Space wthout any bodies being in rerum natura, would not be extended as not having parts in that parts are assigned to it wth respect to body from whence also the notion of distance is taken, now without either parts or distance or mind how can there be space or anything beside one uniform no thing?” (¶ 96). Again, “The greatness per se perceivable of the sight, is onely the proportion any visible appearance bears to the others seen at the same time; of (wch is the same thing) the proportion of any particular part of the visual orb to the whole. but mark that we perceive not it is an orb, any more than a plain but by reasoning. This is all the greatness the pictures have per se” (¶ 204). If Berkeley is concerned about structure and unity, then he also has an explanation of it. What provides the structure is activity, that is, the will: “We cannot possibly conceive any active power but the Will” (¶ 154). ’Tis folly to define volition an act of the mind ordering./ for neither act nor ordering can themselves be understood without Volition (¶ 635).

And this is the same as Mind or Soul: “The soul is the will properly speaking & as it is distinct from Ideas” (¶ 478a). Power in fact is nothing more than the relation of cause and effect: “Power no simple Idea. it means nothing but the Relation between Cause & Effect.” (¶ 493) What we observe by

186 sense are only occasions, not causes (¶ 856, ¶ 853). The causes are the activities that produce the effect consequent upon the cause: “I say there are no Causes (properly speaking) but Spiritual, nothing active but Spirit” (¶ 850). These activities are the volitional activities of a substantial soul. What means Cause as dinstinguish’d from Occasion?/ nothing but a Being wch wills wn the Effect follows the volition. Those things that happen from without we are not the Cause of therefore there is some other Cause of them i.e. there is a being that wills these perceptions in us (¶ 499)

Cartesian body, that is, extended substance, is inert. Locke and Newton also make body inert. Moreover, Newtonian space is inert. On Berkeley’s view, none of these systems is able to solve the problem of structure. But there is such a problem: the ideas of which we are aware in sense experience display an order, and that implies, Berkeley is proposing, an active structuring. And sense gives no necessary connection. Qu: whether possible that those visible ideas wch are now connected with greater extensions could have been connected with lesser extensions. there seeming to be no necessary connexion between those thoughts (¶ 181)

In order to account for structure, one needs necessary connections, that in turn requires activity, and that in its own turn implies a substantial mind. Ideas are in minds because the mind provides the structure that organizes those ideas into structured unities. We thus see that the root of Berkeley’s idealism is the attempt to account for the order that we discover among the sense impressions that are presented to us. Berkeley begins, basically, with Locke’s empiricism. But the substantial mind that creates order is not given in the ordinary awareness through which we are aware of sensible particulars. What he has to do is struggle towards the conclusion that the self that gives order cannot be construed on empiricist lines as a complex entity, “a bundle or collection of different perceptions which succeed one another with an inconceivable rapidity and are in a perpetual flux and movement,” as Hume put it (Treatise, p. 252). Berkeley did attempt this: as he put it in the famous passage from his Commonplace Book, By Soul is meant onely a Complex idea made up of existence, will & perception in a large sense. therefore it is known & it may be defin’d (¶ 154).

But he struggled out of this position:

187

The substance of Body we know, the substance of Spirit we do not know not being knowable. it being purus actus (¶ 701).

it

Or again, Perception is passive but this not distinct from Idea/ therefore there can be no Idea of volition (¶ 756).

Ideas are contrasted to acts of thought or volitions (¶ 808). Ideas and the will of course go together (¶ 841, ¶ 842; cf. ¶ 577); the latter structure the former. But the will is not perceived, nor, therefore, is it the object of understanding (¶ 828, ¶ 829), that is, understanding in the sense of the empiricism of Locke and Hume, according to which one understands those things given in sense by subsuming them under the matter-of-fact regularities that we learn through experience. When Berkeley suggests that the self can be construed as a complex idea he is not so much anticipating the Humean conclusion that Dennett recommends to us, but rather struggling to escape the empiricism that he has inherited from Locke and that will forbid him any account of structure if he continues to adhere to it. We thus see that the problem that Berkeley is addressing is the traditional one of order, not that of understanding meaningful speech. Once we recognize this, then we see his idealism falling into place. We also see why the inherence account of that idealism, as defended by Allaire, Cummins and Watson, is substantially correct32: it simply appeals to the model of causation that the traditional account of order uses.33 Properties are in things, that is, inhere in things construed as substances, and the order among those properties is provided by the immanent causal activities of that substance, or by the transeunt causal activities of another substance.34 The traditional problem for the inherence reading is Principles, ¶ 35 49, where Berkeley denies that ideas are predicated of the mind. ...it may perhaps be objected that if extension and figure exist only in the mind, it follows that the mind is extended and figured; since extension is a mode or attribute which (to speak with the Schools) is predicated of the subject in which its exists. – I answer, those qualities are in the mind only as they are perceived by it; – that is, not by way of mode or attribute, but only by way of idea. And it no more follows the soul or mind is extended, because extension exists in it alone, than it does that it is red or blue, because those colours are on all hands acknowledged to exist in it, and nowhere else.

188

If ideas are not predicated of the mind, so goes the objection, then they cannot inhere in it, since, traditionally, inherence has been reflected in language by predication. Berkeley, however, continues the passage to argue that there is no necessary connection between predication and inherence: As to what philosophers say of subject and mode, that seems very groundless and unintelligible. For instance, in this proposition ‘a die is hard, extended, and square,’ they will have it that the word die denotes a subject or substance, distinct from the hardness, extension, and figure which are predicated of it, and in which they exist. This I cannot comprehend: to me a die seems to be nothing distinct from those things which are termed its modes or accidents. And, to say a die is hard, extended, and square is not to attribute those qualities to a subject distinct from and supporting them, but only an explication of the meaning of the word die. (Principles, ¶ 49)

Given the abolition of material substance, it is necessary to construe ordinary things as bundles of qualities, and to construe predication as reflecting the relation of part to whole. Nonetheless, predication does sometimes reflect the relation of mode to substance, as, for example, when a volition is predicated of a spiritual substance. The point is, of course, that even if predication does sometimes reflect the relation of property to substance, it is clear that Berkeley intends the relation of idea to substance to be other than that of the relation of, say, volition to substance. And that being so, it would seem that the inherence account of Berkeley’s idealism is wrong when it insists that that idealism arises precisely because Berkeley construes, if only unconsciously, the relation of idea to mind as that of property to substance. The correct response should be, I think, to note that Berkeley insists upon distinguishing two relations: ideas are in spiritual substances, and so are activities; but the latter are in substances by way of mode, whereas the former are in substances in another way. Once this is done then we recognize, I think, that this reflects a very traditional view of substances. The active form36 is, traditionally, said to be “predicated of” the substance, while the properties are “present in” the substance. When Berkeley insists upon two relations to the substance, he is simply following this tradition. Where he departs from the tradition is in his insistence that when an idea is “present in” a substance then it is not to be predicated of that substance. The latter is required by his non-traditional account of things as bundles of qualities. But from the fact that being “present in” a substance is no longer

189 reflected in predication, it does not follow that Berkeley is not construing the relation of being “present in”, that is, the relation of an idea to a spiritual substance as the relation of property to substance – with all the ontological freight which that implies, including what is necessary to justify the inherence account of his idealism. In particular, for a property to be “present in” a substance just is for it to be supported by the informed activity of the substance; inherence is, if you wish, inseparable from activity.37 Thus, reading Berkeley as introducing active substances in order to solve the problem of order enables one to meet at least some of the objections to the inherence account of his idealism. But if Berkeley solves the problem of structure by introducing active substances, he nonetheless still faces problems.38 There is of course the traditional problem of how we know substances. This problem he does solve – sort of – by reviving John Sergeant’s thesis about “notions.”39 More deeply, there is the problem of interaction. Berkeley’s account of ordinary things as wholes the parts of which are sensible particulars, and his idealism, together solve the problem of the interaction of mind and body. He must still, however, deal with the problem of the interaction of substances, since he still has interacting minds. Unlike Leibniz, Berkeley never faced up to the problem of interaction. Until he does, he has not solved the problem that he tried to solve.

190 Endnotes to Study Four

1.Alan Hausman and David Hausman, “A New Approach to Berkeley’s Ideal Reality,” in R. Muehlman, ed., Berekeley’s Metaphysics: Structural, Interpretative and Critical Essays (University Park, PA: Pennsylvania State University Press, 1995), pp. 47-66. This nicely done essay is provocative and stimulating, bringing together a number of themes from philosophy and its history. 2.E. B. Allaire, “Berkeley's Idealism,” Theoria, 29 (1963), pp. 229-244; P. Cummins, “Perceptual Relativity and Ideas in the Mind,” Philosophy and Phenomenological Research, 24 (1963), pp. 204-214; and R. A. Watson, “Berkeley in a Cartesian Context,” Revue Internationale de Philosophie, 64 (1963), pp.381-394. For criticism see: R. Muehlmann, “Berkeley’s Ontology and the Epistemology of Idealism,” Canadian Journal of Philosophy, 8 (1978), pp. 89-111, and also Berkeley’s Ontology, (Indianapolis, IN: Hackett Publishing, 1992); Nathan Oaklander, “The Inherence Interpretation of Berkeley: A Critique,” The Modern Schoolman, 54 (1977), pp. 261-269; and Harry Bracken, “Some Problems of Substance among the Cartesians,” American Philosophical Quarterly, 1 (1964), pp. 129-137. Cummins changes his mind in “Berkeley’s Ideas of Sense,” Noûs, 9 (1975), pp. 55-72. George Pappas reviews the whole set of issues and offers a further critique of the inherence account in his “Ideas, Minds and Berkeley,” American Philosophical Quarterly, 17 (1980), pp. 181-194. The best surveys, however, are the judicious papers by Alan Hausman, “Adhering to Inherence: A New Look at the Old Steps in Berkeley’s March to Idealism,” Canadian Journal of Philosophy, 14 (1984), pp. 421-443; and R. Muehlmann, “Introduction” to his Berkeley’s Metaphysics: Structural, Interpretative, and Critical Essays, (University Park, PA: Pennsylvania State University Press, 1995), pp. 1-22. See also, R. Muehlmann, “The Substance of Berkeley’s Philosophy, ibid., pp. 89-106. 3.George Pappas, “Abstract Ideas and the ‘Esse is Percipi’ Thesis,” Hermathena, 139 (Winter, 1985), pp. 47-62. 4.D. Dennett, Brainstorms (Boston: Bradford Books, MIT Press, 1978). 5.Dennett, Brainstorms, p. 122. 6.The example is from May Brodbeck, who long emphasized this point that Dennett has taken up. Brodbeck learned the idea from G. Bergmann, who deserves to be credited for this idea, now presented to the world by Dennett, who would like to claim more originality than seems reasonable: the problems were located, debated and in effect solved before Dennett took up the issues, 7.Dennett, Brainstorms, p. 119.

191

8.This way of speaking derives from Wilfrid Sellars. For a brief account, see below, the essay “Marras on Sellars on Thought and Language.” 9.Dennett, p. 122. 10.Dennett, p. 122. 11.Dennett, p. 122. 12.Dennett, p. 88. 13.Dennett, p. 122; see also p. 101. 14.See D. Hume, Treatise of Human Nature, ed. L. A. Selby-Bigge, revised P. H. Nidditch (Oxford: Oxford University Press, 1978), p. 252. 15.Dennett, p. 101. 16.Dennett, p. 122. 17. Cf. Berkeley, Principles of Human Knowledge, in vol. 2 of The Works of George Berkely, Bishop of Cloyne, 7 vols., ed. A. A. Luce and T. E. Jessop (London: Nelson, 1948–57), ¶ 16. 18.Cf. D. Lewis, “Moore’s Realism,” in L. Addis and D. Lewis, Moore and Ryle: Two Ontologists (The Hague: Nijhoff, 1965), Ch. 3 and 4; and F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill (Toronto: University of Toronto Press, 1990), Ch. 5. 19.This point made by H. Bracken, Berkeley (London: Macmillan, 1974), Ch. 9, esp. pp. 95-6, where he emphasizes that the perceptual object, being extended in time, is not wholly presented to one, and that, if Berkeley did succeed in eliminating the scepticism implicit in the representationalist’s veil of ideas, he did not succeed in eliminating the possibility of error that arises from the veil of time. But for a discussion of this issue and its relation to scepticism, see the essay, “Acquaintance, Ontology and Knowledge,” this volume, above. 20.John Sergeant, Solid Philosophy Asserted against the Fancies of the Ideists (London, 1698). 21.See R. A. Watson, The Downfall of Cartesianism (The Hague: Nijhoff, 1966).. 22.Wilson, “Weinberg’s Refutation of Nominalism,” Dialogue, 8 (1969), pp. 460-474. 23.Weinberg, “The Concept of Relation,” in his Abstraction, Relation, and Induction (Madison, WI: University of Wisconsin Press, 1965), p. 126.

192

24.In his Philosophical Commentaries, in vol. 1 of The Works of George Berkeley; see 19. The Philosophical Commentaries are Berkeley’s commonplace book, in which he worked out his philosophical position prior to writing the New Theory of Vision and the Principles of Human Knowledge. We shall refer to the Commentaries by paragraph number. 25.Cf. F. Wilson, in the present volume, below, “Marras on Sellars on Thought and Language,” following W. Sellars, “Notes on Intentionality,” in his Philosophical Perspectives (Springfield, Ill.: Charles C. Thomas, 1967), and “Reflections on Language Games,” in his Science, Perception, and Reality (London: Routledge and Kegan Paul, 1963). 26.Cf. F. Wilson, “Association, Ideas and Images in Hume,” in P. Cummins and G. Zoeller, eds., Minds, Ideas and Objects (Atascadero, Calif.: Ridgeview Publishing Co., 1992), pp. 255-74. 27.Treatise, I, III, xv. 28.Cf. F. Wilson, “Hume’s Fictional Continuant,” History of Philosophy Quarterly, 6 (1989), pp. 171-188. 29.Siris, in Berkeley’s Works, volume 5. 30.References are to Plotinus, The Enneads, trans. Stephen MacKenna, Fourth Edition, rev. S. Page (London: Faber and Faber, 1969)’ hereafter, Enn, volume, book and section numbers. 31.T. H. Green, Prolegomena to Ethics (Oxford: Oxford University Press, 1906); hereafter, G, followed by page number. 32.Watson, “Berkeley in a Cartesian Context,” points out in detail how Berkeley repeatedly talks as if ideas inhered in substances. 33.Similar points are made in Kenneth P. Winkler, “Unperceived Objects and Berkeley's Denial of Blind Agency,” Hermathena, 139 (1985), pp. 81-100. 34.The connections between order or relations and spirit is emphasized by Berkeley in his additions to the Principles in the second edition; see E. J. Furlong, “Berkeley on Relations, Spirits and Notions,” Hermathena, 107 (1968), pp. 60-66. 35.Cf. Pappas, “Ideas, Minds and Berkeley.” 36.As Winkler, “Unperceived Objects,” points out, Berkeley rejects blind agency; that means that the activities of spiritual substances are always informed, or intrinsically directed.

193

37.This point, too, is emphasized by Winkler, “Unperceived Objects,” p. 87ff, against various comments by Bennett, who accuses Berkeley of using the word ‘depend’ to blur the difference between “the ownership of ideas” and “what causes ideas to be had by minds” (J. Bennett, Locke, Berkeley, Hume: Central Themes [Oxford: Oxford University Press, 1971)], p. 168). 38.While, as I have been arguing, Berkeley does propose an account of relations as contributed by the substantial mind, I think it is also true that he does not develop the position as systematically as one would have hoped. As Luce and Jessop have remarked, “Berkeley nowhere develops this view of relations, which seems to imply that relations among ‘ideas’ are not discovered but instituted by the mental act, or at any rate that activity of relating somehow enters into the content of the relation” (A. A. Luce and T. E. Jessop, The Works of George Berkeley, vol. II, p. 106, note 1). 39.Cf. R. Grossmann, “Digby and Berkeley on Notions,” Theoria, 26 (1960), pp. 1730; and D. Flage, Berkeley’s Doctrine of Notions (New York: St. Martin’s Press, 1987).

Five Bradley’s Account of Relations and Its Impact on Empiricism* Bradley, late in his career, had occasion to respond to criticisms that Russell had made of his idealism.1 On Bradley’s view, Russell had returned to the atomism of Hume and Mill, a view that he (Bradley) believed that he had long ago refuted.2 In effect Bradley is arguing that his own work has had no real or lasting impact on British philosophy. Interestingly enough, this view seems to be shared by David Pears who, in his relatively recent book on Russell, argues that Russell restores the Humean empiricism that had been forced underground by several decades of idealism: “what [Russell] did was to take over and strengthen the type of empiricism whose most distinguished exponent had been David Hume.”3 To be sure, Russell did respond to Bradley,4 but the empiricism that Russell defends is essentially that of Hume. Pears thus seems to share Bradley’s view that Bradley existed but had no lasting impact. This view is quite wrong. It is certainly true that the Russell that emerged in the first two decades of the century was thoroughly empiricist, and that this empiricism has strong roots in the earlier empiricism of Hume and the Mills. But it certainly was not the same empiricism. Russell made important contributions that enabled empiricism to overcome defects that were present in its earlier versions. Those defects had also been noticed by Bradley. Bradley, too, had offered solutions to those problems. Those solutions were, however, idealist rather than empiricist. Russell’s contribution consisted, on the one hand, of responding on behalf of empiricism to the criticisms of Bradley while, on the other hand, criticizing Bradley’s idealist response to the empiricism of Hume and the Mills.5

* Appeared originally as “Bradley’s Impact on Empiricism,” in J. Bradley, ed., Philosophy after F. H. Bradley (Bristol:Thoemmes Press, 1996), pp. 251-82.

196 (I) The Earlier Empiricism There are two aspects of the empiricism of Hume and the Mills to which Bradley attended. One was the doctrine of relations, the other the doctrines of associationism. We shall be concentrating on the former.6 After examining this earlier empiricist ontology of relations, we turn to the argument developed by the empiricists against necessary connections. For Bradley also replied to this argument. (a) Relations The empiricists took over the nominalistic account of relations that had dominated philosophy since Aristotle and before.7 Consider the relational statement that (@) a is R to b or, in symbols that Russell has made familiar to us, R(a,b) On the nominalist account, such a statement has a two-fold analysis. On the one hand, there are the objective truth conditions, the objective facts concerning a and b which determine whether the statement is (objectively) true or false. On the other hand, there is the subjective mental state that the use of the relational statement expresses. As for the former, the objective facts represented by (@) are non-relational: (#) a is r1 and b is r2 or, in symbols, r1(a) & r2(b) The non-relational properties r1 and r2 are the (objective) foundations of the relation. As for the subjective state that the use of (@) expresses, this is a judgement of comparison. Notice that, upon this account of relations, if one of the relata, say b, ceases to exist, so will the relation – objectively, together of course with the possibility of comparison – but this happens with no change in the other relatum: b’s ceasing to exist, or ceasing to be r2, is compatible with a continuing unchanged, and in particular with a remaining r1. This account of relations was explicitly stated by Locke, who tells us that The nature therefore of Relation, consists in the referring, or comparing two things, one to another; from which comparison, one or both comes to be

197 denominated. And if either of those things be removed, or cease to be, the Relation ceases, and the Denomination consequent to it, though the other receive in it self no alteration at all, v.g. Cajus, whom I consider to say as a Father, ceases to be so to morrow, only by the death of his Son, without any alteration made in himself (Essay, II, xxv, 5)

A little later he adds that ... there can be no Relation, but betwixt two Things, considered as two Things. There must always be in relation two Ideas, or Things, either in themselves really separate, or considered as distinct, and then a ground or occasion for their comparison (Essay, II, xxv, 6).

This view of relations and of relational predication is the one that Russell was later to call the “monadistic” account of relations,8 making reference to Leibniz’s adoption of the same position. The position is adopted by Hume, who considers relations as at once philosophical, that is, objectively, or as natural, that is, subjectively. The word Relation is commonly used in two senses considerably different from each other. Either for that quality, by which two ideas are connected together in the imagination, and the one naturally introduces the other ...; or for that particular circumstance, in which, even upon the arbitrary union of two ideas in the fancy, we may think proper to compare them (Hume, Treatise, I, i, 5, p. 13).9

The same monadistic account of relations was also adopted by James Mill In his Analysis of the Phaenomena of the Human Mind, the chapter devoted to ‘Relative Terms’ states that .10

If it is asked, why we give names in pairs? The general answer immediately suggests itself; it is because the things named present themselves in pairs; that is. are joined by association.

The second edition was prepared for publication and edited by the younger Mill, who comments in his Notes on the “set of facts, which is connected by both the correlative terms” that constitute the fundamentum relationis; and goes on to remark that objects are said to be related, when there is any fact, simple or complex, either apprehended by the senses or otherwise, in which they both figure. Any objects, whether physical or mental, are related, or are in relation, to one another, in

198 virtue of any complex state of consciousness into which they both enter; even if it be a no more complex state of consciousness than that of merely thinking them together.11

Prior to Bradley, it is safe to say, the empiricists explicitly adopted the monadistic account of relations. This account was applied to all relations. Thus, for example, it was applied to resemblance. If Peter and Paul resemble each other in respect of being red, then the relation of resemblance is analyzed objectively into two non-relational facts, that of Peter being red and that of Paul being red, and subjectively into the comparison of Peter and Paul in respect of colour.12 It was also applied to causation. Hume argued in detail that the judgment that A’s cause B’s is to be analyzed objectively, that is, in Hume’s terms, as a relation considered philosophically, into a regularity, a mere conjunction if you wish, between A and B, and subjectively, or, in Hume’s terms, as a relation considered naturally, into an association between (the impression or the idea of) A and (the idea of) B.13 The Mills adopt the same view of causation.14 But Hume applies it to all sorts of relations, including spatial and those of degree, e.g., of hot and cold.15 Again he is followed in this by the Mills.16 (b) The Argument Against Necessary Connections A is necessarily connected to B just in case that it is impossible for the one to exist and the other not. The empiricist claims that there are no necessary connections among matters of fact, things that can be known through ordinary sense experience. The empiricist argument against necessary connections among things was first fully developed by Hume. There are two threads to this argument and, it is important to note, these are not carefully distinguished by Hume. The first thread of the argument is this. If we take the idea of some cause and the idea of its effect, then there is no contradiction in supposing the former to exist and the latter not: “the actual separation of these objects is so far possible, that it implies no contradiction nor absurdity; and is therefore incapable of being refuted by any reasoning from mere ideas; without which ’tis impossible to demonstrate the necessity of a cause” (Treatise, p. 80). There is no contradiction in separating the ideas because these ideas derive from perceptions and in perception there is no necessary connection that is presented to us: “as all our ideas are deriv’d from impressions, or some precedent perceptions, ’tis impossible we can have

199 any idea of power and efficacy, unless some instances can be produc’d, wherein this power is perceiv’d to exert itself .... [But] these instances can never be discover’d in any body ...” (Treatise, p. 160; Hume’s italics). Hume’s appeal here is to the empiricist Principle of Acquaintance 17 (PA). Among the entities that are given to us in our ordinary sensory experience or inner awareness we discover no connections such that one of these entities cannot exist unless another of these entities exists. In identifying the property that characterizes the cause we do not have to refer as a matter of necessity to the property that characterizes the cause, one to which it is necessarily tied; the properties are presented as logically selfcontained rather than as necessarily tied to one another. We are, in other words, acquainted with no such entity as a necessary connection in ordinary experience, and there is therefore no necessary connection among the ideas that we use to describe whatever is given to us in experience. This argument from acquaintance derives from Locke. Locke considers the regular activities of external substances. These include the production of the ideas of the secondary qualities, the simple ideas of red, sweet, etc. For these activities to be explained as those who defend necessary connections require, there must be necessary connections between red, sweet, etc., and the natures or real essences of the substances that cause these qualities to appear. These necessary connections must be both ontological, in the entities themselves, and epistemological, giving us, when in the mind, scientific knowledge of those entities. But, Locke argues, we grasp no such connections. ’Tis evident that the bulk, figure, and motion of several Bodies about us, produce in use several Sensations, as of Colours, Sounds, Tastes, Smells, Pleasure and Pain, etc. These mechanical Affections of Bodies, having no affinity at all with those Ideas, they produce in us, (there being no conceivable connexion between any impulse of any sort of Body, and any perception of a Colour, or Smell, which we find in our Minds) we can have no distinct knowledge of such Operations beyond our Experience; and can reason no otherwise about them, than as effects produced by the appointment of an infinitely Wise Agent, which perfectly surpasses our Comprehension (Essay, IV, iii, 28).

Locke’s appeal to a Principle of Acquaintance is clear. Properties are presented as logically self contained; there is nothing about them as presented that requires us when we are identifying them to refer as a matter of necessity to other properties, those to which they are necessarily tied.

200 We are, in other words, not presented with necessary connections, and we can therefore, by PA, not admit them into our ontology. But there is a second strand in Hume’s argument against necessary connections. It is this. There is no necessary connection between cause and effect for the reason that ... as all distinct ideas are separable from each other, and as the ideas of cause and effect are evidently distinct, ’twill be easy for us to conceive any object to be non-existent this moment, and existent the next, without conjoining to it the distinct idea of a cause or productive principle (Treatise, p. 79).

This is not mere logical separability. It is, rather, the stronger claim that the effect can exist as it is, that is, unchanged, were the cause to exist or not. To be sure, this latter, stronger claim entails the former. But the converse does not hold. It is one thing to hold that there is nothing about Caius minor qua effect that entails that Caius major qua cause must exist. In that case it would be contradictory to suppose that Caius minor qua effect exists while Caius major qua cause does not. But this does not entail that Caius minor would have all the same properties, that is, remain unchanged, if Caius major did not exist. This is what Hume is now claiming: the effect and the cause can exist independently, and unchanged, even if the other does not exist, or were to cease to exist. This latter is, to make the point again, a stronger claim than the former. If a being F causes b to be G, then the first argument claims that it is not self-contradictory to assert both that a is not F and that b is G. This does not entail the b would be unchanged were a to cease to exist. But this latter is what is asserted by the second thread of the argument. The first strand of the argument against necessary connections proceeded on the basis of an appeal to the empiricist's PA. The second strand has a different basis. Quite clearly it is rooted in the monadistic account of relations that Hume had inherited from the tradition. Hume is clearly assuming that all relational statements can be analyzed into nonrelational statements about foundations. For, on that account, one of a pair of relata could cease to exist while the other remains unchanged. Hume notes that his opponents – he considers the Cartesians to be the most significant among those who hold that there are necessary connections – admit that necessary connections are not given in ordinary sensory experience.

201 The small success, which has been met with in all attempts to fix this power, has at last oblig’d philosophers to conclude, that the ultimate force and efficacy of nature is perfectly unknown to us, and that ’tis in vain we search for it in all the known qualities of matter (Treatise, I, iii, 14, p. 159).

The entity thus transcends ordinary experience. This implies that it must be located ontologically outside ordinary experience and, epistemologically, that we must know it by some means other than ordinary experience. How this will be done will depend upon one's specific views. The Cartesians locate the necessary connections in the activities of the Deity, and as a means of knowing these invent the mechanism of innate ideas. ... ’tis impossible we can have any idea of power and efficacy, unless some instances can be produc’d, wherein this power is perceiv’d to exert itself. Now as these instances can never be discover’d in body, the Cartesians, proceeding upon their principle of innate ideas, have had recourse to a supreme spirit or deity, whom they consider as the only active being in the universe, and as the immediate cause of every alteration in matter (Treatise, p. 160; Hume’s italics).

Others, such as the Aristotelians, locate real powers and necessary connections in finite substances, and invent special abstract ideas as ways of knowing these powers. In general, “...since we can never distinctly conceive how any particular power can possibly reside in any particular object, we deceive ourselves imagining we can form any such general idea” (Treatise, p. 162). The result of this analysis of the thesis that there are necessary connections is the conclusion that upon that view we end up in a state of complete scepticism with respect to causation: “... when we speak of a necessary connection betwixt objects, and suppose, that this connexion depends upon an efficacy or energy, with which any of these objects are endow’d; in all these expressions, so apply’d, we have really no distinct meaning, and make use only of common words, without any clear and determinate ideas” (Treatise, p. 162; Hume’s italics). In contrast, with Hume’s own account of causation, conforming to the empiricist’s PA, there is no similar impossibility of knowing causal relations. To be sure, since causal judgments are judgments of general regularity about a population that go beyond the limited sample to which we, as finite human beings, are inevitably restricted, they cannot be known infallibly. But at least they can be known fallibly, which is rather more than is possible on the view that causation is a matter of necessary connections. The introduction of transcendent entities inevitably leads to scepticism;

202 conforming one’s ontology to the empiricist’s Principle of Acquaintance permits one to avoid such radically sceptical conclusions.18 (II) Bradley’s Critique of Empiricism Bradley criticized the monadistic account of relations that Hume and the Mills had adopted.19 In the context of criticizing the account of relations he proposes his own alternative account. This account is that which Russell was later to call the “monistic” account.20 (a) Relations Bradley’s contention is that the monadistic account of relations in effect eliminates everything that is makes a relation relational, and he concludes that, since clearly there are facts which are essentially relational, the monadistic account of the empiricists simply will not do. Consider again the relational statement that (@) a is R to b or, in symbols, R(a,b) The monadistic account proposes to reduce this to (#) a is r1 and b is r2 or, in symbols, r1(a) & r2(b) On this view, the objects a and b are independent in the sense that if the one ceased to exist the other would continue to exist unchanged. Bradley argues that there are relational wholes in which the relata are not in this way independent. He proposes that genuine relations are incompatible with the independence that is a consequence of monadistic view. “... [A] mode of togetherness such as we can verify in feeling,” he tells us, “destroys the independence of our reals.”21 Conversely, if we do make the relata independent or absolute, then we destroy their relatedness: “Relations are unmeaning except within and on the basis of a substantial whole, and related terms, if made absolute, are forthwith destroyed” (A & R, p. 125). Bradley’s point is that, if (@) is genuinely relational, then what is predicated of a is the property “R to b”:

203 (@') a is (R to b) and what is predicated of b is the property “R-ed by a”: (@") b is (R-ed by a) This being so, if a ceases to exist, then perforce b ceases to have the property of being “R-ed by a”. That is, if a ceases to exist then, when one has genuine relations, it is not the case that b remains unchanged. In this sense, genuine relations always affect the being of things. As Bradley puts it, “if the relations in which the reals somehow stand are viewed as essential, that, as soon as we understand it, involves at once the internal relativity of the reals. And any attempt to maintain the relations as merely external must fail” (A & R, p. 125). Nor is this simply a matter of trying to turn relations into properties, that is, non-relational properties, as Pears naively suggests.22 This is what the monadistic view attempts, and it is precisely this that Bradley is arguing against. Pears tells us that Bradley ... would argue that, if two people are married, the relation must make some difference to them. If this contention is interpreted in the ordinary way, nobody would deny it. But he means that the two people must each possess a property which makes the relationship intelligible. If this were interpreted to means that they mst possess traits of character which would explain the marriage, it would seldom be denied. But he means something very different. He means that the relation can be understood only if it is taken to be identical with certain properties of the individuals.23

But of course, allowing for definitional shorthand that allows the re-writing of (@) as (@') and as (@"), the latter two non-relational predications ARE identical with the relational predication (@). The relation simply is identical to certain non-relational predications of the relata. In fact, the monadistic account agrees with this. What is crucial about Bradley’s argument is that where the relation is genuine, the relation makes a difference to the being of the relata in this way: if one of the relata ceased to exist the being of the other would change, in the sense that something previously predicated of it could no longer be predicated of it. That is, if one of the relata ceased to exist, then one of the ways in which the other relatum is, one aspect of its being, would cease to be a way in which that thing is, cease to be an aspect of its being. Bradley is not here attempting to turn relations into properties; to the contrary, he is arguing against the monadistic attempt to do that. He is simply insisting that, where the relation is genuine, the being of one relatum, that is, the way in which the

204 relatum is or may be said to be, is not separable from the being of the other relatum, the way in which that other relatum is or may be said to be. John Stuart Mill argued that “The only difference between relative names and any others consists in their being given in pairs; and the reason of their being given in pairs is not the existence between two things, of a mystical bond called a Relation, and supposed to have a kind of abstract reality” (“Notes,” II, pp. 7-8). In asserting this, Mill is simply re-stating the monadistic account of relations. But he also holds that at least some relations are presented to us in experience: A rudimentary conception [of extension, that is, of structure] must be allowed, for it is evident that even without moving the eye we are capable of having two successive sensations of colour at once, and that the boundary which separates the colours must give some specific affection of sight, otherwise we should have no discriminative impressions capable afterwards of becoming, by association, representative of cognitions of lines and figures which we owe to the factual and muscular sense.24

Now, the relation is experienced, so, by PA, we must admit it into our ontology. But the monadistic account of relations will require that we treat it, not as tying the two relata into a unity, but as somehow a further thing alongside the two relata. But this simply won’t do: we have lost, according to Bradley, the essential feature of a relation, namely, that where there is a relation things form unified wholes, not parts lying as it were side by side. Let us abstain from making the relation an attribute of the related, and let us make it more or less independent. ‘There is a relation C, in which A and B stand; and it appears with both of them.’ But here again we have made no progress. The relation C has been admitted different from A and B, and no longer predicated of them. Something, however, seems to be said of this relation C, and, again, of A and B. But this something is not to be the ascription of one to the other. If so, it would appear to be another relation, D, in which C, on one side, and, on the other side, A and B, stand (A & R, pp. 17-18).

The infinite regress is clear. So is its vicious nature. One aims to capture the relational fact that is presented to one, but every time one does that the monadistic analysis eliminates the relational feature. Thus, even where the empiricists are prepared on the basis of their PA to admit relations into their ontology, the monadistic account immediately eliminates those relations.25

205 What Bradley is insisting upon is that where the relations are genuine the relata are not independent in way in which they must be if the monadistic account is correct. Thus, if (@) is genuinely relational then a and b are not independent as required by the monadistic account of relations. Nor will it do for the empiricist to suggest that she does insist upon there being a connection between the relata, save that it is subjective rather than objective. What is needed is an objective relation. As Bradley puts this point, ... to understand a complex AB, I must begin with A or B. And beginning, say, with A, if I then find B, I have either lost A or I have got beside A something else, and in neither case have I understood. For my intellect cannot simply unite a diversity, nor has it in itself any form or way of togetherness, and you gain nothing if beside A and B you offer me their conjunction in fact (A & R, p. 509).

So much the worse, then, for the monadistic account of relations: since the empiricist account of relations makes it unintelligible how any two things can be genuinely related, that position is just wrong. (b) Bradley’s Alternative What Bradley proposes is an alternative account of relations which can, unlike the monadistic, allow that there are relations that are genuine in the sense that their relata are not independent, or, equivalently, in the sense that the being of one relatum is not separable from the being of the other relatum. If it [a relation] is to be real, it must be so at the expense of the terms, or, at least, must be something which appears in them or to which they belong. A relation between A and B implies really a substantial foundation within them. This foundation, if we say that A is like to B, is the identity X which holds these differences together. And so with space and time – everywhere there must be a whole embracing what is related, or there would be no differences and no relation. It seems as if a reality possessed differences A and B, incompatible with one another and also with itself. And so in order, without contradiction, to retain its various properties, this whole consents to wear the form of relations between them (A & R, p. 18).

Bradley concludes against the monadistic account that “there must be a whole embracing what is related.” From this he infers his own account of

206 the nature of this whole. The relata A and B are different things within a whole (A, B). This whole then “consents to wear the form of a relation”; thus, if A and B stand in the relation R, then the correct representation of this fact consists in attributing a property corresponding to R, say r, to (A, B). Thus, according to Bradley’s account, the correct way to represent the fact reported by (@) a is R to b is given by (*) (a, b) is r or, r(a, b) The whole (a, b) is itself a particular thing,26 of which the two terms a and b are but aspects, and where the arrangement r characterizes the whole. But this whole consists of the relata as parts. Thus, the relation holds of the relata, not separately as in the monadistic account, but jointly:27 “where the whole, relaxing its unity, takes the form of an arrangement, there is coexistence with concord” (A & R, p. 19). Bradley applies this account to all relations. He mentions something being like something else, which is to say, he applies it to resemblance. He applies it to space and time. He applies it to quality orders (A & R, p. 19), and to causation (A & R, p. 46ff). He applies it to contrariety among qualities, so that the general facts of the sort (!) (x)[Fx e ~Gx] which record such contrariety become relational facts among qualities (A & R, p. 19). He applies the account to all predication (A & R, pp. 16-17). At each stage newer, and more embracing unities appear to support the relations that structure the world. In the end all the apparently separable particulars, and all the apparently separable qualities or properties of things that we are aware of in ordinary experience disappear into one allembracing unity. The ultimately real thing is one substantiality (A & R, p. 124-6). This is the Absolute (A & R, p. 151). Thus, Russell’s characterization of Bradley’s account of relations as “monistic” is quite apt. Not only do relations characterize unities, but in the end it requires all reality to be a unity of seamlessly inseparable parts. We should also note that Bradley’s account of relations introduces particulars that are not given in ordinary experience. In experiencing, say, a being to the left of b, we experience two particular things, a and b. We also experience the unity of a being to the left of b, the sort of unity represented by (@). But Bradley asserts that there is a further entity, a further

207 particular thing, present in the situation. That is the whole (a, b) that is the subject of the arrangement. This further whole is not given in ordinary experience. It is rather, something discovered by the faculty that Bradley calls “thought.” The object of thought is not something given in sense experience: “That [the object of thought] is not mere sense-experience should be a commonplace.”28 Rather, “judgement, on our view, transcends and must transcend that immediate unity of feeling upon which it cannot cease to depend.”29 Thought “grows from, and still it consists in, processes not dependent on itself. And the result may be summed up thus; certainly all relations are ideal, and as certainly not all relations are the product of thinking” (A & R, p. 426). And since relations are in effect not known by sense experience our knowledge of them is a priori; and since they are all ideal, transcending the entities of sense experience, the hold necessarily. In short, Bradley’s monistic account of relations requires the introduction of transcendent entities and the creation of a species of cognition through which we are supposed to be aware of, or cognize these entities. And since these entities are necessary and are known a priori, they in effect re-introduce the necessary connections against which Hume and the empiricists argued. But Bradley’s case for necessary connections goes beyond dogmatism: it is based on his account of relations which responds to real difficulties in the monadistic account that was accepted by the empiricists. (III) Russell’s Reply to Bradley Bradley’s case against empiricism rests on the monistic account of relations; remove that, and the case disappears. The monistic account of relations is offered as a solution to certain problems about relations that are insoluble upon the monadistic account of relations that the associationists had adopted. Russell accepted Bradley’s criticism of the monadistic account of relations. But he rejected Bradley’s own monistic account. Rather, he proposed a still different, third account of relations. This third account of relations is not subject to the criticisms that Bradley advanced against the monadistic account; it thus fulfils the same needs that the monistic account aims to fulfil. Yet it also perfectly compatible with the associationist account of thought; adopting it, one has no need to abandon associationism. Nor is one required to abandon empiricism and introduce

208 transcendent entities; in other words, it is also compatible with the empiricist’s PA. Russell’s account of relations does accept that one must abandon the monadistic account of relations that earlier empiricists had taken over from the tradition, but it also shows that that account is not essential to empiricism. Conversely, it shows that one can acknowledge the reality of genuine relations while rejecting Bradley’s claim that to do so requires one to accept the monistic account of relations and the consequent abandoning of empiricism. In effect, Bradley’s critique of empiricism depends upon two things. The first is the need to account for genuinely relational facts. The second is the assumption that the only way to satisfy the first need is by accepting the monistic account of relations. Russell accepts the first, while rejecting the second. He rejects the second by finding a third alternative that Bradley has not considered. (a) Russell’s Account of Relations Russell’s ontology of relations has become the commonplace. It does not therefore require much development. Consider once again the relational statement that (@) a is R to b or, in symbols, R(a,b) Russell rejects both the monadistic account in which (@) is to be analyzed into (#) a is r1 and b is r2 or, in symbols, r1(a) & r2(b) and also the monistic account in which (@) is to be analyzed into (*) (a, b) is r or, r(a, b) where the whole (a, b) is taken itself to be a particular thing. Both the monadistic and the monistic accounts of relations assume that predication always involves only one term. This is the “common opinión ... that all propositions, ultimately, consist of a subject and a predicate” (Russell, Principles of Math., p. 221). Russell rejects this common assumption. Russell’s account of relations takes the grammatical form of (@) to perspicuously represent its logical form. The objective fact represented by

209 (@) does not dissolve into a pair of facts about individuals – (#) – as on the monadistic account. Rather, as on the monistic account, a and b are located in a genuine unity such that, if one of a or b were not to exist, that unity could not exist. But on the other hand, this unity is not a whole of which the relation is predicated, as on the monistic account. Rather, the relation is predicated of the terms jointly. It is a and b being related that is the unified whole, rather than a and b being constituted into a whole of which the relation is then predicated. (b) Russell’s Criticisms of Bradley Russell offers three sorts of objection to the monistic account of relations. In the first place, there are certain relations for which the theory can provide no satisfactory account; it therefore fails as an ontology of relations. (Russell directs the same criticism at the monadistic view of relations.) Secondly, the account leads to a radical scepticism in which in the end nothing can be known. And thirdly, it introduces entities that violate the empiricist’s Principle of Acquaintance. First: On the monistic view, relational states of affairs consist of a relational property r being predicated of a complex individual whole (a, b). The relation of a to the whole (a, b) is the same as the relation of b to that whole. That is, the role of a in that whole is symmetrical with the role of b. Thus, (*) r(a, b) represents indifferently both the fact that (@) a is R to b and its converse, the fact that b is R to a (@c) Where R is a symmetrical relation, one for which we have (x)(y)[R(x,y) e R(y,x)] then we have no problem: if (@) obtains so does (@c), and (*) can represent the two indifferently. But the same does not hold for asymmetrical relations (Principles, p. 221). Where R is asymmetrical, we have (+) (x)(y)[R(x,y) e ~R(y,x)] In this case (@) obtains while (@c) does not. In the case of an asymmetrical relation, there is a difference – an ontological difference – between a relational fact and its converse that is not captured in any

210 account, like the monistic, that requires both facts to be represented indifferently by the same notation. But the monistic account was introduced to solve the problem of relations. Since it cannot do that, it must be rejected.30 To this objection, Bradley replies (“Relations,” p. 672), that the incompatibility between a relation and its converse, that is, the law (+), if it is to be more than a matter of chance, must be the expression of a real relation that obtains between a relational fact constituted by R and the converse of that fact. It will, therefore, be a necessary fact about R, part of the meaning of R, that its obtaining excludes its converse obtaining. Now, this may well be the case, given the monistic account of relations. It does not, however, adequately reply to Russell. For, even if it is somehow a necessary truth that the obtaining of a relation excludes the obtaining of its converse, that is, if it is somehow a necessary truth that a relation is asymmetrical, it still does not follow that one has provided an account of relations that adequately captures the difference between a relation and its converse. The point remains that (*) represents both (@) and (@c) indifferently, this because a and b occur symmetrically in (a, b). There is therefore nothing that accounts for the difference between (@) and (@c). Bradley looks at the contrariety between (@) and (@c) rather than the difference between them that is presupposed by the contrariety.31 Second: As we saw, upon the monistic account of relations, all relational statements are ultimately about a single individual whole, the Absolute, that contains all reality within it. Hence, in order to know any truth, one must grasp this individual totality, the Absolute itself. But we are finite; it is impossible for us to fully grasp this one subject of predication. It follows that we, that is, we as we in fact are in our ordinary experience, finite beings, cannot know any relational truth. In short, a radical scepticism about relational facts results from the monistic account of relations, even though that theory was introduced precisely to account for certain truths about relations that the empiricists had neglected. It is worse than this, however. Bradley applies his account of relations to predication itself. The judgement that c is H, i.e. that c is qualified by H, becomes a judgment in which the property of “qualifying” is now predicated of the whole (c, H). But this leads to an infinite regress, one which is vicious in the sense that in order to make one judgment one must make an infinite number of further judgments. To avoid such a

211 regress, one must insist in the end that one arrives at a whole in which the difference between it and its predicates disappears. Bradley accepts this conclusion: “... to reach a mode of apprehension, which is quite identical with reality, surely predicate and subject, and subject and object, and in short the whole relational form, must be merged” (A & R, pp. 151-2). But any judgment, any proposition, involves a distinction between subject(s) and predicate. So, on Bradley’s view, no judgment about the Absolute can ever be wholly true. In fact, any judgment must be self-contradictory. For, on the one hand, it purports to be true. On the other hand, it purports to ultimately be about reality, that is, the Absolute since it is the only reality, and this reality is such that it must be false. Any judgment must therefore claim of itself to be both true and false. This, moreover, must be equally true of the judgment that no statement about the Absolute is ever fully true: this statement of Bradley’s view is itself contradictory upon that view. This, surely, is to condemn it. As Russell puts it, ... we find monists driven to the view that the only true whole, the Absolute, has no parts at all, and that no propositions in regard to it or anything else are quite true – a view which, in the mere statement, unavoidably contradicts itself. And surely an opinion which holds all propositions to be in the end selfcontradictory is sufficiently condemned by the fact that, if it be accepted, it also must be self-contradictory (Principles of Math., p. 226).32

Oddly enough, Bradley himself accepts this conclusion. In primitive feeling, instinct, we are aware of the world and of ourselves in it as a simple indivisible unity. But we are led to develop arguments to lead to this conclusion. Unfortunately, since these arguments involve judgements, they must all be unsound. Hence, as Bradley says, “Metaphysics is the finding of bad reasons for what we believe upon instinct...” (A & R, p. x). In all his sturdy Victorian manliness, Bradley accepts all the consequences of his account of relations, however absurd they may be. For the rest of us, these consequences condemn it, if not as silly, then as philosophically unacceptable: any view that, in attempting to solve philosophical problems, leads to a radical scepticism must be rejected. Third: We noted above that Hume discovered a radical scepticism implicit in the rationalist doctrine that there are necessary connections. This radical scepticism derived from the fact that the rationalists introduced transcendent entities into their ontology. We now see that Bradley’s

212 monistic account of relations similarly introduces transcendent entities, and similarly leads to a radical scepticism. It is the introduction of transcendent entities that generates the scepticism. One can eliminate the scepticism by eliminating the transcendent entities upon which it is parasitic. Hume achieved this elimination by appeal to the empiricist's PA. Russell does the same. If Bradley is correct, then one can identify a quality of a thing only if one goes beyond that quality and relates it to other qualities. For, qualities of things stand in the relation of contrariety one to another, that is, for qualities we have laws of the sort (!) (x)[Fx e ~Gx] If this regularity is to be more than a matter of chance, it must be the expression of a real relation, or, what amounts to the same, a necessary connection, that obtains between the qualities F and G. Thus, in order to identify the quality of a thing as F, one must also refer as a matter of necessity to the property G, to which it is necessarily tied. Russell argues to the contrary that the identification of a property is a matter of that property alone; identifying it as what it is does not require that we also identify what it is not. Russell (Principles of Math., p. 448) refers to G. E. Moore’s essay on “The Nature of Judgment.”33 Moore makes the point that we have seen Russell make, that Bradley’s view of relations requires us to ascend to the Absolute before we can make any true judgment at all34: it requires “the completion of an infinite number of psychological judgments before any judgment can be made at all” (Moore, p. 178). The problem is Bradley’s theory of relations. This account ...presupposes that I may have two ideas, that have a part of their content in common; but ... at the same time compel[s] us to describe this common part of content as part of the content of some third idea (Moore, p. 178).

That is, Bradley's account of relations requires the introduction of a third particular, the Whole, over and above the two particulars that stand in relation to each other. This relation is such that the one quality cannot be identified independently of its necessary connections to other qualities. But in fact qualities can be identified as themselves without reference to the relations in which they stand to other qualities and other things. Moore makes the point with specific reference to the relation between qualities and the mind that knows them:

213 It is indifferent to their nature whether anybody thinks them or not. They are incapable of change; and the relation in to which they enter with the knowing subject implies no action or reaction (Moore, p. 179).

But Moore implies it in full generality, and Russell says just that: To say that two terms which are different if they were not related, is to say something perfectly barren; for if they were different, they would be other, and it would not be the terms in question, but a different pair, that would be unrelated. The notion that a term can be modified arises from neglect to observe the eternal self-identity of all terms and all logical concepts, which alone form the constituents of propositions [here Russell refers to Moore’s essay]. What is called modification consists merely in having at one time, but not at another, some specific relation to some specific term; but the term which sometimes has and sometimes has not the relation in question must be unchanged, otherwise it would not be that term which has ceased to have the relation (Principles, p. 449).

Note that here Russell is allowing Bradley’s point against the monadistic account of relations. On the latter, the predications of one term of a relation would not change if the other relatum ceased to exist. Russell accepts this. What he is denying is the implication of the monistic account of relations that there is something about properties or qualities as presented that requires us when we are identifying them to refer as a matter of necessity to other properties, those to which they are necessarily tied. Russell is holding that properties are presented to us as logically selfcontained rather than as necessarily tied to one another; he concludes that there are no such necessary connections. But such connections are required by the monistic account of relations. The falsity of the latter view follows. Russell’s rejection of Bradley’s monistic account of relations in the basis of an appeal to the empiricist’s Principle of Acquaintance is clear. Russell’s general position against Bradley thus accepts the latter’s criticism of the monadistic account of relations that had been adopted by the earlier empiricists. Yet Russell also attacked Bradley’s own positive account of relations on the basis of an appeal to the empiricist's PA. Russell was thus, under the impact of Bradley, rejecting part of traditional empiricism, while at the same time appealing to the central empiricist principle of Locke, Hume and the Mills, that is, PA, to reject Bradley’s monistic relations. How much of empiricism survived?

214 (c) Empiricism and Necessary Connections Once Again Bradley’s monistic account of relations re-introduced the necessary connections against which Hume had argued. Hume’s argument against those connections had two strands. The second of these was based on an appeal to the monadistic account of relations. Bradley made a compelling case against the latter, a case that Russell accepted. So this part of the argument against necessary connections is eliminated. What of the first strand? This argued that there is no necessary connection between cause and effect because it is not self-contradictory to affirm the existence of the effect while denying the existence of the cause. Among the entities that are given to us in our ordinary sensory experience we discover no connections such that one of these entities cannot exist unless another of these entities exists. In identifying the property that characterizes the effect we do not have to refer as a matter of necessity to the property that characterizes the cause, one to which it is necessarily tied; the properties are presented as logically self-contained rather than as necessarily tied to one another. We are, in other words, acquainted with no such entity as a necessary connection in ordinary experience, and there is therefore no necessary connection among the ideas that we use to describe whatever is given to us in experience. The appeal is to PA to establish that presented properties are logically self-contained. But we have seen that Russell adopts this very same appeal to PA to argue against Bradley’s necessary connections. So the first strand of the empiricist’s case against necessary connections stands. Bradley attacked the Humean account of causation and of laws. This case was based on the alleged need to adopt the monistic account of relations. But Russell showed that that argument was unsound by introducing his third alternative to the monadistic and monistic accounts. And he showed that in any case the monistic account was untenable, on precisely the same ground that Hume used to argue against necessary connections. So the empiricist account of causation and laws emerges saved by Russell from Bradley’s critique. Russell thus ensures not only that empiricism in its central tenets survives Bradley’s critique, but also that it emerges strengthened, with a new account of relations that is at once compatible with PA and meets the objections that Bradley had raised against the monadistic account that earlier empiricists had taken over uncritically from the previous tradition.

215

(IV) Bradley vs. Russell Bradley did reply to Russell critique of his monistic account of relations. He argues that relations understood as holding between their terms generate a vicious infinite regress. We have seen above that this is so if one adopts the monadistic view of relations. Russell argues that the same is not true of his account of relations, so that this reply of Bradley has no force. To be sure, if we have the relational fact R(a, b) then one can define the propositional function R(r, x, y) and assert that R(R, a, b) Clearly, this process can be repeated indefinitely. We thus have an infinite regress. But Russell notes that it is an infinite regress of implied propositions, and is no more vicious than the regress p, ~~p, ~~~~p, ... of implied propositions following upon the law of double negation. “The endless regress is undeniable, if relational propositions are taken to be ultimate, but it is very doubtful whether it forms any logical difficulty” (Principles of Math., p. 99; cf. p. 51). Against this, Bradley was simply to restate his point that a relation taken joining two terms can never achieve a genuine unity.35 In an experienced relational whole, “You can have the terms, without which you cannot have the relation, only so far as (in order to have the relation) you abstract from the ... unity, on which (to keep your relation, which requires some unity) you are forced vitally to depend” (“Relations,” p. 637). But this is merely a re-assertion of his own monistic account of relations, rather than a reply to Russell’s point that the empiricist can simply take relational unities to be “ultimate.” This ultimacy of relational facts means in part that such wholes cannot be analyzed as Bradley requires if he is to get his regress off the ground. If we have the relational fact that (a) R(a, b) then this can be analyzed into three parts: (b) a, b, R

216 But the fact (a) cannot be re-constructed from the list of parts (b). For that list is compatible with another, quite different fact: (c) R(b, a) In this sense the crucial unity of the fact that is created by the relation always escapes analysis. As Russell once put it, “I do not admit that, in any strict sense, unities are incapable of analysis; on the contrary, I hold that they are the only objects that can be analysed. What I admit is that no enumeration of their constituents will reconstitute them, since any such enumeration gives a plurality, not a unity.”36 Bradley holds that if we are to discern three entities in (a) as Russell does, then we need yet another entity to achieve the unity of the fact. “The relation ... has been admitted different from [its terms], and is no longer predicated of them. Something, however, seems to be said of this relation ..., and said, again of [its terms] .... it would seem to be another relation ... in which [the first relation], on the one side, and, on the other side, [its terms] stand” (A & R, p. 18). What Russell does, on his account of relations, is insist that the relation is indeed to be admitted as different from its terms and that it is predicated of them. What is called analysis consists in the discovery of the constituents of a complex. A complex differs from the mere aggregate of its constituents, since it is one, not many, and the relation which is one of its constituents enters into it as an actually relating relation, and not merely as one member of an aggregate. I confess I am at a loss to see how this is inconsistent with [his (Russell’s) own] account of relations, and I suspect that the meaning which I attach to the word “external” is different from Mr. Bradley’s meaning; in fact he seems to mean by an “external” relation a relation which does not relate.37

As Russell’s last remark indicates, Bradley’s argument works against the monadistic account but not against Russell’s which denies one of its crucial premises. But in denying that crucial premise Russell’s account requires that one distinguish the relational fact from the set of its constituents. In this sense, any relational fact is more than the sum of its constituents. Since the fact cannot be reconstructed from the analysis, Bradley rejects the claim that one can distinguish by analysis three constituents in the relational fact (a). “Since what I start with in fact is this, and what analysis leaves to me instead is that – I therefore can not but reject, at least in part, the result of análisis.”38 Elsewhere he characterizes Russell’s position in this way:

217 On the one side I am led to think that [Russell] defends a strict pluralism, for which nothing is admissible beyond simple terms and external relations. On the other side, Mr. Russell seems to assert emphatically, and to use throughout, ideas which such a pluralism surely must repudiate. He throughout stands upon unities which are complex and which cannot be analysed into terms and relations. These two positions to my mind are irreconcilable, since the second, as I understand it, contradicts the first flatly.39

But this is simply to deny Russell’s account of relations; it is not to argue against it. In fact this is all that Bradley ever does in reply to Russell. Thus, once again, Bradley, writing in response to Russell’s insistence that there is a distinction between a fact and the set or aggregate of its constituents, Is there anything ... in a unity besides its ‘constituents’, i.e. the terms and the relation, and, if there is anything more, in which does this ‘more’ consist? Mr. Russell tells us that we have not got merely an enumeration or merely an aggregate. Even with merely so much I should still have to ask now even so much is possible. But, since we seem to have something beyond either, the puzzle grows worse.40

Bradley is here doing nothing more than assert that that which yields the unity of the fact must appear in the analysis. If it does not, then one has not captured the unity. He concludes that any attempt like Russell’s to distinguish among the constituents of a relational fact will not do. Rather, one needs an account of relational unities such that one does not lose the unity. What this means is that one in effect achieves unity while losing the diversity of the parts. This is, of course, precisely what the monistic account of relations ultimately implies. “For me,” Bradley asserts, “immediate experience gives us a unity and unities of one and many, which unities are not completely analysable or intelligible, and which unities are self-contradictory unless you take them as subject to an unknown condition. Such a form of unity seems to me to be in principle the refutation of pluralism...”.41 But all this is simply to beg the question against Russell. Bradley is denying precisely what is central to Russell’s account of relations, that the unity of a relational fact is something ultimate and does not appear among the constituents of the fact. But he merely denies. When it came to Bradley’s monistic account of relations, Russell provided an extended argument against it. In contrast, when it comes to Russell’s account of relations, Bradley is content simple to deny,

218 recognizing no need to make a case. Naturally, we are not convinced. Russell’s account of relations emerges unscathed.42

(V) Conclusion For many years Bradley was taken to be a prime example of a metaphysician gone mad, and Appearance and Reality was read for no other reason than that of finding “metaphysical” statements that could be held up for ridicule as metaphysical “nonsense.” Thus, A. J. Ayer, in Language, Truth and Logic, uses a quotation “taken at random” from Appearance and Reality to show the way in which metaphysicians are inclined to make statements which “have no literal significance, even for themselves.”43 The assumption was that the whole business had been radically mistaken. That philosophers could arrive at that opinion was due in no small part to Russell’s critique of Bradley. But if the opinion was in a way justified by that critique, it was nonetheless radically mistaken. That Bradley required a sustained critique was itself the measure of the philosophical significance of his views. Moreover, a good part of Bradley’s critique of earlier empiricism survived in Russell. Russell recognized the force of Bradley’s critique of the monadistic account of relations that earlier empiricists had adopted. Russell recognized that if empiricism was to be defended it required an account of relations that was at once compatible with PA and met the objections of Bradley to the doctrine accepted by Hume and the Mills. It was Bradley’s prodding and Bradley’s challenge that moved Russell to develop a doctrine of relations and a version of empiricism that was free from the problems of its earlier versions. Bradley, in short, has had a significant impact on British empiricism. If Russell re-established empiricism after a long bout of idealism, this empiricism was no longer merely that of Hume, but something more defensible. That it was more defensible was due in no small part to Bradley. If empiricists have come to be able to look back from a defensible empiricism to Bradley as a metaphysician gone mad, it is only because Bradley was not that but rather a searching and careful philosopher whose effective criticisms of empiricism and defense of rationalism moved

219 Russell to create an empiricism that could in fact be reckoned as defensible. Even where we disagree with Bradley, we must, like Russell, give him his due.

220 Endnotes to Study Five 1 F. H. Bradley, “Relations,” in his Collected Essays, 2 vols. (Oxford: Oxford University Press, 1935), vol. 2, pp. 628-76. 2 “Relations,” p. 656f. 3 D. F. Pears, Bertrand Russell and the British Tradition in Philosophy (London: Collins, 1967), p. 11. 4 Ibid., p. 165ff. 5 This is not the only point where Russell can be seen as responding to Bradley; see the essay by N. Griffen, “F. H. Bradley’s Contribution to the Development of Logic,” in J. Bradley, ed., Philosophy after F. H. Bradley (Bristol: Thoemmes, 1996), pp. 195230. 6 For the latter, see P. Ferreira, “F. H. Bradley’s Attack on Associationism,” in J. Bradley, ed., Philosophy after F. H. Bradley, pp.283-306; and also F. Wilson, “Bradley’s Critique of Associationism,” Bradley Studies, 4 (1998), pp. 5-60. 7 See J. Weinberg, “The Concept of Relation,” in his Abstraction, Relation, and Induction (Madison, WI: University of Wisconsin Press, 1965). 8 B. Russell, Principles of Mathematics, Second Edition (London: Allen and Unwin, 1937) [first edition 1903], p. 221f. 9 These two sides of relational statements, the objective and the subjective, are reflected in the two definitions of ‘cause’ that Hume offers, Treatise (ed. L. A. SelbyBigge, rev. P. H. Nidditch (Oxford: Oxford University Press, 1978), p. 172; see F. Wilson, “Hume’s Theory of Mental Activity,” in D. F. Norton et al., McGill Hume Studies (San Diego: Austin Hill Press, 1979). See also F. Wilson, “Bergmann’s Hidden Aristotelianism,” this volume, below. 10 James Mill, Analysis of the Phaenomena of the Human Mind (first edition, London: 1825; reprinted ed. J. S. Mill, with notes by J. S. Mill, A. Bain, et al., [London: Longmans, Green, Reader and Dyer, 1869]); page references are to the latter). The passage quoted is from vol. II, p. 7; see also vol. I., p. 185. 11 John Stuart Mill. “Notes,” to James Mill, Analysis of the Phaenomena of the Human Mind, 1869, vol. II, pp. 9-10. 12 Cf. Hume, Treatise, p. 14.

221 13 Treatise, Bk. I, Part iii; the conclusion occurs on p. 169ff. See Wilson, “Hume’s Theory of Mental Activity,” and also Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1997) For an extended defence of the Humean account of causation, see F. Wilson, Laws and Other Worlds (Dordrecht, The Netherlands: 1989). 14 James Mill, Analysis, vol. I, pp. 362-74, 389-91; John Stuart Mill’s “Notes” to his father’s Analysis, vol. I, p. 412, 437-8; and John Stuart Mill, System of Logic, vol. 7 of The Collected Works of John Stuart Mill, ed. J. Robson (Toronto: University of Toronto Press, 1974), p. 338ff. 15 Treatise, pp. 14-15. 16 .James Mill, Analysis, vol. II, p. 7, vol. I, p. 185; J. S. Mill, “Notes” to his father’s Analysis, vol. II, pp. 7-8. 17 For a discussion of this principle, see the essays, “Acquaintance, Ontology and Knowledge,” this volume, above; “Effability, Ontology and Method,” this volume, below; and “Universals, Particulars, Tropes and Blobs: The Role of a Principle of Acquaintance in Ontology,” this volume, below; and also F. Wilson, “The Lockean Revolution in the Theory of Science,” in S. Tweyman and G. Moyal, eds., Early Modern Philosophy: Epistemology, Metaphysics and Politics (New York: Caravan Press, 1986), pp. 65-97. 18 Hume thus spots a problem that is general to any non-empiricist metaphysics that attempts to solve philosophical problems by introducing transcendent entities. Cf. the essay, “Acquaintance, Ontology, and Knowledge,” this volume, above. 19 This was the foundation of other of his criticisms of empiricism as he knew it, e.g. the associationism of Hume, the Mills, and Spencer. But we cannot go into these other issues here. See, however, F. Wilson, “Bradley’s Critique of Associationism.” See also F. Wilson, “The Significance for Psychology of Bradley’s Humean View of the Self,” Bradley Studies, 5 (1999), pp. 5-44. 20 Principles of Mathematics, p. 222. 21 F. H. Bradley, Appearance and Reality, second edition (Oxford: Oxford University Press, 1897), p. 125; hereafter cited as “A&R”. 22 Pears, Bertrand Russell and the British Tradition in Philosophy, p. 165. 23 Ibid.; his italics. 24 J. S. Mill, Examination of Sir William Hamilton’s Philosophy, vol. 9 of The Collected Works of John Stuart Mill, ed. J. Robson (Toronto: University of Toronto Press, 1979), p. 230.

222 25 Standard discussions of Bradley’s case generally fail to note that it is perfectly sound against the monadistic account; cf. B. Blanshard, “Bradley on Relations,” in G. Manser and G. Stock, eds., The Philosophy of F. H. Bradley (Oxford: Oxford University Press, 1984), p. 215f. 26 Cf. Bradley, “Relation,” pp. 635-6. 27 Ibid., p. 636. 28 F. H. Bradley, “Association and Thought,” in his Collected Essays, 2 vols. (Oxford: Oxford University Press, 19335), vol. 1, p. 208. 29 F. H. Bradley, Essays on Truth and Reality (Oxford: Oxford University Press, 1914), p. 231. 30 For a more recent discussion of the problem of order, see E. B. Allaire, “Relations: Recreational Remarks,” Philosophical Studies, 34 (1978), pp. 81-89; and also the essay, below, “Burgersdijck, Coleridge, Bradley, Russell, Bergmann, Hochberg: Six Philosophers on the Ontology of Relations.” 31 In fact the difference is there also in the case of symmetrical relations, and is equally overlooked in such cases also. But in those cases it is open to the monists to dismiss the difference as ontologically irrelevant. In the case of asymmetrical relations, such dismissal is not possible. 32 Russell elaborates this case in detail in his essay on “The Monistic theory of Truth,” in his Philosophical Essays, rev. ed. (London: Allen and Unwin, 1966 [originally pub. 1910]), pp. 131-146 Russell was not the only one to advance this sort of criticism; see the discussion of C. A. Campbell, in Lorne Maclachlan, “F. H. Bradley and C. A. Campbell,” in J. Bradley, ed., Philosophy after F. H. Bradley, pp. 73-90. 33 G. E. Moore, “The Nature of Judgment,” Mind, n.s. 8 (1899), pp. 176-93. 34 Moore does not add the further point that Russell makes, that even then we cannot make a true judgment. 35 For more recent discussions of Bradley's regress, see E. B. Allaire, “Wolterstorff and Bradley on Ontology,” Journal of Philosophy, 89 (1968); Kenneth Barber, “A Note on the Paradox of Analysis,” Philosophical Studies, 19 (1968) ; and Richard Gull, “Bradley’s Argument against Relations,” The New Scholasticism, 45 (1971). 36 B. Russell, “Some Explanations in Reply to Mr. Bradley,” Mind, n.s. 19 (1910), p. 373.

223 37 Russell, “Some Explanations in Reply to Mr. Bradley,” p. 374. 38 F. H. Bradley, The Principles of Logic, 2nd ed., rev., with commentary and terminal essays, 2 vols. (London : Oxford University Press, 1922), vol. 2, p. 693. 39 F. H. Bradley, “On Appearance, Error, and Contradiction,” Essays on Truth and Reality, p. 281. 40 Bradley, Essays on Truth and Reality, pp. 288-9. 41 Bradley, “On Appearance, Error, and Contradiction,” p. 281. 42 For another perspective on the disagreement between Bradley and Russell on the possibility of analysis, see P. Dwyer, “Bradley, Russell, and Analysis,” in J. Bradley, ed., Philosophy after F. H. Bradley, pp. 332-348. 43 A. J. Ayer, Language, Truth and Logic, rev. ed. (New York: Dover, 1946), p. 36.

Six Moore’s Refutation of Idealism* G. E. Moore’s essay on the “Refutation of Idealism”1 is apparently wellknown. Yet one can in fact find very few analyses of its argument. Even where these exist, they remain sketchy. Thus, the commentary by D. Lewis,2 admirable as it is, deals only with what Moore designates as the second part of his argument, by-passing the first as if it were unimportant.3 Yet, if it can simply be ignored, in the way in which Lewis ignores it, then why did Moore include it? Moore’s argument is clearly important, if only symbolically: though others were making some of the same points as was Moore, it is Moore’s essay that has come to mark the point at which idealism began its retreat after its apparent triumph over empiricism in the final quarter of the nineteenth century. But if it is so important, then precisely what was that argument? The general thrust of Moore’s argument is of course to refute the principle that Berkeley was the first to defend, that esse is percipi, to use the Berkeleyan formula, which Moore in fact quotes, but which should more exactly be expressed as the principle that esse MUST BE percipi. It is clear that one can defend the principle that what is perceived exists. It is clear that one can also defend, at least on empiricist grounds, that what exists must be perceivable, in the sense of being related to the perceiver in such a way that, if the perceiver were appropriately situated, or if his or her senses were acute enough, then it would be perceived. This is the force of the following argument in Berkeley’s Three Dialogues4: ... you neither perceive matter objectively, as you do the inactive being or idea, nor know it as you do your self by a reflex act neither do you mediately apprehend it by similitude of the one or the other: nor yet collect it by reasoning from that which you know immediately (p. 232).

This much Berkeley seems clearly to establish at the beginning of the ‫٭‬

Originally appeared as “Moore’s Refutation of Idealism,” in P. Coates and D. Hutto, eds., Current Issues in Idealism (Bristol: Thoemmes Press, 1996), pp. 23-58.

226 Principles. What seems puzzling is the stronger conclusion that what exists is perceived and the still stronger conclusion that what exists must be perceived. As Berkeley puts it in the Principles,5 All those bodies which compose the mighty frame of the world, have not any subsistence without a mind, that their being is to be perceived or known; that consequently so long as they are not actually perceived by me, or do not exist in my mind or that of any other created spirit, they must either have no existence at all, or else subsist in the mind of some eternal spirit: it being perfectly unintelligible ... to attribute to any single part of them existence independent of a spirit (§ 6).

Now, it is clear that Berkeley conceives the relation between the mind and what is perceived as that of substance to property. He repeatedly makes this point: ... spirits and ideas. The former are active, indivisible substances: the latter are inert, fleeting or dependent beings, which subsist not by themselves, but are supported by, or exist in minds or spiritual substances (§ 89). ... if we had a new sense, it could only furnish us with new ideas or sensations, and then we should have the same reason against their existing in an unperceiving substance that has already been offered with relation to colour, and the like. Qualities ... are nothing else but sensations or ideas, which exist only in a mind perceiving them (§ 78).

In fact, this seems to be his basic argument: What is perceived is either a quality or a set of qualities; it is a metaphysical principle that there are no qualities apart from substances; substances must be active – that is what it is to be a substance – while to be a mind or spirit and to be active amount to the same; hence, what is perceived must be perceived. As Berkeley puts it in the Principles: ... the objects perceived by sense, are allowed to be nothing but combinations of those qualities, and consequently cannot subsist by themselves (§ 46).

Briefly put, the argument is simply: What we perceive are qualities; qualities cannot exist apart from substances; the only substances are minds; hence, what is perceived is dependent upon a perceiver, what is perceived must be perceived. Berkeley’s case, however, will not withstand scrutiny. The principle that qualities cannot exist apart from substances was intended to have an

227 anti-platonic force: qualities cannot exist apart from the entities of which they are predicated. Since, in the tradition, substances were the entities of which qualities are predicated, the metaphysical principle follows. But Berkeley has already analyzed ordinary things into collections of qualities, and has insisted that qualities are predicated of the collections of which they are parts: ... to me a die seems to be nothing distinct from those things which are termed its modes or accidents. And to say a die is hard, extended, and square, is not to attribute those qualities to a subject distinct from and supporting them, but only an explication of the word die (§ 49).

In fact, it can be argued that this is Berkeley’s major contribution to philosophy: no longer was philosophy to be wed to the dogma that an ordinary thing had to be a substance because the qualities predicated of ordinary things had to be predicated of substances. Since the principle upon which Berkeley relies to establish that esse is percipi has now lost its traditional rationale, it follows that Berkeley’s principle, too, has lost its rationale.6 Nonetheless, it is this principle, that qualities are inseparable from minds, that subsequent idealists were prepared to defend, and it is this principle that Moore proposes to refute. Moore does take up this Berkeleyan idealism, in which the properties known are dependent upon the mind as predicates are dependent upon particulars. He does this in the second part of his essay. Moore’s argument here consists in his pointing out that we in fact have two distinct relations, that between a predicate and its subject and that between the mind and what it knows. The latter has, since Brentano, often been referred to as the “intentional” connection. Moore’s argument, then, consists in pointing out that there is a difference between predication and intentionality. As Moore puts it, he is asserting ...two distinct propositions, (1) that blue is probably not part of the content of the sensation at all, and (2) that, even if it were, the sensation would nevertheless not be the sensation of blue, if blue had only this relation to it (“Refutation,” p. 26).

Now, Moore holds that blue is probably not part of the content of the sensation of blue at all because that would mean that blue would be predicated of the sensation; one would have in fact a blue awareness:

228 ...offence may be taken at the expression, but yet it expresses just what should be and is meant by saying that blue is, in this case, a content of consciousness or experience. Whether or not, when I have the sensation of blue, my consciousness or awareness is thus blue, my introspection does not enable me to decide with certainty: I only see no reason for thinking that it is (ib.).

However, whether my awareness is indeed blue or not is unimportant, ...for introspection does enable me to decide that something else is also true: namely that I am aware of blue, and by this I mean, that my awareness has to blue a quite different and distinct relation (ib.).

What is predicated of a subject is dependent upon that subject: it cannot exist apart from it. So at least, the tradition, as Berkeley well knew. Provided that there is no distinction between the relation of predication and the relation of knowing or intentionality, then it follows that what is known is dependent upon the knowing mind. However, if Moore is correct, then there is a distinction between the two, and so the fact that a thing is known or thought of, the fact that a thing stands in the intentional relation to mind, does not imply that it is dependent upon the mind for its existence, at least not in the traditional sense in which properties are dependent upon substances. Specifically, there is nothing about the intentional relation, unlike the relation of predication, that can be taken to imply that the perceived object will cease to exist if it were unperceived. ... we recognise that this awareness is and must be in all case of such a nature that its object, when we are aware of it, is precisely what it would be, if we were not aware...(p. 29).

The argument is as clear in its simplicity as it is convincing. Who, in fact, would disagree? Certainly, if one were to canvass philosophers today it is more than likely that many if not most would agree with Moore. To that extent Moore’s argument has succeeded in persuading the twentieth century that idealism is wrong.7 But in fact, there were those who disagreed. Thus, Berkeley does, for one, as well as the idealists who form the immediate background to Moore's essay. But before turning to the latter, who were Moore's main target, we had best complete our task of unravelling the structure of Moore’s essay. Given the structure of Moore’s essay, the second portion of the overall argument against idealism that we have just examined, the portion di-

229 rected specifically at Berkeleyan idealism, is clearly intended to presuppose the first part. That is why Lewis is therefore wrong to ignore this first part of Moore’s argument. For all that, it is not too hard to think why Lewis might have ignored this part. Here is how Moore summarized the case he is presenting in this portion of his argument: ‘Esse is percipi,’ we have seen, asserts of two terms, as distinct from one another as ‘green’ and ‘sweet’ that whatever has the one has the other: it asserts that ‘being’ and ‘being experienced’ are necessarily connected: that whatever is is also experienced. And this, I admit, cannot be directly refuted. But I believe it to be false; and I have asserted that anybody who saw that ‘esse’ and ‘percipi’ were as distinct as ‘green’ and ‘sweet’ would be no more ready to believe that whatever is green is also sweet (p. 16).

The conclusion of the first part of Moore’s argument would thus seem to be the simple one that, where things are distinct, they are distinct. It is a point that Moore’s argument against Berkeley presupposes. For, after all, that argument depends upon his insisting that the relation of intentionality is distinct from the relation of predication. Moreover, the second part of the argument also presupposes that the sensation – what is perceived, the colour blue – is distinct from the perceiving of it – the act of awareness: the latter is a sensation of blue but is distinct from the blue. So the conclusion of the first part of Moore’s argument is indeed necessary for the second part, but at the same time it is a point about which it hardly seems worth arguing. It is that, in fact, which likely explains why Lewis did not think it worthwhile to examine this section in detail. However, obvious as the point is, there were philosophers who were clearly prepared to deny it. Otherwise, why would Moore have included it? It is the philosophers who deny this who are his opponents. Of these opponents, the only one who is explicitly mentioned by Moore is A. E. Taylor, whose essay on “Mind and Nature”8 is cited as making the claim that esse is, and inseparably is, percipi. Moreover, he at least must be among those who hold that ‘esse’ and ‘percipi’ are not, contrary to all appearances, as distinct as ‘green’ and ‘sweet.’ Now, in point of fact, Taylor in this essay does not come out and assert that esse and percipi are not distinct. What he says is that they are inseparable, and Moore takes this to amount to the same. Whether it is as Moore takes it to be is another issue. However, Taylor does offer an argument why they must be taken to be inseparable. It may be that, if we examine this, then we can find out whether he really is committed to the view that Moore clearly attributes to

230 him. But when we examine Taylor’s argument, it is hard to see, at first anyway, why it should have seemed to him or to anybody to be persuasive. Taylor undertakes to show that, whatever be the fact that one is prepared to say is real or actual, “what makes it real can be nothing but its presence as an inseparable aspect of a sentient experience” (p. 58). When the fact is chosen, call it “A”, then, says Taylor, ...we shall invite you to think, as you always can, of a corresponding A which is not real but merely imaginary, and then to say what it is that makes the difference between the real and the imaginary or unreal A. If you will try the experiment, you will always find, as Kant proved in the historical case of the hundred dollars, that the difference does not lie in the addition of a new predicate to those by which the imaginary A is characterized, but always in the actual presence of the real A to a sentient experience, its entrance into some immediately apprehended whole (pp. 58-59).

Taylor seems to have liked this argument, because he again presented it his later Elements of Metaphysics9 (p. 24f). The argument is there expanded, but it is essentially the same. The difficulty is that it hardly seems to establish the inseparability of esse and percipi. To be sure, it is good empiricist sense that something, A, if it is to be affirmed as real, must stand in some appropriate relation to actual experience. Specifically, it must have effects leading up to some actual experience such that the chain of these effects can be traced back in a way that justifies our affirming that A is indeed real or actual. For something to be affirmable as real is for it to be part of a structure or set of relations that connect it appropriately to actual experience. And for something to be real tout court is for it to be part of a structure or set of relations such that, if a subject of experience were appropriately situated, those relations would connect it appropriately to actual experience. This is empiricism, however, and not idealism: the inseparability of the experienced from the experiencing of it seems scarcely to follow. Things become even more mysterious if we turn to what Taylor says about the concept of reality. In fact, we are told two things. First, “what is real is never self-contradictory” is said to be the universal criterion of reality (p. 19). The problem here is, as many have suggested, that there are many things that are not self-contradictory, or, what is the same, are selfconsistent, and yet are not real, e.g., unicorns and hyperbolic spaces. Second, we are also told that “the ‘unreal’ [is] that with which we have not, for any human purpose, to reckon” (p. 51). “Stubborn” facts are precisely

231 those that we must recognize, and take account of, when we try to achieve our purposes. Or, to put it another way, something is unreal for me in case that I have no need to take account of it as I pursue my special purposes. What is real, upon this definition, then, is what is part of the coherent structure that permits me to pursue and, often enough, to fulfil my purposes. Although there do seem to be three things here – experience, consistency, and purpose – one can begin to discern a pattern. For, among my purposes are my cognitive purposes: I aim to understand. In fact, pursuit of other purposes always presupposes that I do understand reality, at least that part that is relevant to achieving those other purposes. Reality, then, is what I must take account of if I am to achieve my cognitive purposes. But to understand is simply to trace out the connections that link things elsewhere and elsewhen with the reality that I presently grasp in experience. The definition of reality in terms of my purposes can thus be linked up with the definition of reality through structures that relate things to experience provided that we take into account cognitive purposes. This in turn can be related to the definition of reality as what is self-consistent if we say that the connections that we trace must be self-consistent. And so Taylor says that “... the ‘non-existent’ primarily means that which finds no place in the scheme of objects contemplated by consistent scientific thought ...” (p. 51). But, though we can begin to find connections among the three definitions of reality or esse that Taylor offers, it has again to be stated that consistency hardly seems to be the criterion of reality, and, moreover, has again to be stated that the fact that something is consistent hardly seems to imply that it is inseparable from experience. The empiricist will not disagree with regard to the claim that to understand something is to locate it as part of a pattern. These patterns are causal patterns, where causes are to be understood in terms of regularities or patterns that are discerned to hold among the objects we have experienced. Taylor, however, disagrees (p. 172ff). Upon his view, any attempt to understand things as instances of regularities is to draw a distinction between one event as a cause and another event as an effect; it is to insist that these events are distinct. Genuine understanding requires us to place the cause and the effect in a process that is continuous and in which, therefore, cause and effect are not distinct. To understand two things as part of a process requires us to grasp them as parts of a single thing, a single process, in which they are inseparable aspects. On this point the empiricist will, it is clear, disagree. The empiricist

232 will insist that to the contrary to understand is to subsume distinguishable events under regularities. Taylor is proposing a sense of ‘understanding’ which, insofar as it insists upon something more than this, is unreasonable. We have, however, made some progress. For, we have come to see that genuine understanding, upon Taylor’s view, involves grasping a relational structure in which is itself a unity and in which the relata are inseparable parts. And this is real progress, for it makes clear how Taylor gets to esse being inseparable form percipi. If the reality of something is a matter of its being related to experience, as indeed the empiricist insists, then, given Taylor’s view of relations, it follows that the thing that exists, on the one hand, is related as part of a unitary continuous structure to, on the other hand, experience, a structure in which the thing and the experience are inseparable parts. This can be part only of the picture, however. Experience in the end is not simply one of the relata in a relational whole. To the contrary, it is the very substance of the relation itself, the relation as relating, constituting out of itself the whole of the relational whole. For, according to Taylor, ... the real is a single all-embracing whole of experience of psychical matter of fact, determined entirely from within by a principle of internal structure, and therefore completely individual. Because the matter of the system is in all its parts experience, the principle of its structure must be teleological (p. 118).

This is a much stronger view, of course. How do we get to it? How, at least, does Taylor think that we get to it? The view of relations that Taylor defends leads to the following conception of reality: ... there will ultimately be only one “substance” − the central nature or principle of the system itself − of which all subordinate aspects or parts of existence will be the attributes (p. 140).

An ordinary thing is a “unity” of qualities (p. 123). This unity is provided by a relation: it is “the unity of coherent structure” (p. 126). But this structure is teleological: “That is one thing which functions as one, in other words, which is the systematic embodiment of a coherent scheme or structure” (p. 125). The qualities of ordinary things are dependent upon “its mode of relations to other things” and upon “its relation to our percipient organism” (p. 140). Moreover, the qualities themselves are related to each other, everywhere by “the relations of identity and difference” (p. 140),

233 and beyond this by other relations such as contrariety and the laws that describe how they succeed one another in the history of a thing (p. 137, p. 140). Reality cannot be simply a set of relations, since “every relation implies two or more terms which are related” (p. 141). At the same time, however, reality cannot consist of discrete individuals, as empiricists such as Hume maintain, since relations are given in experience: even “sensation itself is a continuous process containing a multiplicity of ‘marginal’ elements which in all sorts of ways modify the character of its central or ‘focal’ element” (p. 143). To be sure, the relational aspects of the situation are not themselves given in sense experience. It does not follow, however, they are not given in experience. There is in fact a higher form of experience, namely thought or reason, that apprehends the relation that structures the parts into a whole. The problem of the Humeans is that they insist, wrongly as Taylor suggests, that there is “an absolute distinction between the sensory and the intellectual factor in cognition” (p. 143). Reality is an Individual of which the elements are lesser individuals.... A thing ... is individual just so far as it is unique, and only that which is the embodiment of a single purpose or interest can be unique. A single whole of experience, owing its unity as a whole precisely to the completeness and harmony with which it expresses a single purpose or interest, is necessarily an individual. The allembracing experience which constitutes Reality is thus in its inmost nature a complete individual. And the lesser experiences which form the elements or material content of Reality are each, just so far as each is truly one experience, individual in the same sense as the whole. We may thus call Reality a complete or perfect individual consisting of minor or incomplete individuals (pp. 98-99).

The whole provides the truth, and insofar as we attempt to contemplate an individual apart from the whole we not only have a grasp of the truth that is incomplete but a grasp of something that is not the truth: there is no truth short of the whole. ... the structure of the whole is so repeated in any and every one of its members that what is not the truth about the whole is never the ultimate truth about anything, precisely because there is ultimately nothing apart from the whole, and the whole again is nothing apart form its members (pp. 154-5).

How is this ultimate reality known? Not by discursive thought or reason, according to Taylor. Such reason grasps the relational structure but the ultimate reality is beyond reason. “...[T]he relational scheme which discursive thought uses does not adequately express the true nature of the real...”

234 (p. 152). Thought which gives access to the thing that is most real, the individual in which all lesser individual appear but appear ultimately as unreal, is “the experience of the Absolute” and this “must be suprarelational”: it is ... an experience which contains the results of an elaborate process of distinction and relation, but contains them in a way which transcends the relational form and reverts in its directness to the unity of immediate feeling (p. 152).

All this from the account of relations. Reality is a whole; it is the whole truth and the only absolute truth; it is structured, but the structure is a matter of its own teleological self-causation or self-determination; insofar as the whole is teleological, it is spiritual, mind or self-consciousness; everything, then, insofar as it is real, is inseparable from experience. The esse of everything really is percipi. All this is indeed impressive, but simply raises another issue: it all follows, given the view of relations, but why accept this view of relations? Taylor does argue for this view. If we take “pluralism” to be the view that there are separable parts of things, that there are, as Moore insists, entities that are distinct, then, according to Taylor, it “must be resolute enough to dismiss the idea of systematic interconnection between its independent realities as an illusion of the human mind” (p. 90). But if it does accept that view, then “I should have no means of knowing it to be true” (ib.). For, to know it to be true would require one to know that the other entities are real and that would require that one be related to them. The point is correct, but there is a response to it: Why accept an account of relations that makes it impossible for things that are related to exist as distinct entities in the sense of being separable, that is, to be such that they can exist otherwise as they are even if the relation were not to obtain? After all, this is precisely what Moore insists upon in the case of the intentional relation. Why not accept this account of relations rather than one of the sort Taylor is trying to articulate in which relata are always inseparable? In response Taylor provides a second argument. Because the world as known, or again as providing for the coherent realisation of practical purposes, is an orderly system, and on any other supposition coherent knowledge and consistent action are alike impossible, the world must for Metaphysics be regarded as the complete embodiment and expression of a single ultimate principle (p. 94).

235 And so, “we are ... committed to some form of theory of the type generally known as Monism” (ib.). Unfortunately, this remains pretty obscure. On the face of it, it seems to beg the question, and that in at least two different ways. In the first place, even if we grant that the world requires, for knowledge and for action, a coherent structure, why should that structure be one in which the parts are united inseparably into unitary continuous wholes? Why does the existence of order require that the order be of the sort that Taylor likes? Why does the fact of order imply that Taylor’s view of relations is correct and that of Moore is wrong? In the second place, why should we assume that coherent knowledge and consistent action are always possible? Why should we grant the world has to be a place in which our practical purposes can always find fulfilment, or that the world is a place in which our cognitive purposes can always find fulfilment? Moreover, to refer back to Taylor’s criterion of self-consistency as definitory of reality or existence, why, in the first place, does coherence in the sense of self-consistency require coherence in the sense of unitary continuous wholes? And why, in the second place, does the self-consistency of reality guarantee that my interests and purposes, including my cognitive interests, will find their true fulfilment in that reality? Taylor leaves all these questions more or less unanswered. Essentially he takes his own position to be so obvious that it requires neither detailed defence nor a careful delineation of its own internal logic. Moore’s implicit suggestion that his opponents were more inclined to pontificate than to argue, more inclined to preach than to demonstrate, is not without justification. Other idealists, however, were not as inclined as Taylor was to leave the appeal to the obviousness of their position: they were prepared to address the sorts of question that we have raised. Such a one was John B. Watson (the philosopher).10 To argue against empiricism, Watson sets out to prove that “our intelligence is not infected with an absolute limit” (p. 26). This judgment, that my intelligence is absolutely limited, must itself be absolute. If there is any doubt of the truth of this judgment [that my intelligence is absolutely limited], it becomes doubtful whether other judgments may not be absolutely true; and if this judgment is false, its contradictory must be true, and therefore it must be false that all other judgments are not absolutely true (ib.).

The judgment that there is a limit is either a perfect judgment or not. But a perfect judgment is not limited. Hence, the judgment that there is a limit, if it is perfect, contradicts itself, and there is no limit after all. But if it is im-

236 perfect, then there is, in the end at least, a perfect judgment by which we judge it to be so. In thinking that there is a limit we must go beyond the limit, eventually to arrive at a perfect judgment which judges all others. ...granting the truth of the hypothesis that the universe is rational, we cannot without contradiction defend the thesis, that there is an absolute limit in the human intelligence, which prevents it from knowing that the universe is rational (ib.).

We shall take up the point about the hypothesis in a moment. But first we have to notice a further conclusion that Watson deduces. The possibility of a perfect judgment in turn implies, Watson argues, that the universe itself contains no irrationality that is beyond comprehension: “if the human intelligence is not infected with an absolute limit, it cannot be shown that the universe is any sense irrational” (p. 31) Intelligence is not an abstract power which works in a vacuum, but essentially consists in the comprehension of reality. If we suppose the universe to be infected with irrationality, the intelligence, if it proceeds upon the assumption that the universe is completely rational, is bound to find itself checked and frustrated in its effort to comprehend that which, as irrational, must be absolutely incomprehensible....Such an external limit to the intelligence necessarily implies a limit in the intelligence itself; for the intelligence can only exist, and possess the nature of intelligence, provided that it is consistent with the total nature of things. A perfectly rational intelligence cannot exist in a partially rational universe; and therefore the complete rationality of the universe is the indispensable condition of an intelligence free from any absolute limit. The rational and real must coincide ... (pp. 31-32).

Simply put, because we know that there is no limit to intelligence, intelligence can achieve a perfect judgment. But a perfect judgment which does not encompass all reality would be imperfect; hence, a perfect judgment implies that there is no reality that is not rational. Since a perfect judgment is realizable, it follows that it comprehends all reality, and therefore that reality and the rational indeed coincide. Watson, then, makes clear, what Taylor only alludes to, namely the coincidence of the real and the rational. This is not to say, however, that Watson’s argument is free from difficulties. In the first place, if he is to go from here to the idealism that Taylor holds and that Moore is aiming to refute, then he must establish that the rational is a seamless web of inseparable parts. Moreover, in the second

237 place, he must establish his own premises. For, the empiricist, if he or she is wise, will hardly agree with what Watson has to say about judgments perfect and imperfect. The empiricist understanding of what it is to comprehend a portion of the universe is for that portion to be subsumed under some matter-offact causal generality or regularity. This is the empiricist understanding of understanding or reason. The cognitive ideal, then, is for one to subsume every event under some regularity. This will be possible provided that it is in fact true that every event has a cause. However, as Hume argued, so long as the concepts we use are distinct, then the distinction that can be drawn between any event and its cause ensures that the two events are in fact not inseparable. Indeed, it ensures that the event in question might have existed without a cause. An event without a cause would, from the viewpoint of the cognitive ideal, be irrational; we would not be able to understand or comprehend such an event. However, since there is no logical guarantee that every event has a cause, it follows that there is no logical guarantee that there are no irrational, incomprehensible events in the universe. Certainly, the cognitive ideal, the ideal of a perfect comprehension of the universe, does not imply that the universe is comprehensible throughout. The real need not, on this concept of reason, be comprehensible, intelligible or rational. What Watson has to do, in order to make his case, is rule out antecedently this empiricist notion of what it is to understand. More strongly, he must establish a concept of reason such that the possibility of a perfect judgment ensures that whatever is real is rational. And in fact, Watson himself is aware of this. For, as we have seen, he does base his point about the unlimited nature of reason upon what he calls the “hypothesis” that “the universe is rational.” It is this hypothesis that he must establish. And, moreover, establish it in such a way that the rationality of the universe consists throughout of a seamless web of inseparable parts. Watson proposes, to begin, yet another criterion of the real: a thing which is real “has an existence of its own that is quite independent of other things” (p. 46). Here is yet another strand to weave into the tapestry that constitutes the idealism that Moore is trying to take apart. Nor is it at all clear exactly how it ties in with the other strands that we have got from Taylor and Watson. Let us go with it, however, and see where it leads. Watson contrasts three views of the world. The first view of the world construes ordinary things as real in the sense just indicated. The world, in other words, consists of parts that “must

238 themselves be real, self-complete and independent” (ib.). The world is thus an aggregate of small reals. Any relations that hold among things are only apparent, the result of the activity of a mind that contemplates the independent reals, and compares them in various respects. And clearly, this – what Russell was to refer to as the “monadistic” account of relations – is to deny that there is any objective relational structure. The problem with this point of view is that it neglects the fact that objectively things are related one to another, and each to everything else. Once this fact is recognized we are led to the second view of things. ...since no object can be found that does not in some way depend upon other objects, the doctrine now formulated is, that there are no independent and selfcomplete things, such as we had at first supposed, and that relations are by no means due to external comparison, but are absolutely essential to the reality of anything whatever (p. 48).

But this view, too, must be false. For, it implies that there is nothing real, and therefore nothing that can be known. What we experience is a continual phenomenal process, one in which each part is related to every other part, but in which there are no independent things, and therefore nothing real. There are, to be sure, patterns, and it is to these that the phenomenalisitic empiricist appeals in order to get on in the world. But the view “denies that we are entitled to say that they are subject to any absolute law” (p. 49). Relations and patterns obtain, but they are not absolute, not themselves real, because they are not grounded in anything real. What we need, then, are not just relations, but relations that are grounded in something real. The first view is incorrect because, while it allows for real things, it has no relations. The second view is wrong because, while it allows relations, these are not founded in any reality. To obtain a correct picture of the world we need it to consists of parts that are related where the relations are grounded in something real, that is, selfcomplete and independent. “Evidently, the reason why the mind cannot be satisfied with this view of things is that such a series does not take us beyond a reality that is dependent, and we therefore affirm it not to be true reality but only appearance” (p. 51). However, we can judge it to be only appearance, only imperfect, because we presuppose a conception of something perfect, namely self-dependent being. ... we tacitly assume that only that which is self-dependent can be real. Dependent being, in other words, presupposes self-dependent being (ib.).

239

But now we have returned to the notion of perfection, and we can infer, as before, that “if the ideal is not real, it becomes a mere fiction with which we may vainly seek to comfort our hearts, [in which case] ... the real becomes an insoluble enigma” (p. 53). The ideal, here, is the notion of a being that self-dependent and selfcomplete. Watson infers that if a being is self-dependent, it is “therefore self-causing and self-differentiating” (p. 52). The concept of cause is not the empiricist’s one of regularity but that of teleology: “We must ... substitute for the idea of external design the deeper notion of a world the very constitution of which involves immanent purpose, order and system” (p. 138). It is only from the point of view of separate events as occurring in time that we speak of causality in the ordinary sense of the term; and when we have discovered that this mechanical and external mode of thought is obviously not ultimate, we are forced to advance first to the idea of a self-developing total reality, and ultimately to the conception of God as the absolute source of all modes of being (p. 136).

The concept of the ideal being, then, is the idea of a being that causes itself to be real and causes reality to conform to its concept, that is, concept of itself as a systematic whole. As it causes itself to be real, it differentiates itself into the manifold of inseparable parts that constitute the world. The world, insofar as it is real, comes to be through the activity of the ideal being creating reality to fit the concept that describes it. The ideal being in its self-creation guarantees the truth of the concept that describes it, and guarantees the truth, too, what truth there is, with regard to all beings into which it differentiates itself and which are dependent for their reality upon it. The reality of both ordinary individual things and of their relations thus must derive in each case from a self-complete and wholly independent reality. Insofar as anything is separable from this reality, it is not real. It is in this context that we can see how the criterion of consistency applies. Take any ordinary thing. It is what it is in part at least by virtue of the relations which it bears to other things. What we cannot consistently assert is that it is both real and related to other things. That at which reason aims is a reality that explains all other things, makes them comprehensible. This real must be consistent. It follows that it must be independent of all other things, related to nothing, self-complete. Watson’s argument, here, is parallel to that of another idealist,

240 Wilbur Marshall Urban. For Urban, values are demands for their realization in fact.11 The basic moral requirement is not so much “do this” as “be this”, and a moral ideal is a demand upon our wills that it, the ideal, be realized. As a demand, it moves us, both individually and collectively. As a demand it involves an element of will, an element that transcends our own will. The “demand for reality” of our ideals (Valuation, p. 400) thus implies an “inner identity and continuity of the will with its objects or with itself” (p. 401), and, beyond that, an “identity of subjective will with a meta-empirical will not completely expressed in any of these forms” (p. 402). Our ordinary obligations are limited in various ways, but beyond them is an unconditional ideal, an unconditional value that is limited in no way, and in which our partial or limited obligations all find their appropriate place. In fact, this ultimate and unconditional ideal is the first premise of idealism: “Every idealistic theory of the world ... has as its ultimate premise a logically unsupported judgment of value – a judgment which affirms an end of intrinsic worth and accepts thereby a standard of unconditional obligation” (Intelligible, p. 92).12 This judgment of value includes our cognitive values; “the intellect is ultimately oriented towards value ...” (p. 216). This judgment of the intellect in its own way, this ultimate value judgment, encompasses all others. The logically unsupported judgment of value of the idealist is precisely the ideal of genuine knowledge which underlies his thinking (p. 92).

Urban argues as Watson does that the attempt to assert that knowledge is limited, the attempt to deny that knowledge of the whole can be attained, requires one to go beyond those limits to the whole: the attempt itself testifies to a knowledge of the whole. Every effort to deny knowledge of the whole involves ... such knowledge of the whole, every effort to deny an absolute experience involves the assertion of such experience (p. 93).

The intelligibility of things presupposes this ultimate value judgment. “... [I]t is only by assuming ... a timeless order of values that intelligible interpretation ... [is] possible” (p. 431). As for intelligibility, this involves, on the one hand, the notion of system. “The entire significance of system lies in the idea that it is necessary to make the world intelligible” (p. 430). The notion of a system is the notion of a set of parts systematically related into a whole. “The term ‘sys-

241 tem’ is in general used to designate the orderly relation of parts within a significant whole” (p. 432). Urban, like Watson, explicitly contrasts the notion of an intelligible system or whole to the notion of an ‘aggregate.’ Within the systematic whole, the parts are so related that they are inseparable each from every other. We cannot change our concept of life without changing the meaning of the other concepts in the system or hierarchy in which its meaning and validity are found. We cannot change our concept of purpose or development without changing the whole system of meanings and values in which it is found. ... Every philosophical concept is an abbreviation of system, and this fact is the deepest ground for those internal relations of concepts, categories, and values which we call philosophic system (p. 435).

Watson argued that “a rational universe must be not only one and selfdifferentiating, but it must be a coherent system. Every element in the whole must be related to every other; so that nay change in one element will involve a correspondent change in all” (Watson, p. 45). As for Watson, so for Urban: the ultimate judgment concerning the systematic whole will not be ultimate unless it includes everything. “Philosophic system must ... be all-inclusive” (p. 434). Now, “... an intelligible system, one that shall make the parts intelligible ..., must ... be explanatory” (p. 433). But we can explain, we can understand, only with reference to cause in the sense of purpose. “...[T]he meaning and value of anything ... can be expressed only in terms of purpose or finality. Consequently its ultimate origination, its sufficient reason, requires that this element of finality be included in the notion of cause or reason” (pp. 216-17). We cannot think of life except as a centre of values, except as a movement towards the good. No intelligibility without finality (p. 209).

Thus, the notion of intelligibility involves the notion of system but also, on the other hand, the notion of purpose, will, and finality. “The first element in an explanatory system is the idea of purpose or end” (p. 433). If life, reality, is a river without source or end, if the world process is from an unknown centre to an unrecognizable goal, there can be no system. But there can also be no intelligibility (p. 430).

The ultimate value judgment is, like all basic values, not an obliga-

242 tion to do but an obligation to be: value judgments acknowledge “not being in the sense either of existence or subsistence, but of worthiness to be” (p. 144). The ultimate ideal is thus a demand that it be, that is, the demand that the ideal be real. But the ideal is a systematic structure, one that proceeds from parts to ever more inclusive wholes to the final whole that encompasses all things. This whole, if it itself is to be intelligible, must involve the notion of cause and finality. ... we cannot postulate the series of degrees without bringing causality and finality together in one non-temporal system (p. 454).

It must be an end that causes itself to be real, an end that brings itself into being. As Watson put it, it must be self-dependent, self-causing and selfdifferentiating. Once again we see that for the idealist, the ideal being in its selfcreation guarantees the truth of the concept that describes it, and guarantees the truth, too, what truth there is, with regard to all beings into which it differentiates itself and which are dependent for their reality upon it. We see, therefore, how Watson and Urban provide answers to questions that Taylor leaves unanswered. With regard to Taylor’s view we saw that he left unanswered the question, why should we assume that coherent knowledge and consistent action are always possible? Why should we grant the world has to be a place in which our practical purposes can always find fulfilment, or that the world is a place in which our cognitive purposes can always find fulfilment? Watson and Urban argue that coherent knowledge is possible − the world must be a coherent systematic whole. And they argue further that this whole is the source of all that is truly valuable. It is in it, therefore, that our practical purposes, insofar as they are genuine, will find fulfilment. We asked, moreover, referring back to Taylor’s criterion of self-consistency as definitory of reality or existence, why does coherence in the sense of self-consistency require coherence in the sense of unitary continuous wholes? We now see that Watson and Urban provide the response that if the doctrine of relations is correct, then the notion of self-consistency requires the real to be an independent and self-complete whole. We further asked, in the same context, why does the self-consistency of reality guarantee that my interests and purposes, including my cognitive interests, will find their true fulfilment in that reality? We now have the answer: the self-complete whole is the source of all value insofar as it is genuine or true, and so it is in that whole, the self-consistent whole, that my purposes and interests, including my cognitive interests,

243 will find the completion. One more example will illustrate clearly the dependence upon a specific doctrine of relations of the idealism that Moore aims to refute in the first part of his argument. Consider Bosanquet’s lecture on The Distinction between Mind and Its Objects.13 Bosanquet makes perfectly clear the idealist argument is not simply Berkeley’s, that it is a doctrine of relations that for him guarantees idealism: wherever there is a relation then there is consciousness, and that the unity achieved upon this account by relations implies inseparability of the relata, the parts of the unity. Bosanquet considers a blue sensation, a patch of blue that I sense, and suggests What I see when I look at a blue thing has unity, and life. Its parts, that is, though varied, confirm, support and determine one another by explicit “compresence.” It pulsates with feeling, a common tone, which involves the presence of a whole all at once, reinforcing and modifying every part by the simultaneous effects of all (p. 32).

The various parts of the sensation are related to one another by relations of similarity and dissimilarity; these relations create a unity; this relational unity cannot exist except through the activity of mind. How do the elements of the [sensation] hang together? What makes the blue reinforce or modify the blue? There is no push or pull between them. They work on each other through their identity and difference; or, to avoid disputes, here irrelevant, through their likeness and unlikeness. What sort of medium does such a unity involve? Surely, that of consciousness and no other. Blue, then, while it retains the characters of blue, must have it in the life of the mind (p. 33).

As Watson and Urban emphasized, so does Bosanquet, that relations not only imply consciousness but also the notion of a complete system. Moreover, the parts of this system cannot be as they are apart from the system: to be a sensation that is blue is for the sensation to be related to other entities, if only by relations of likeness and unlikeness, and to be related is to be inseparable from consciousness. ...no world can be synthetic in itself, that is, can possess universals as a part of its own nature, if its elements have not, pervading them, the living nexus and endeavour towards a whole which indicates participation in the nature of minds. I cannot understand any attempt to explain a universal which does not recognise that it absolutely consists in the effort of a content to complete itself as a system

244 (p. 35).

What establishes idealism, on this view, is not the collapse of the act-object distinction, as in Berkeley. Rather, what establishes idealism is that there can be no structure apart from a structuring activity. “There can be no concrete whole but a whole centering in mind, and no self-existent whole but a concrete whole” (p.p. 39-40). It is not so much that qualities, whether secondary or primary, are mind-dependent, as Berkeley would have it, that is, cannot exist save as predicated of a substantial mind. It is instead that they are parts of a unity that is achieved through the activity of consciousness: the idealist conclusion concerning all qualities “is drawn..., one might say, from their being mind-component; that is, possessing a logical character or implicit unity, which finds completion only in the focus of mind, which, in turn, it constitutes” (p. 42). The root of idealism ... is not the failure to distinguish between an act and an object of mind. It is not any simple prejudice that mind can apprehend only what is part of itself. But it is the insight – an insight substantially just – that a universe severed from the life of mind can never fulfil the conditions of self-existence.... to overlook the character of mind which bears on this point, when stating the simplist points of perception, is to be misled ab initio. Mind is always a world; its objects are always fragments (p. 38).14

The idealism that forms the background to Moore’s essay, and which he is trying to refute in the first part of his argument, thus crucially presupposes a doctrine of relations. There remains one question, however, which does not receive a clear answer. The crucial point in the case developed at length by Watson and Urban, and sketched in his lecture by Bosanquet, is that the self-consistent whole of which they speak will do the work that it has to do provided that one accepts the doctrine of relations that they propose. They can provide what they do only if one accepts this doctrine. Now, in contrast to traditional empiricism, this idealist ontology does not deny that there are relations among things. Watson is correct in holding that in this respect the idealist picture of the world is the more accurate: relations do exist. But we have not just the affirmation that there are relations but rather something much more specific, a particular doctrine of relations. It is a doctrine of relations that has the consequences we have seen concerning the whole, and, in particular, the consequence that esse is, in a

245 very strong sense, percipi.15 Specifically, we recognize the following. The doctrine of relations must be such that where one has a relation predicated of two things, there is a third individual, the whole. The two relata are parts or aspects of this whole, and insofar as they are such parts or aspects they are not genuine individuals. The whole alone is a truly real individual. The whole itself not only has these aspects but is characterized as being a whole of a certain sort depending upon the way in which the parts are related. The original things are inseparable parts of the relational whole. Moreover, the whole is active: it causes its own differentiation into the parts according to the relational structure that characterizes it. Unless the doctrine of relations had these features, the consequences held to follow from the fact of order by idealists such as Taylor, Watson, Bosanquet, and Urban – and by Bradley, for it is above all his doctrine of relations that we are talking about – these consequences that the idealists draw simply would not follow from the fact that things do, really, stand in relations to one another. Thus, for example, the empiricist is going to deny that in order to understand one must have recourse the category of finality. To be sure, purposes and values, that is, valuings, are among the relevant variables to which appeal must be made if we are to understand certain things. But the understanding consists in general, and in this case in particular, in subsuming events and processes under regularities. It is just that in these cases the regularities describe among other factors the purposes and values of certain beings. This brings us, of course, to the one question that we asked regarding Taylor that Watson and Urban do not answer. This is the question. Even if we grant that the world requires, for knowledge and for action, a coherent structure, why should that structure be on in which the parts are united inseparably into unitary continuous wholes? Why does the existence of order require that the order be of the sort that Taylor, Watson, Urban, Bosanquet and Bradley like? Why does the fact of order imply that this view of relations is correct and that of Moore is wrong? In fact, there is no argument. As Moore and Russell never refrained from pointing out, the doctrine of relations defended by the idealists was taken for granted. As Russell remarked in his review of Joachim’s The Nature of Truth,16 “The arguments in the pages we have been considering are...such as will only appear cogent to those who already admit the conclusions which the arguments are intended to prove.”17 No arguments for

246 the doctrine of relations were in fact offered − other than the argument that we found in Watson, that this idealist account of relations, unlike the monadistic account found traditionally in empiricism, could account for order. Moore, however, wants to do more than just show that the idealist doctrine of relations has nothing much in its favour. He wants to argue that the doctrine is false, to refute it. Why he wishes to do this becomes apparent from another discussion of his. It is part of the idealist doctrine of relations that things that are related become, insofar as they are real, inseparable parts of a more comprehensive, more comprehendable, whole. Since intentionality constitutes a structure relating mind to what it is thinking about, it will be taken, upon the idealist account, to be a connection such that the relata, insofar as they are real, are inseparable parts of a greater whole, that in which mind and its object are unified into one thing or individual. Indeed, as Watson puts it, “[t]o distinguish an object from himself, the subject must comprehend within his embrace both himself and the object” (p. 61). Moore, of course, wishes to dispute that conclusion. If he is to do that then he must refute the doctrine of relations that implies it. Beyond the simple issue of intentionality there is in fact more. If the idealist doctrine of relations is true, then entities that enter into different relational wholes will be, insofar as they are real, different. Hence, if I think of X and you think of X, X will enter into two different relational wholes, one involving my mind and one involving yours. It follows that the X insofar as it is related to my mind will be different from the same X insofar as it is related to your mind. Thus, the X that I think is in fact not the same X that you think. Upon the idealist account of relations, in other words, it is impossible for two of us to think exactly the same thing. But idealists wish to hold that two persons can experience the same fact. Thus, Watson, for example, holds that a perceived fact “is objective, because it is actually experienced under certain definite conditions, and under those conditions is identical for all” (Watson, p. 63). Moore’s point is that you can’t have it both ways. And if you want the latter, then so much the worse for your idealist doctrine of relations. Moore suggests, therefore, that one can convince the idealist of the incorrectness of his or her doctrine of relations by getting him or her to agree, on the one hand, that two persons can indeed think the same thing, exactly the same thing, and by getting him or her to agree, on the other hand, that his or her doctrine of relations entails that this is impossible.18 This makes clear, I think, why Moore thinks that it is important to

247 refute the idealist: only if idealism is refuted can one consistently hold that there is objective knowledge, truth that is accessible by several ordinary – or “finite” – minds. Moore clearly holds that such objectivity is evident, and that is it evident moreover to the idealists themselves. It is, therefore, a fact to which he can appeal in order to get the idealists to recognize the incoherence of their position. This is the sort of appeal to commonsense that Moore would continue to use throughout his career. It is, moreover, not only evident, not only commonsensical, but important. So it should be defended. It is this, I would suggest, that motivates Moore in his efforts not merely to convince the idealist that he or she is wrong but more strongly to refute idealism. However, Moore’s attempt to refute the idealist, as opposed to convincing him or her of the incoherence of the position, does not appeal to this fact of objectivity but to a rather different premise. The appeal is made in order to show that the idealist doctrine of relations is false. For, that is the basis upon which rests the idealism that Moore aims to refute in the first half of his argument. Moore uses his premise to show the falsity of the proposition, implied by the idealist doctrine of relations, that relata must always be inseparable. Now, while Moore proposes thus to argue against the idealist account of relations, it is also true that he, Moore, unlike the traditional empiricists, does not reject relations. Indeed, he accepts them, in contrast to the traditional empiricists. That, after all, is his point against Berkeley. What he does do is insist that things, properties, and relations are distinct, that, while they are related one to another, they are also separable. Moore urges, in other words, that by itself RELATEDNESS DOES NOT IMPLY THAT DISTINCT RELATA ARE LOGICALLY INSEPARABLE. The idealist, Moore suggests, holds that yellow and the sensation of yellow are indistinguishable. Hence, “to assert that yellow is necessarily an object of experience is to assert that yellow is necessarily yellow – a purely identical proposition, and therefore proved by the law of contradiction alone” (“Refutation,” p. 14). But unlike “A is A”, the proposition, that yellow is a sensation, is substantive, and worth defending. Hence, “the proposition also implies that experience is, after all, something distinct from yellow – else there would be no reason for insisting that yellow is a sensation... ” (ib.). Thus, the very assertion of the idealist, that yellow is inseparable from the experiencing of it, is one that would not be made unless in fact yellow and the experience of it are in fact distinct and therefore separable. The doctrine of relations proposed by the idealists maintains that, since yellow and the experience are indeed related, and therefore also

248 maintains that yellow and the experience are inseparable and therefore not distinct. But to state the doctrine presupposes that yellow and the experience are distinct. The very attempt to state the doctrine of relations thus refutes that doctrine. The doctrine of relations thus does not so much establish that there is no distinction as obscure the fact that such a distinction exists. When, therefore, we are told that green and the sensation of green are certainly distinct but yet are not separable, or that it is an illegitimate abstraction to consider the one apart from the other, what these provisos are used to assert is, that though the two things are distinct yet you not only can but must treat them as if they were not. Many philosophers, therefore, when they admit a distinction, yet (following the lead of Hegel) boldly assert their right, in a slightly more obscure form of words, also to deny it. The principle of organic unities ... is mainly used to defend the practice of holding both of two contradictory propositions, wherever this may seem convenient. In this, as in other matters, Hegel's main service to philosophy has consisted in giving a name to and erecting into a principle, a type of fallacy to which experience had shown philosophers, along with the rest of mankind, to be addicted. No wonder that he has followers and admirers (pp. 15-16).

Moore’s argument is clear. It is in effect an appeal to the empiricist's Principle of Acquaintance (PA).19 We do in fact know yellow; it is presented to us. We also know experience; this too is presented to us. These entities that are presented to us are presented as distinct. They are presented as self-complete, and to know the one does not require us to know the other. The entity yellow as we know it, as it is presented to us, is presented as an entity such that there is nothing about it that requires it to be connected inseparably with experience. To identify it as yellow does not require us to identify it as somehow connected to any other property, which would after all be required if it were, as the idealists assert, inseparably connected to some other entity, e.g., experiencing. This argument from acquaintance can be found in earlier philosophers. It can be found, for example, in Locke. Locke considers the regular activities of external substances. These include the production of the ideas of the secondary qualities, the simple ideas of red, sweet, etc. For these activities to be explained as those who defend necessary connections require, there must be necessary connections between red, sweet, etc., and the natures or real essences of the substances that cause these qualities to appear. These necessary connections must be both ontological, in the entities themselves, and epistemological, giving us, when in the mind, scientific

249 knowledge of those entities. But, Locke argues, we grasp no such connections. ’Tis evident that the bulk, figure, and motion of several Bodies about us, produce in use several Sensations, as of Colours, Sounds, Tastes, Smells, Pleasure and Pain, etc. These mechanical Affections of Bodies, having no affinity at all with those Ideas, they produce in us, (there being no conceivable connexion between any impulse of any sort of Body, and any perception of a Colour, or Smell, which we find in our Minds) we can have no distinct knowledge of such Operations beyond our Experience; and can reason no otherwise about them, than as effects produced by the appointment of an infinitely Wise Agent, which perfectly surpasses our Comprehension (Essay,20 Bk. IV, Ch. iii, sec. 28).

Locke’s appeal to a Principle of Acquaintance is clear. Properties are presented as logically self contained; there is nothing about them as presented that requires us when we are identifying them to refer as a matter of necessity to other properties, those to which they are necessarily tied. We are, in other words, not presented with necessary connections, and we must therefore, by PA, exclude them from one’s ontology. It is this Lockean argument that Moore re-introduced into philosophy. In fact he had already made much the same point in his earlier essay on “The Nature of Judgment.”21 Moore there makes the point that the idealist account – Bradley’s account – of relations requires us to ascend to the Absolute before we can make any true judgment at all.22 For Bradley,23 a sign is an entity with meaning; it could be linguistic or it could be mental. But, contrary to the classical empiricists, the meaning is not an image: the details of images are irrelevant to the meaning they have for us. So far, Moore agrees with Bradley: meaning is not an image. Now, for Bradley, a sign, like any entity, has, in his terminology, a content: its content is the totality of its properties – “the complex of qualities and relations that it contains” (Principles of Logic, p. 3). A sign’s meaning cannot be its total content, since several signs can have the same meaning. Its meaning is but a part of the content of the sign, a part that is “cut off, fixed by the mind, and considered apart from the existence of the sign” (p. 4). Blue is, of course, the meaning of the awareness of blue. This implies that blue is only part of the content of the sign, the awareness. Insofar as this blue is isolated and abstracted from the totality of the relational complex, it is not grasped as it really is. In order to identify truly, or, more truly, precisely what it is that is presented to us, we must locate it relative to the other things to which it is related. Any two ideas, or, more generally, any two signs, will

250 have different contents, if only because they will stand in different relations to different things. This implies that, if two ideas, perhaps both mine, or perhaps mine and thine, have the same meaning, then they have but part of their content in common. More accurately, each will have a part, and these parts will be the same. However, to say this is to place those parts in the context of certain relations of similarity and dissimilarity.24 In order, therefore, to know that these contents are the same, we must become aware of the whole, the relational complex, by virtue of which they are the same. This is what is required by Bradley’s idealist account of relations. In fact, in order to know precisely what the meaning of an idea is we must place it in its total context, Since everything, meanings included, is in the end related to everything else one way or another, it follows that with respect to any meaning, we cannot know precisely what it is, we cannot identify it as it is in itself, unless we proceed through all those relations to a grasp of the infinite whole. Thus, Moore correctly infers, we can know what really is the meaning of one of our ideas only if we have passed through an infinite regress of judgements: as Moore puts it, the task of knowing precisely what we mean requires “the completion of an infinite number of psychological judgments before any judgment can be made at all” (“The Nature of Judgement,” p. 178). Again, the idealist account of relations ...presupposes that I may have two ideas, that have a part of their content in common; but...at the same time compel[s] us to describe this common part of content as part of the content of some third idea (p. 178).

That is, Bradley’s idealist account of relations requires the introduction of a third particular, the Whole, over and above the two particulars that stand in relation to each other. This relational whole is such that the one quality cannot be identified independently of its necessary connections to other qualities.25 But, Moore indicates, in fact qualities can be identified as themselves without reference to the relations in which they stand to other qualities and other things. Moore makes the point in his “The Nature of Judgment” with specific reference to the relation between qualities and the mind that knows them, in a way therefore that touches specifically on the concern of “The Refutation of Idealism”: It is indifferent to their nature whether anybody thinks them or not. They are incapable of change; and the relation in to which they enter with the knowing subject implies no action or reaction (“Nature of Judgment,” p. 179).26

251 Moore makes the point about qualities being identifiable apart from the relations in which they stand with specific reference to the cognitive relation, but he implies the point in full generality (and was so to state it elsewhere, in his Principia Ethica, as we shall see in a moment), and Russell27 was later to cite Moore's specific statement but only to assert it himself in that full generality: To say that two terms which are different if they were not related, is to say something perfectly barren; for if they were different, they would be other, and it would not be the terms in question, but a different pair, that would be unrelated. The notion that a term can be modified arises from neglect to observe the eternal self-identity of all terms and all logical concepts, which alone form the constituents of propositions [here Russell refers to Moore’s “The Nature of Judgment”]. What is called modification consists merely in having at one time, but not at another, some specific relation to some specific term; but the term which sometimes has and sometimes has not the relation in question must be unchanged, otherwise it would not be that term which has ceased to have the relation (Principles, p. 449).

Note that here Russell is allowing the idealist’s point against the monadistic account of relations. On the latter, the predications of one term of a relation would not change if the other relatum ceased to exist. Russell accepts this. What he is denying is the implication of the idealist account of relations, made explicit as we saw by Bosanquet, that there is something about properties or qualities as presented that requires us when we are identifying them to refer as a matter of necessity to other properties, those to which they are necessarily tied. Russell is holding that properties are presented to us as logically self-contained rather than as necessarily tied to one another; he concludes that there are no such necessary connections. But such connections are required by the idealist account of relations. The falsity of the latter view follows. Russell’s rejection of Bradley’s idealist account of relations on the basis of an appeal to the empiricist’s Principle of Acquaintance is clear. Moore argued the same point in his Principia Ethica.28 Here he looked at the issue from the perspective of the idealistic doctrine that parts in relation formed an “organic whole,” where, given the idealist account of relations, “the whole would not be what it is but for the existence of the parts, [and] ... the parts would not be what they are but for the existence of the whole,” understanding this to mean not merely causal dependence but more strongly a dependence such that “the part is no distinct object of thought” in as much as “the whole, of which it is a part, is in its turn a part

252 of it” (p. 33). Moore objects that That this supposition is self-contradictory a very little reflection should be sufficient to shew. We may admit, indeed, that when a particular thing is a part of a whole, it does possess a predicate which it would not otherwise possess – namely that it is a part of that whole. But what cannot be admitted is that this predicate alters the nature or enters into the definition of the thing which has it. When we think of the part itself, we mean just that which we assert, in this case, to have the predicate that it is part of the whole; and the mere assertion that it is a part of the whole involves that it should itself be distinct from that which we assert of it. Otherwise we contradict ourselves since we assert that, not it, but something else – namely it together with that which we assert of it – has the predicate which we assert of it (p. 33).

The contradiction is clear29: in asserting that “this (entity) is part of such and such relational whole” we are at once distinguishing the this from other entities, thus asserting that it is distinct from those entities, but at the same time asserting that it is not distinct from those entities, since the relational connection means that it is not after all distinct; we are asserting both that the entity is distinct and that it is not distinct. This contradiction would in fact not disturb the idealists, as we have seen. From this it follows that truth is attained only with the absolute, and not with any part of it. The thesis that assertions about parts are always self-inconsistent and the thesis that truth is attained only with the absolute are both theses that follow from the idealist account of relations. Moore’s response is to appeal to the empiricist’s PA: in our judgments about things we do experience them as distinct: “it is obvious that no part contains analytically the whole to which it belongs, or any other parts of that whole” (p. 33). We “cannot admit” that a “predicate alters the nature or enters into the definition of the thing which has it” without violating what we know to be true in our judgments, that the things about which we judge are distinct, that the relations they have to predicates and to other things do not alter their natures.30 Moore and Russell, we see, appealed systematically to the empiricist’s Principle of Acquaintance to make their case against the idealists. It was this appeal that Moore re-introduced into philosophy. Philosophers seem in fact to have found it fairly convincing. After all, the idealist doctrine of relations which denied that there are any real distinctions is no longer with us. Moore’s argument is that a relational whole “does contain analytically that which is said to be its part” (Principia Ethica, p. 34), contrary to the claim of the idealists that to analyze is to falsify. Hence the origin of the designation “analytical philosophy,” which has come to charac-

253 terize at least English-speaking philosophy ever since. That is how persuasive Moore’s argument was, how persuasive his appeal to the empiricist’s Principle of Acquaintance. In fact, if we turn to more recent philosophers who have found various aspects of Bradley’s philosophy attractive, at least so far as it emphasized things neglected by various other of their contemporaries, have stated those points in a way that is compatible with Moore’s appeal to the empiricist's PA. While these philosophers are inclined to emphasize, as did the idealists, the relational structure of the world, they do so in a way that allows that there are in fact real distinctions among things, properties and relations. But in allowing the latter, in allowing that there aspects to the world – e.g., mind and what the mind knows – which are not logically, or ontologically, inseparable. Thus, to take one, though an important, example, consider some of the views of T. L. S. Sprigge,31 perhaps the leading idealist in Britain today. What Sprigge wants to do is to rehabilitate some of the idealist notions, particularly some of those concerning relations. One example is his attempt to render plausible the holistic principle of the idealists that “to be related requires belonging together in a more genuine whole” (p. 267). He asks us to consider the streets of a town. ... one could say that the streets, understood as the full social realities which they are, belong together in the life of the town, region, or country, and that at some level one of these is going to be more of a genuine individual than the individual streets. It will be more of a genuine individual in that it makes more sense to try to understand it in its own terms. To try to understand one individual street, in the sense of explaining, or even giving much of a description of, what is going on there, while confining oneself simply to what lies within its bounds, would be hopeless; it requires rather a study of the town as a whole, or perhaps of the region or country to which it belongs. Thus if you take the street as a ‘concrete’ reality, as the whole of one distinguishable part of the universe, then it seems reasonable to say that its relations to other things are always a matter of how it unites with them to constitute something larger which is more of a genuine individual than they are (p. 267).

Now, if the empiricist knows what he or she is about, it is unlikely that they will disagree with this. What Sprigge is insisting is that in order to explain a social phenomenon, it is necessary to introduce social relations as relevant variables. Just as, in order to understand the motions of the planets, it was necessary for Newton to consider the distances of the bodies of the system relative to each other as among the relevant variables.32 There

254 have been those who neglected the role of social relations. This can be seen in those who demand individualistic explanations. It can be seen in those who insist that the atomistic economic man is the paradigm case for social explanations. It can be seen in some philosophies of the social sciences, John Stuart Mill’s to give one notable example.33 But that said, it is also true that there is nothing in empiricism that requires one to ignore either relations or social relations: the existence of the former generically and the latter specifically is quite compatible with accepting the injunction to do philosophy within the constraints established by PA. In any case, upon Sprigge’s usage it is clear that a genuine individual is one that is causally complete, one where the individuals within it exemplify jointly a complete set of relevant variables, and where no individual outside the system is (terribly) relevant.34 Such a system is the ideal of scientific explanation – it is the sort of explanation achieved by Newton with regard to the solar system. In contrast to the way in which a system that is complete and closed stands alone causally, completely determined within itself, individuals within such a system do not in terms of their causal relations stand alone. The point is that Sprigge is here not only correct but states a position that is quite compatible with an empiricism defined by the PA to which Moore appealed in his “refutation of idealism.” The point about causal relations was in fact emphasized by the idealists against those who had neglected it. The radical individualism that ignores social relations has returned in spite of the idealists. One can find it, for example, in Popperian “methodological individualism.” It is good that Sprigge should recall to us the idealist insight that this is a view that simply will not do. The point is, however, that when Sprigge does so recall us he takes for granted Moore’s critique of idealism, and recalls us in a way that turns out to be fully compatible with empiricism’s PA to which Moore appealed. This is also evident in another point made by Sprigge in defence of the idealists. Sprigge points out that “the qualities which things exhibit when seen in isolation are incompatible with the qualities they present as elements in larger totalities” (James and Bradley, p. 416). This is something with which no one who disagrees with the idealists need disagree. If we have a physical object moving in isolation, then it will have one velocity. If it moves into a certain relation with another object, then it will, by virtue of the interaction, move in a different way. In other words, it will have a property when it is related to another object that it would not have if it were not so related. Similar processes occur in perception. An object in

255 isolation will present one colour to the observer, but if it is presented standing in certain relations to other individuals of a somewhat different colour, both those individuals will present a different colour to the observer. Again, put a straight stick in relation to water and it appears bent, presents to one that geometrical shape where it presented before a different geometrical shape. Once one takes into account the causal relevance of relations none of this is problematic. In contrast, for the traditional empiricists it was indeed problematic, since their monadistic account of relations made it difficult for them to allow that relational as opposed to non-relational properties were causally relevant. The idealists made a sound point against this traditional empiricist position. But once relations are admitted, as Moore and Russell admit them, there is no reason why this point should not be granted. To grant it requires only that relations be admitted, not that the idealist account of relation be accepted. Sprigge in effect recognizes this when, having drawn attention to these cases, he does not proceed from them to the further conclusion, implied by the idealist account of relations, that the properties of the entities, either when they are related or when they are not, are so inseparably united that they are in the end not distinct. This implication, clearly suggested by, for example, Bosanquet’s discussion above, is simply not accepted by Sprigge. To that extent, then, while he accepts insights defended by the idealists against the monadistic account of relations of the earlier empiricists, rejects what Moore set out firmly to reject, the implication of the idealist account of relations that entities that are related lose their distinctness in an inseparable union. Sprigge, in other words, on this point accepts the conclusion justified by Moore’s appeal to PA. Sprigge’s conformity to Moore’s PA is still more clear in another appeal which he makes in defence of the idealists. Russell argued against the idealists that their account of relations – what Russell called the monistic account of relations – could not in the end deal satisfactorily with a number of issues.35 Bradley’s was, of course, the best articulated of the idealist accounts of relations.36 As did the other idealists, Bradley insisted, against the traditional monadistic account of relations that the empiricists had adopted, that there are relations that are genuine in the sense that their relata are not independent, or, equivalently, in the sense that the being of one relatum is not separable from the being of the other relatum.37 If it [a relation] is to be real, it must be so at the expense of the terms, or, at least, must be something which appears in them or to which they belong. A re-

256 lation between A and B implies really a substantial foundation within them. This foundation, if we say that A is like to B, is the identity X which holds these differences together. And so with space and time − everywhere there must be a whole embracing what is related, or there would be no differences and no relation. It seems as if a reality possessed differences A and B, incompatible with one another and also with itself. And so in order, without contradiction, to retain its various properties, this whole consents to wear the form of relations between them (A & R, p. 18).

But like the other idealists, Bradley insisted that where there is a relational fact there is an individual constituted by the two related individuals forming a whole which is a further individual. “Nothing in the end is real but the individual...” (“Relations,” p. 663), is the way Bradley puts it, and therefore infers that the whole must itself be an individual. Bradley thus applies the dictum in particular to relations: “A relation, to be experienced and to be actual, must be more than a mere abstraction. It must be an individual or particular fact, and, if less than this, it cannot be taken as itself” (“Relations,” pp. 635-6). Bradley concludes against the monadistic account that “there must be a whole embracing what is related”, and imposes the condition that the parts be inseparable in this whole. Thus, “Your individual term is an abstraction always. It implies what we may call a selection from the concrete fact of the whole and entire experience” (“Relation,” p. 646). This relational whole must furthermore be itself a substantial particular. From these two conditions Bradley infers his own account of the nature of this whole: The relata A and B are different things within a whole (A, B), which itself is an individual and which has those two terms as aspects. This whole then “consents to wear the form of a relation”; thus, if A and B stand in the relation R, then the correct representation of this fact consists in attributing a property corresponding to R, say r, to (A, B). Thus, according to Bradley’s account, the correct way to represent the fact reported by (@) a is R to b is given by (*) (a, b) is r or, r(a, b) The whole (a, b) is itself a particular thing, of which the two terms a and b are but aspects, and where the arrangement r characterizes the whole. But this whole consists of the relata as parts. Thus, the relation holds of the relata, not separately as in the monadistic account, but jointly: “where the

257 whole, relaxing its unity, takes the form of an arrangement, there is coexistence with concord” (A & R, p. 19). Russell criticized this account on the ground, among others, that it did not account for the fact of order.38 If we consider the example of a cup upon a saucer, the two particulars stand in a certain relation, namely that of being on top of: the cup is on top of the saucer − and not conversely. This relation orders the objects in the fact in a certain way: the relation goes from the cup to the saucer. But the two particulars are to be treated symmetrically, upon Bradley’s analysis, relative to each other, relative to the whole of which they are both parts, and relative to the relational form exemplified by the whole: at no point are we able to distinguish, upon this account, whether it is the saucer that is upon the cup or the cup that is upon the saucer. Sprigge has attempted to defend Bradley on this point. In fact, as it turns out, while this attempted defence succeeds in making a valid phenomenological point, it fails to touch the substance of Russell’s ontological critique of Bradley in particular and the idealist’s monistic account of relations in general. More importantly, as we shall see, Sprigge’s defence turns out to be compatible with Moore’s critique of idealism. Sprigge tells us that The monist must interpret the proposition ‘This cup is above the saucer’ as a characterisation of a certain total situation comprising the cup and the saucer. Now certainly if he took it as the predication of a mere aboveness of that situation, then indeed he could not distinguish its significance from that of ‘That saucer is above that cup’. It is evident, however, that he would take it as the application to a sensibly presented totality of the predication being (the totality) of a cop on top of a saucer, which is a quite different gestalt quality from that of being (the totality) of a saucer on top of a cup (p. 407).

This is how Sprigge reads such Bradleyan passages as the following: We saw that all judgement is the attribution of an ideal content to reality, and so this reality is the subject of which the content is predicated. Thus in “A precedes B,” the whole relation A-B is the predicate, and, in saying this is true, we treat it as an adjective of the real world.39

Sprigge comments that “It is evident that by ‘the relation A-B’, Bradley means the whole relational situation of A preceding B, not the mere relation of precedence which happens to hold between A and B” (James and Bradley, p. 407).

258 Sprigge argues that the judgment that the cup is on the saucer, understood as attributing a gestalt quality of the whole is not the same a conjunctive judgment to the effect that “This is a cup & that is a saucer & this is on top of that” (pp. 407-8). This is undoubtedly a correct phenomenological point. He also makes the point that in making the judgment our judgment is to the effect that “This totality is a cup on a saucer,” rather than a judgment in which we attend to two separate particulars (pp. 409-10). That is, the gestalt quality is predicated of a single subject; it is treated as a monadic predicate and not a dyadic predicate with two subjects. This, too, is a valid phenomenological point. Now, it is clear that Sprigge does have a point, that if Russell insists that the predication of a gestalt quality is simply a conjunction of three predications, two non-relational and one relational, then Bradley does better justice to the phenomenology of judgement. But it is clear from the overall thrust of idealism that it is doing more when it asserts its theory of relations than simply insisting that there are circumstances in which we predicate gestalt qualities of totalities. The idealist theory involves much more than this; otherwise (what Sprigge does not note) the account of relations by itself would not imply idealism, the inseparability of any set of related things both from themselves and from mind. Still, it might nonetheless be true (as Sprigge does suggest) that predications of gestalt qualities enables one to capture, contrary to what Russell argues, the notion of order. This might not be the whole of the idealist account of relations, but it will be part of it. Still, even if, as I think we must do, grant the correctness of Sprigge's phenomenological points about judgements involving gestalt qualities, it does not follow that Sprigge has shown that the idealist account of relations can capture, as Russell rightly insists it must, the relevant relational order. There is, to be sure, a problem that Sprigge creates for Russell, namely that of providing a way of formulating a text that at once expresses the content of the gestalt judgements to which Sprigge directs our attention, and does so in a phenomenologically correct way. But it does not follow that the notion of a gestalt quality captures the feature of order in a way that makes it possible for Bradley to avoid Russell’s criticism. In fact, what I want to suggest is that we can provide for Russell a phenomenologically adequate account of judgements which predicate gestalt qualities of totalities. It turns out, however, that these predications presuppose order rather than account for it. They therefore cannot provide a basis for defending Bradley’s idealistic account of relations in which the

259 relata always occur symmetrically relative to each other. Now, to speak of a “gestalt” quality is to speak of a structure that is predicated of a whole. There certainly are such qualities. Think of four objects so arranged that the relations that they are in determine that they form a square. This property, that of being a square, is one that is exemplified by the four objects jointly, by that totality, in other words, and not by any of the four individual objects that, related as they are, form that totality. Sprigge is quite correct when he points out that these gestalt qualities are often presented to us in experience. But do they provide any sort of defence of the idealist account of relations? The question must become, What would Russell be able to say about such qualities? In thinking about relations we must distinguish < a, b > from < b, a > as two different “ordered pairs.” These ordered pairs are taken to satisfy such axioms as < a, b > = < b, a > . = . a = b The usual axioms for ordered pairs establish that such pairs satisfy the same formal axioms as pairs ordered by relations. Sets of ordered pairs are what Russell called relations in extension. As for relations in extension, Russell introduces these into Principia Mathematica in Section *21.40 He defines a predicate (i)

ûŵ[R(u, w)](x, y)

This is a predicate expression in just the sense in which ‘R(x, y)’ is a predicate expression. (i) is so defined that it holds of x and y if and only if (ii) R(x, y) (i) is a class of ordered pairs; it can be proved that the members of (i) satisfy the axioms that must be satisfied by order pairs. Specifically, (i) is the class of pairs ordered by the relation R; it is, in other words, the relation R considered in extension. The order in the pairs is that which is established by the relation R. Thus, the concept (i) does not introduce the notion of order but rather presupposes it. Russell does not make use of the notion of an ordered pair, but his apparatus enables one to introduce such a notion. An ordered pair is a set, a totality; it is a set which has but two members, where these members are

260 taken in a certain order. Recognizing this, we see that the ordered pair < a, b > can be defined as (iii)

ûŵ[u = a & w = b)](x, y)

(iii) will hold of x and y just in case that we have (iv)

x=a&y=b

It is thus analytically true that (iii) holds of the pair a and b and analytically false that it holds of any other pair. In particular, it is analytically false that (iii) holds of the pair b and a. Here again, (iii) does not introduce order but rather presupposes it. The order which it presupposes is represented by the typographical order of the conjuncts in (iii), or, what amounts to the same, in (iv). If it is Bertrand who has selected the arrangement (iv), then the order among the individuals a and b that is represented by the typographical order is given by the relational fact: (v)

a and b stand in the relation of prior to posterior in Bertrand’s selection

(v) is of course an objective relational fact, though it is, to be sure, a fact that exists by virtue of an arbitrary choice made by Bertrand. In this sense, the arrangement is indeed arbitrary, but it is still an arrangement, a relational fact about the individuals a and b. It is the relation (v) that orders the pair consisting of a and b, and it is this order that is represented by the typographical order in (iii). Here, as in the case of (i), we see that the defined predicate (iii) does not introduce the notion of order but rather presupposes it. The notation for an ordered pair allows us to refer to the pair of entities that are ordered or structured by some relation or other, but without making specific reference to the relation that does the ordering. The terms referring to ordered pairs can function grammatically as subject terms of sentences that are used to make statements about the structured wholes to which the term refers. Russell in this provides a notation for referring to the structured wholes to which Sprigge directs our attention. We can similarly introduce by contextual definition one-pace or monadic predicate terms that can be predicated of ordered pairs. Thus, we can introduce

261 R (< x, y >) = Df R(x, y) ‘R’ is a one-place predicate that takes as its subject terms the special terms that refer to ordered pairs. Thus, to say that an ordered pair is R is to make a non-relational predication of a structured whole understood in context to be a single unified totality. Similarly, we can define (F-R-G) ( < x, y >) = Df Fx & R(x, y) & Gx which again constitutes a one-place, or non-relational, predicate applying to ordered pairs understood in context to be single unified totalities. Now, with regard to Sprigge’s cup and saucer, what he is saying is that we often make judgments about such situations that cannot reasonably be expressed as merely conjunctive. Rather, they are judgments best seen as predicating a non-relational gestalt predicate of a single complex taken as a unitary whole. Neither Russell nor the empiricist need disagree with Sprigge that there is a phenomenological difference here in the two sorts of judgment that must be captured. What we have just seen is that Russell provides us with the resources for expressing this sort of difference on the side of the judgment. If a and b are Sprigge’s cup and saucer, then what he is saying is that the judgment that is expressed by (+)

(F-R-G) (< a, b >)

is different from the judgment that is expressed by (++)

Fa & R (< a, b >) & Gb

or by (+++)

Fa & R (a, b) & Gb

The point that Sprigge is making concerns the text that is most appropriate for expressing a judgment. He insists, quite correctly, that certain judgments about wholes or totalities must be expressed by sentences making non-relational, non-conjunctive predications of wholes (pp. 407-8). The issue is whether this supports the idealist account of relations or is, rather, compatible with the account of relations given by Russell. It is of course true that the phenomenological point about relations that Sprigge

262 makes comes down on the side of Bradley against the monadistic account of relations that the early empiricists had taken over from the tradition. But it comes down equally in favour of any account that asserts, against the monadistic view, that relations must be given genuine ontological status. It thus tells equally in favour of Russell. It tells against Russell only if one insists upon accepting as principle the thesis that different judgments expressed by sentences of different forms must be understood as being about different complexes. In the case of relations, the idealist account of relations forces us to accept something like this principle. But that presupposes that one has antecedently accepted that account for relations, where the issue that is now before us is whether we have any reason to accept that account of relations. The argument has to be from the principle to the account and not conversely. And unfortunately, perhaps, Sprigge gives us no reason to accept that principle. His example therefore provides no basis for rejecting Russell’s view of relations and accepting that of the idealists. In fact, Sprigge accepts the point that Russell is making against Bradley and the idealists. This is the fact that relations are inseparable from order. Bradley and the idealists do not account for order. Russell takes it to be a basic feature of relations. When Sprigge takes the predication of the gestalt quality to apply to a structured whole he is accepting, as Bradley does not, that the relational predicate R, or the more complex predicate (FR-G) – in Sprigge’s example, the predicate “(is-) a-cup-on-a-saucer” – , presuppose an order. For he insists repeatedly that such non-relational predicates are predicated of wholes or totalities that are structured: they are, in other words, predicated of ordered pairs of individuals (or ordered n-tuples). If (F-R-G) [ = “a-cup-on-a-saucer”] were really a monadic predicate, say P and < a, b > were really an individual constant, say

263 i then (+) (F-R-G) (< a, b >) would be represented by Pi and no order would be evident, nothing that would indicate that either the quality or the totality were anything other than non-relational entities. But in fact Sprigge insists that the quality that is predicated is of a special sort, namely a gestalt quality. It is a quality which has an order evident in it. So long as the complex predicate that represents this property includes, implicitly at least – and that is all we need to make Russell's point – a reference to the relation between the individuals through having built into its own complexity the term that represents the relation, namely, R (“on”), then we have in fact presupposed in the term that represents this special quality the relation and the order upon which Russell insists. Sprigge’s complex predicates referring to the gestalt qualities thus presuppose, rather than account for, a relational order, precisely that order for which the idealist view of relations is unable to account. Moreover, Sprigge also insists that the whole of which the gestalt quality is predicated is also a special sort of entity, something which is a “whole” or a “totality,” something which is more than a mere set, or in other words, something more than a mere unordered set. Further, Sprigge, unlike the idealists, does not insist that the structured wholes are individuals over and above the related individuals, nor insist with the idealists that the related individuals somehow lose their individuality in the superindividual that constitutes their relatedness. The gestalt qualities and structured wholes or totalities to which Sprigge directs our attention thus presuppose a structure or ordering, and therefore presuppose the relation R that Russell insists is basic. Sprigge is thus allowing that (+) and (+++) express different judgments but are nonetheless both about the same objective state of affairs. Sprigge is thus accepting that the various forms of judgment that we use to describe the cup being on the saucer all presuppose the ordering of the cup and saucer induced by the relation that holds between them. In short, Sprigge is in effect agreeing with Russell that the issue of

264 order cannot be avoided. Moreover, Sprigge does not suggest that the cup becomes inseparable from the saucer once the relation between them is recognized. The whole consisting of the cup and saucer, the former on top of the latter, is itself distinguishable from other wholes, and, moreover, the cup is distinguishable from the saucer. But this is to say that things are after all not inseparable. That is, he recognizes that one can consistently affirm both that Fa and ~ R(a, b) or that one could consistently affirm that the cup a is on the saucer b while denying that the saucer is on the table c, even when it is in fact upon the table. In allowing that the parts that are related to each other are thus distinguishable from each other, and therefore separable, he is denying that his account of relations implies such inseparability. What this means is that Sprigge has accepted Moore’s basic claim against the idealists and against their account of relations, that when things are related they become inseparable parts of a single whole. Sprigge makes a good number of valuable, and valid points, in his defence of Bradley. He shows that Bradley is not simply the silly metaphysician that Ayer, for example, took him to be.41 Indeed, he indicates his acceptance of several idealist theses, becoming thereby a major exponent in the late twentieth century of the idealist positions developed some hundred years earlier. But at the same time, as we now see, he also takes for granted the fundamental criticism, based in the empiricist’s PA, that Moore offered of the idealist account of relations in his “Refutation of Idealism.” Moore, of course, was not the only one to attack idealism at the turn of the century. Another was Samuel Alexander.42 Like Moore, Alexander insisted against Berkeley that the relation between an awareness and the object of the awareness is not that of predication: “...the perceiving is but the act of the subject, and the perceived is not an element in the perceiving, but distinct from it and related to it in the single situation which is described as their togetherness” (p. 316). Or, as he put it elsewhere, “The experience, ‘I perceive a tree,’ is the statement that there are two existents, the percipience or consciousness and the different and therefore non-

265 mental thing the tree. The fact that these two things exist together is a statement precisely of the same character as that there are two existents, the tree and the grass in which it stands, the only difference being that one of the two things which are together is conscious.”43 This relation is a real relation; for, like idealists and like Moore and Russell he insisted that the monadistic account of relations would not do: “you cannot explain how a continuum can be generated from elements supposed to be discontinuous and completely independent” (“On Relations,” p. 311). We cannot “separate the world into terms and their relations”; rather, “The world consists of things in their relations” (p. 310). But contrary to the idealist account of relations, Alexander maintains that the fact that two things are related does not imply that those things cannot be distinguished in their being apart from their being so related: “...paternity is not external to a father but it is external to a man before he is a father. Thus a relation can be considered external to its terms if those terms are considered out of the relation, or rather if they can as in the case of paternity be so considered” (p. 310). Contrary to the idealists, entities are wholly self-contained. Commenting on Bosanquet’s formula that denies this, that “Mind is always a world; its objects are always fragments,”44 Alexander asserts the opposite,45 “Mind and its objects are both alike and in the same sense worlds or fragments. So far as the object is a fragment, mind is a fragment too” (p. 18). The idealist position, that things cease to be fragments only by ceasing, when in relation, to be themselves, depends upon ...a confusion between the notion of independence and that of absence of relation. It is clear that in the experience both object and subject enter into a relation, that of being known on the one hand and of knowing on the other. But dependence is a specific kind of relation. When A is dependent on B, we mean at least that without B, A would not have the qualities for which it is said to be dependent on B. Thus when we speak of a very dependent person, we mean that he can do nothing without the help of some one else. To be independent in any respect is for A to have these qualities in the absence of B. Now the object is clearly dependent on the mind for being known. But it is, as the mind itself declares in the experience, not dependent upon the mind for the qualities which make it what it is. The mind is indispensable to blue in so far as it is sensed, but not for the blueness (p. 24).

Alexander thus agrees with Moore that in the sensation of blue, blue is what it is, distinct from other things, including consciousness, even when it is in relation to consciousness. Alexander, moreover, agrees with Moore that this fact of distinctness is given to us in experience; it is something

266 which “the mind itself declares in experience.” Alexander, in other words, makes the same appeal to the empiricist’s Principle of Acquaintance (PA) as does Moore. What distinguishes Moore from Alexander is neither their philosophical position, nor the principles upon which they defend it, but rather simply the fact that Moore squarely places the argument from PA for the distinctness of things in the context of an attempt to refute idealism. It is primarily this, I suspect, that has given Moore’s version of the argument pride of place in our collective image of the history of philosophy. In any case, the argument that Moore and others offer against Berkeleyan idealism has not universally been accepted. There are, for example, those who hold what has been called an “adverbial” account of perception.46 On this account, the relation of blue to the experiencing of it is akin to the relation of the speed of a dance to the dance. We predicate “dancing slowly” of a person without implying a relation between the dancing and its speed akin to the relation between, say, a cup and the saucer upon which it is sitting. Rather, the adverb ‘slowly’ indicates that one is talking not about something external to the dancing but merely about a determinate form of dancing. Similarly, upon this account of perception, we predicate “experiencing blue” of a person not implying that there is a relation akin to that of “being upon” between the experiencing or the awareness and the blue. Instead, blue is simply a way of experiencing, and linguistically is an adverb applying to the verb. Just as we dance slowly so (we are told) we experience bluely. The virtue of this account is supposed to be that we can have perception without having to introduce sensations as separate particulars to which experiencing relates us. However, on this view, blue is a determinate form of the determinable experiencing. As such it is predicated of the person. That brings the position close to the view of Berkeley. The view is in fact quite implausible. Blue is a determinate form of colour, not a determinate form of awareness. The determinate forms of the latter are such things as perceiving, feeling, desiring, etc. But in any case, the point is that the philosophers who accept this adverbial account of perception have rejected the conclusion of the second half of Moore’s argument. But these philosophers have also accepted, as we saw Sprigge accept, the conclusion of the first half of Moore’s argument that things and their qualities are distinct from other things, they are what they are independently of other things and entities, including those entities to which they are related. In this case, it is the quality of, say, experiencing bluely that is distinct from and independent of other qualities. To this extent, contemporary philosophers

267 have largely accepted the conclusion of the first half of Moore’s argument in refutation of idealism. We started by wondering why it was that the first half of Moore’s argument tended to be overlooked by those discussing Moore’s “Refutation of Idealism.” What I am now suggesting is that Moore’s argument has become so thoroughly accepted by the philosophical community that we have tended to forget that once upon a time that argument really was needed. We have forgotten that there were those who denied Moore’s conclusion, and therefore have forgotten not only that the argument was needed but therefore also exactly what its force is. It has become a puzzle. When idealism is now to be defended this is to be done, as it is by Sprigge and by the defenders of the adverbial account of perception, by means that are compatible with Moore’s point, and, it would seem, with PA. That shows how much Moore’s paper marks a real turn away from the idealism of Bradley in the 20th century: whatever idealism we now have it is not that idealism, nor can it be based on the theory of relations that we find in Bradley and the other idealists against whom Moore was arguing.

** Esse may well be percipi, but that is hardly plausible unless one can somehow establish that it must be that way. Moore’s point – or, what is the same, Locke’s point – was that an appeal to PA argues against the inseparability of such connections, and against the idealist theory of relations that entailed such inseparability. This point and this principle seems by now to have securely precluded any possibility that we can conceive that Berkeley’s principle that esse is percipi can be made plausible by an account of relations that, on the one hand, makes relational states dependent upon consciousness, and, on the other hand, concludes that related entities become so inseparable as to be indistinguishable.

268 Endnotes to Study Six

1G. E. Moore, “The Refutation of Idealism,” in his Philosophical Studies (London: Routledge and Kegan Paul, 1922), pp. 1-30. 2D. Lewis, “Moore’s Realism,” in L. Addis and D. Lewis, Moore and Ryle: Two Ontologists (The Hague: Nijhoff, 1966), Ch. I. 3More recently, David Crossley, “Moore’s Refutation of Idealism: The Debate about Sensations,” Idealistic Studies, 24 (1994), pp. 1-20, also ignores the first part of Moore’s argument. John Skorupski, English-Language Philosophy 1750-1945 (Oxford: Oxford University Press, 1993), refers (p. 152) to Moore’s “Refutation of Idealism,” but mentions only the second part of Moore’s argument, where he insists upon the act-object distinction as a way of refuting Berkeley. This fits with Skorupski's reading of T. H. Green and the other idealists as beginning with a phenomenalism similar to that of Berkeley (p. 84), but in fact, as we shall see, the phenomenalism, or, what is the same, the idealism, is not a premise but a conclusion derived from the idealist account of relations. Skorupski does recognize the importance of the doctrine of relations (p. 84ff), but does not probe it deeply enough to see how the idealism is implied by that account (see note 15, below). Since the idealism of such philosophers as Bradley and Bosanquet depends upon the doctrine of relations, and since the first part of Moore’s argument in the “refutation” is directed at that account of relations, we can see why Skorupski overlooked the significance of the first part of Moore’s argument. More oddly, Alan White, G. E. Moore (Oxford: Blackwell, 1958), not only ignores the first part of Moore’s argument but the second part also: Moore’s essay on the “Refutation of Idealism” is not even discussed. Unfortunately, Lewis’ “Moore’s Realism” appeared too late for White to be able to make use of its insights. Referring to it, White might have become able to see the significance of Moore’s early arguments against the idealists, and their relevance for his (Moore’s) defense of realism. White might also have benefited from Lewis’ often very insightful account of Moore on perception. The notable exception to the rule that the first part of Moore’s argument is ignored is Thomas Baldwin, G. E. Moore (London: Routledge, 1990), who devotes a chapter to analyzing Moore on idealism, both the arguments of the “Refutation” and also those elsewhere in Moore’s work. We shall have something to say about Baldwin’s discussion below. 4G. Berkeley, Three Dialogues between Hylas and Philonous, in vol. 2 of the Works, ed. A. A. Luce and T. E. Jessop (London: Nelson, 1948-57).. 5G. Berkeley, Principles of Human Knowledge, in vol. 2 of the Works..

269

6E. B. Allaire, “Berkeley’s Idealism,” Theoria, 29 (1963), pp. 229-44; P. D. Cummins, “Perceptual Relativity and Ideas in the Mind,” Philosophy and Phenomenological Research, 24 (1963), pp. 204-14; R. A. Watson, “Berkeley in a Cartesian Context,” Revue Internationale de Philosophie, 65 (1963), pp. 381-94. For criticism see: R. Muehlmann, “Berkeley's Ontology and the Epistemology of Idealism,” Canadian Journal of Philosophy, 8 (1978), pp. 89-111; Nathan Oaklander, “The Inherence Interpretation of Berkeley: A Critique,” The Modern Schoolman, 54 (1977), p. 261-9; and Harry Bracken, “Some Problems of Substance among the Cartesians,” American Philosophical Quarterly, 1 (1964), pp. 129-37. Cummins changes his mind in “Berkeley’s Ideas of Sense,” Noûs, 9 (1975), pp. 55-72. See also Alan Hausman and David Hausman, “A New Approach to Berkeley’s Ideal Reality,” in R. Muehlmann, ed., Berkeley’s Metaphysics: Structural, Interpretative, and Critical Essays (College Park, PA: State University of Pennsylvania Press, 1995), pp. 47-66; and F. Wilson, “Comments on the Hausmans’ ‘New Approach to Berkeley’s Ideal Reality’,” in R. Muehlmann, ed., Brekeley’s Metaphysics, and also in this volume, above. 7.Baldwin (G. E. Moore, p. 19) objects that Moore has not refuted phenomenalism. But that was not Moore’s intention. Phenomenalism is not the same as idealism. The latter holds that there is an ontological dependence of sensations upon the mind; phenomenalism holds that sensations are causally dependent for their existence upon the state of our sense organs, and in this way causally dependent upon the perceiving of them. Hume was a phenomenalist but not an idealist; he is careful to argue that, although sensations depend for their existence upon our organs (D. Hume, A Treatise of Human Nature [ed. L. A. Selby-Bigge and P. H. Nidditch, 2nd edition (Oxford: Oxford University Press, 1978)], Bk. I, Part III, sec. 2, pp. 210-11), it is logically or ontologically possible for them to exist apart from minds (Treatise, I, III, 2, p. 207). (On this point see, F. Wilson, “Was Hume a Subjectivist?” Philosophy Research Archives, 14 (1989), pp. 247-82, and “Was Hume a Sceptic with regard to the Senses?” The Journal for the History of Philosophy, 27 (1989), pp. 49-73; and in greater detail, The External World and Our Knowledge of It: Hume’s Critical Realism: An Exposition and a Defence [Toronto: University of Toronto Press, forthcoming]). It is no doubt true that pains, if not sensations, exist depending upon the states of our bodily organs. Hume’s point is that it is still logically and ontologically possible for these entities to exist apart from any perceiving mind. Moore’s is the same. What Moore establishes, in other words, is that even if pains are causally dependent upon the perceiving of them, nonetheless, to experience a pain is already to be outside the circle of ideas, that is, outside the circle of things that are predicated of the mind and in that sense are ontologically dependent upon the mind. This establishes that idealism is false, but not that phenomenalism is false. But Moore’s point is to refute idealism, not phenomenalism. Douglas Lewis sees this quite clearly. Baldwin does not. So Baldwin ends up unfairly criticizing Moore’s argument on the grounds that it fails to do what it did not set out to do!

270 It should also be added that, while to get to the blue is to get one outside the circle of ideas, it does not necessarily get one to things in the sense of physical objects. This point has been made by others. Thus, Henry Baker proposed “to use the word ‘object’ ... to denote what is actually experienced by the individual subject, and to use the word ‘thing’ to denote the standardised object of ordinary speech” (“Can There Be Anything Obscure or Implicit in a Mental State?” I, Proceedings of the Aristotelian Society, 13 [1912], p. 258). Douglas Lewis, “Moore’s Realism,” is also clearly aware of this distinction. The point is, as Lewis emphasizes, that to refute idealism is not yet to secure realism. 8A. E. Taylor, “Mind and Nature,” Ethics (formerly The International Journal of Ethics), 6 (1902), pp. 55-86. 9A. E. Taylor, Elements of Metaphysics, second edition (London: Methuen, 1909). 10J. B. Watson, The Interpretation of Religious Experience, Part Second: Constructive, (Glasgow: James Maclehose and Sons, 1912). 11W. M. Urban, Valuation: Its Nature and Laws. (London: Swan Sonnenschein, 1909); and The Intelligible World: Metaphysics and Value (London: George Allen and Unwin, 1929).. 12Urban is here quoting with approval from A. S. Pringle-Pattison; he gives no reference. 13Bernard Bosanquet, The Distinction between Mind and Its Objects (Manchester: Manchester University Press, 1913). 14See also note 44, below. 15Thus, the phenomenalism, or idealism, is a consequence of the doctrine of relations, not a premise for the theory – contrary to what Skorupski, English-Language Philosophy 1750-1945, p. 84, suggests. See note 3, above. 16H. H. Joachim, The Nature of Truth (Oxford: Oxford University Press, 1906); B. Russell, Review of Joachim’s The Nature of Truth, Mind, n.s., 15 (1906), pp. 528-33. 17Russell, Review of Joachim, p. 531. 18G. E. Moore, “Mr. Joachim’s Nature of Truth,” Mind, n.s. 16 (1907), pp. 229-235. 19For a discussion of this principle, see F. Wilson, “Acquaintance, Ontology and Knowledge,” in the present volume, above; and “Effability, Ontology and Method,” and “Universals, Particulars and Tropes: The Role of a Principle of Acquaintance in

271 Ontology,” both in this volume, below. 20John Locke, Essay concerning Human Understanding, Fourth Edition, ed. A. C. Fraser (Oxford: Oxford University Press, 1894; reprinted New York: Dover, 1959). See also F. Wilson, “The Lockean Revolution in the Theory of Science,” in S. Tweyman and G. Moyal, eds., Early Modern Philosophy: Epistemology, Metaphysics and Politics (New York: Caravan Press, 1986)., pp. 65-97; and The Logic and Methodology of Science in Early Modern Thought (Toronto: University of Toronto Press, 1999). 21G. E. Moore, “The Nature of Judgment, ”Mind, n.s. 8 (1899), pp. 176-93. 22Moore does not add the further point that Russell makes, that even then we cannot make a true judgment. 23F. H. Bradley, The Principles of Logic, Second Edition(London: Oxford University Press, 1922), p. 3ff. 24See the remarks of Bosanquet, discussed above. 25Thomas Baldwin, G. E. Moore, p. 14, takes Bradley’s “content” and “part of the content” to be simply a collection of qualities, but has not explored the idealist account of relations sufficiently to note that when one speaks in the context of that account of relations of a quality which is a “common part” to two contents one cannot identify that quality independently of its necessary connections to other qualities. He thus fails, as he himself suggests, to understand the significance of Moore’s argument. 26Thomas Baldwin, G. E. Moore, pp. 13-15, misses this part of Moore’s argument. Again, the failure to provide a detailed account of the idealist theory of relations leads Baldwin to fail to recognize the thrust of Moore’s argument and how devastating Moore’s appeal to our ordinary experience of qualities is to that account of relations, and therefore to the idealism which rests upon that account. 27B. Russell, Principles of Mathematics, Second Edition (London: Allen and Unwin, 1937 [first edition 1903]). 28G. E. Moore, Principia Ethica (Cambridge: Cambridge University Press, 1903). 29Contrary to what Thomas Baldwin, G. E. Moore, p. 23, suggests. 30 Cf. Wilson, “Acquaintance, Ontology and Knowledge,” this volume, above. 31T. L. S. Sprigge, James and Bradley: American Truth and British Idealism (La Salle, Ill.: Open Court, 1993).

272

32Cf. M. Brodbeck, “Methodological Individualisms” in M. Brodbeck, ed., Readings in the Philosophy of the Social Sciences (New York: Macmillan, 1968); and also L. Addis, The Logic of Society (Minneapolis: University of Minnesota Press, 1975). 33Cf. F. Wilson, “John Stuart Mill on Psychology and the Moral Sciences,” in J. M. A. Skorupski, ed., Cambridge Companion to John Stuart Mill (Cambridge: Cambridge University Press), pp. 203-254. 34Cf. G. Bergmann’s notion of “process knowledge”; see his Philosophy of Science (Madison, WI: University of Wisconsin Press, 1956), Ch. 2. See also F. Wilson, Explanation, Causation and Deduction (Dordrecht, The Netherlands: D. Reidel, 1985), and The Logic and Methodology of Science and Pseudoscience (Toronto: Canadian Scholars Press, 2000). 35B. Russell, Principles of Mathematics, p. 221ff, p. 226, pp. 448-9. 36F. H. Bradley, Appearance and Reality, Second Edition (Oxford: Oxford University Press, 1897); and also “Relations,” in his Collected Essays, 2 vols. (London: Oxford University Press, 1935), vol. 2. Cf. also F. Wilson, “Bradley’s Contribution to Empiricism,” in this volume, above. 37Cf. Wilson, “Burgersdijck, Coleridge, Bradley, Russell, Bergmann, Hochberg: Six Philosophers on the Ontology of Relations,” in this volume, below. 38Russell, Principles of Mathematics, p. 221. 39F. H. Bradley, Principles of Logic, p. 28. 40B. Russell and A. N. Whitehead, Principia Mathematica, Second Edition, 3 vols. (Cambridge: Cambridge University Press, 1927), vol. 1, p. 200. 41Cf. Wilson, “Bradley’s Account of Relations and Its Impact on Empiricism,” in this volume, above. 42Samuel Alexander, “On Relations; and in Particular the Cognitive Relation,” Mind, n.s. 22 (1912), pp. 305-28. Nor is Alexander the end of the list; see also, for example, T. P. Nunn, “The Aims and Achievements of Scientific Method,” Proceedings of the Aristotelian Society, 6 (1905), pp. 141-82. 43Samuel Alexander, “On Sensations and Images,” Proceedings of the Aristotelian Society, 10 (1909), pp. 1-35, at p. 3.

273

44See footnote 14, above. 45Samuel Alexander, The Basis of Realism, British Academy Lecture, Jan. 28, 1914. 46Such a position was developed by C. J. Ducasse and has more recently been defended by R. Chisholm and W. Sellars. The relevance of this example was pointed out to me by Paul Coates. For discussion, see the discussion of Sellars in the Appendices to “Acquaintance, Ontology and Knowledge,” this volume, above.

Seven Burgersdijck, Coleridge, Bradley, Russell, Bergmann, Hochberg: Six Philosophers on the Ontology of Relations* The discussion of relations early in the 20th century was dominated by Russell and Bradley. They of course disagreed with each other, but at the same time they agreed that the tradition concerning the ontology of relations that had dominated the thought of their predecessors was wrong. The early modern philosophers took over the monadistic account of relations that had dominated philosophy since at least Aristotle and perhaps before.1 It is to be found, for example, in the well-known Monitio Logica of Franco Burgersdijck, first published in Leyden where Burgersdijck taught, and republished in his native Holland several times, as well as being published in England, including a translation of 1697.2 Bradley built on insights developed by Coleridge and offered as an alternative to this account what Russell called the monistic account of relations. Russell showed that Bradley’s view encountered some of the same difficulties that refuted the monadistic view, and more besides. He was led to propose yet another account to solve the various problems. This proposal has become much of a commonplace. Recently, however, Gustav Bergmann has criticized Russell and suggested yet another option. I shall argue, though, that neither is the criticism of Russell valid nor is the suggested alternative capable of withstanding philosophical criticism. Russell’s, I suggest, remains the only reasonable account of relations. Nor, I shall also argue, do we need the emendation of Russell offered by Hochberg. ‫٭‬

A shorter version originally appeared in The Modern Schoolman, 72 (1995), pp. 283-310.

276

1. Burgersdijck and the Logic of Relations: The Tradition Burgersdijck describes the category of relation in this way − which of course goes all the way back to the first chapter of Aristotle’s Categories: Those things are said to be related, which in Respect of what they are, are said to be others, that is, of others or in any other Manner or Respect are referred to another (Monitio Logica, p. 19).

According to Burgersdijck every relation involves two entities, that from which the relation originates and that in which it terminates. “In every Relation,” he tells us, “are required Subject and Term.... That is called the Subject to which the Relation is attributed; or that which is referred to some other thing.... That, the Term to which the Subject is referred” (p. 21). The subject is the relate and the term the correlate. Burgersdijck recognizes that every relation has a converse. Thus, Related and Correlated are mutually referred each to the other; and that not in one, but a two-fold Relation...In which Reciprocation, or mutual Relation, that which is the Subject of one Relation is the Term of the other; and so on the contrary.... As also that which is in one Relation the Relate, is in the other the Correlate; and so on the contrary (p. 21).

The example that Burgersdijck gives is that of father and son; where the father is referred to the son we have the relation of paternity, while where we have the son referred to the father we have the relation of filiation or son-ship.3 Burgersdijck is here making the important and correct point that a relation and its converse are two relations and not one. In recognizing this, Burgersdijck recognized the important point that relations intrinsically involve an order. Burgersdijck goes some way towards classifying different relations. Thus, he distinguishes those relations that are synonymous or of aequiparency, such as those of friend or rival, and those relations that are heteronymous or of disquiparancy, such as those of father and son and of master and servant. From the examples it is clear that Burgersdijck is distinguishing symmetrical from asymmetrical relations. Relations presuppose qualities in both the subject and the term. The quality may either be a property, that is, a quality had universally by

277 subjects of the relevant sort, or an accident, that is, a property that is not present universally. In the latter case Burgersdijck speaks of the relation having a foundation. “Some Relations are supposed, supposing the Subject and Term: Others besides these do require a Foundation.... And a Foundation is that by whose Means the Relation acrews to the Subject” (p. 22). He gives the following examples: When an Egg is said to be like an Egg, the Similitude between these two Eggs arises in each as soon as they begin to exist; nor is there any thing required towards their Relation, besides the Existence of two Eggs. But the Relation of Servant does not presently arise in the Subject so soon as the Term exists; but it behoves that something else also do intercede upon which this Relation is founded: For a Servant is therefore the Servant of one, because by him he has been either saved or purchased, &c. (p. 22).

The distinction between property and accident was to become increasingly unimportant, so for the sake of convenience let us speak of those qualities in a thing by which the relation “acrews to the subject” as the foundation. Now, in the basic instance the subjects and terms of relations are substances. The substances are related to each other by virtue of the foundations which are present in them. As the examples make clear, the qualities which are the foundations of the relation are themselves nonrelational. Moreover, each of the substances which are the subject and the term is, as Burgersdijck puts it, following the ancient formula, “a Being subsisting of it self, and subject to Accidents” (p. 8). Unlike accidents, substances do not depend for their existence on something else: “To subsist by it self is nothing else but not to be in any thing as a Subject...” (p. 8). Now, an accident is “a being inherent in a Substance”; it “cannot exist without a Subject”; nor can it “pass from one Subject to another” (p. 10). Thus, an accident cannot exist apart from the substance in which it is present. So substances can exist apart from other substances. If a substance ceases to exist, so do its accidents; but since substances subsist by themselves, a substance can exist unchanged, that is, remain the same in its being, even if other substances were to cease to exist. Since it is not predicated of other things, the non-existence of other things does not affect its existence or being. What is distinctive of relation, then, is not that it introduces an entity over and above substances and the qualities that are present in them as properties and accidents. What is important about relation is that by virtue of the qualities in them one substance is referred to another in a judgement.

278 Objectively, the subject and term are separable; objectively, there is no unity. The only unity consists in that contributed by the judgement in which the subject is referred to the term. We find this view of relations repeated in Locke, but in the context of a systematic account of the operations of the mind.4 Thus he tells us that The nature therefore of Relation, consists in the referring, or comparing two things, one to another; from which comparison, one or both comes to be denominated. And if either of those things be removed, or cease to be, the Relation ceases, and the Denomination consequent to it, though the other receive in it self no alteration at all, v.g. Cajus, whom I consider to say as a Father, ceases to be so to morrow, only by the death of his Son, without any alteration made in himself (Essay, II, xxv, 5, p. 321).

A little later Locke adds that ...there can be no Relation, but betwixt two Things, considered as two Things. There must always be in relation two Ideas, or Things, either in themselves really separate, or considered as distinct, and then a ground or occasion for their comparison (Essay, II, xxv, 6, p. 321).

Here we have the same doctrine of relations that one finds in Burgersdijck: relational judgments are about facts that are essentially non-relational, facts in which a non-relational quality is present in a substance, while the bringing together of these two into a unity of the subject and term is provided by the mind referring the one to the other in a mental act of comparison. Leibniz, in his commentary on Locke’s philosophy in the New Essays concerning Human Understanding agrees with this account5 (Bk. II, Ch. 25). Elsewhere, in a letter to des Bosses, he puts it this way: You will not, I believe, admit an accident which is in two subjects at once. Thus I hold, as regards relations, that paternity in David is one thing, and filiation in Solomon another, but the relation common to both is a merely mental thing, of which the modifications of singulars are the foundations.6

In his Fifth Letter to Clarke, he illustrates the monadistic account with a mathematical example, contrasting it with the view that he rejects: La raison ou proportion entre deux lignes L et M peut etre conçue de trois façons: comme raison du plus grand L au moindre M, comme raison du moindre M au plus grand L, et enfin comme quelque chose d’abstrait des deux,

279 c’est à dire comme la raison entre L et M, sans considerer lequel est l’anterieur ou le posterieur, le sujet ou l’objet.... Dans la premiere consideration, L le plus grand est le sujet; dans la seconde, M le moindre est le sujet de cet accident, que les philosophes appellent relation our rapport. Mais quel en sera le sujet dans le troisieme sens? On ne sauroit dire que tous les deux, and ensemble, soyent le sujet d'un tel accident, car ainsi nous aurions un Accident en deux sujets, qui auroit une jambe dans l’un et l’autre, ce qui est contre la notion des accidens. Dont il faut dire, que ce rappport dans ce troisieme sens est bien hors des sujets; mais que n’étant ny substance ny accident, cela doit etre une chose purement ideale, dont la consideration ne laisse pas d’etre utile.7

We can represent this account of relations as follows. Consider the relational statement that (@) a is R to b or, in symbols that Russell has made familiar to us, R(a,b) On the monadistic account, such a statement has a two-fold analysis. On the one hand, there are the objective truth conditions, the objective facts concerning a and b which determine whether the statement is (objectively) true or false. On the other hand, there is the subjective mental state that the use of the relational statement expresses. As for the former, the objective facts represented by (@) are non-relational: (#) a is r1 and b is r2 or, in symbols, r1(a) & r2(b) The non-relational properties r1 and r2 are the (objective) foundations of the relation. As for the subjective state that the use of (@) expresses, this is a judgement of comparison. Notice that, upon this account of relations, if one of the relata, say b, ceases to exist, so will the relation – objectively, together of course with the possibility of comparison – but this happens with no change in the other relatum: b’s ceasing to exist, or ceasing to be r2, is compatible with a continuing unchanged, and in particular with a remaining r1. On this view of relations, though the objects a and b are supposed to be related to each other, they are in fact ontologically and objectively independent of each other in the sense that if the one ceased to exist the other would continue to exist unchanged. This account was applied to all relations. Thus, for example, it was applied to resemblance. If Peter and Paul resemble each other in respect of being red, then the relation of resemblance is analyzed objectively into two non-relational facts, that of Peter being red and that of Paul being red, and

280 subjectively into the comparison of Peter and Paul in respect of colour.8 It was also applied to causation. Hume argued in detail that the judgment that A’s cause B’s is to be analyzed objectively, that is, in Hume’s terms, as a relation considered philosophically, into a regularity, a mere conjunction if you wish, between A and B, and subjectively, or, in Hume’s terms, as a relation considered naturally, into an association between (the impression or the idea of) A and (the idea of) B.9 Hume in fact applies it to all sorts of relations, including spatial and those of degree, e.g., of hot and cold.10 Hume was the most articulate of the defenders of the monadistic account of relations. When stated with Hume’s lucidity it gradually became clear that there were real problems, On the one hand, relations, if analyzed monadistically, were not really relations, understood as carriers of order − even though logicians such as Burgersdijck did recognize that relations implied order, asymmetrical relations at least. On the other hand, relations, understood monadistically, are not really relations: what is relational about them is analyzed away into unconnected non-relational facts. Hume’s clarity in effect made these problems visible. It was the English philosopher and poet, Samuel Taylor Coleridge, who first saw these problems. 2. Colerdige on Relations The publication by William Wordsworth and Coleeridge of their joint collection of poetry, their Lyrical Balads11, effected a revolution in poetry (though that was not known at the time). It also had a philosophical significance. In this latter, the central figure is Coleridge. He had been deeply influenced by David Hartley’s expansive associationist account of the human mind, his Observations on Man.12 In fact, Hartley’s picture of human being is not different in central features from that developed by Hume. But at the same time, there is material that would be uncongenial to Hume. Specifically, there is the material, occupying in fact almost the whole of volume two of the observations. This where Hartley gives an associationist account of our love of God, where this God is much of a piece with the God of the Newtonians: a wise and powerful constructor of the beautiful mechanisms of the universe, and clearly benevolent, but otherwise distant and hardly knowable, a person but only vaguely so. How the near-mysticism is to be fit into the semimechanistic associationism is never made clear: putting the parts of his philosophy – the Hume-like associationism and a sort of neo-platonic

281 mysticism – is never made clear: Hartley seems so to like the parts that they are never forged into a whole.13 Coleridge was never really distant from his God. But at the time of Lyrical Ballads, the vision of God was embedded in the Hartleian view of human being and of human reason. It was this associationist account of human being and human reason that Coleridge came to reject. To be sure, Coleridge was not to reject that account as wholly false. It does describe correctly the lower aspects of human being, where humans are passive. It is the pattern of what Coleridge calls the understanding, and is useful for getting along in the world of ordinary life. But there is a deeper, or, perhaps, higher aspect to human reason, Coleridge came later to argue, that moves us away from the realm of ordinary experience towards the higher realms that provide the unity and structure of ordinary reality. The ordinary person, he came to argue, “from the more imperfect development of his faculties, and from the lower state of their cultivation, aims almost solely to convey insulated facts, either those of his scanty experience or his traditional belief; while the educated man chiefly seeks to discover and express those connections of things, or those relative bearings of fact to fact, from which some more or less general law is deducible.”14 He made the same point in his Treatise on Method. Here he asserts that All things, in us, and about us, are a Chaos ...: and so long as the mind is entirely passive, so long as there is an habitual submission of the Understanding to mere events and images, as such, without any attempt to classify and arrange them, so long the Chaos must continue.15

The world of Locke and Hume, and, of course, of Hartley, is such a Chaos – that is, Chaos, given Coleridge’s premises: it contains no relations. Or rather, their view of the world is one in which there are no relations and their (associationist) view of the human mind is one in which only isolated events are grasped. But there is structure. In fact, there is structure in a sense in which the existence of structure is inconsistent with the traditional monadistic analysis. Hartley saw as much in his (Newtonian) vision of God as a supreme infinite being who actively orders and structures the finite entities in the world. But that vision cannot, as Hume saw, be accommodated within the associationist and empiricist philosophy. As Coleridge was to state in his Logic, upon Hume’s view, “experience itself is the sole ground in every instance of the legitimacy or objective truth of the conception.” Coleridge rightly infers from this, as Hume had inferred, that

282

In all reasoning therefore the materials of which or subject-matter are not drawn from experience, the notion of cause and effect is not only inapplicable but without meaning.16

Hartley’s notion that one can infer from finite things the infinite power of God simply will not hold up. Hume’s philosophy thus becomes “a means of undermining the foundations of our religious convictions or principles...”17 Coleridge made the same point in one of his letters, where he refers to “Hume’s system of Causation – or rather of non-causation”, and states that This is the pillar, & confessedly, the sole pillar, of modern Atheism.18

Now, Coleridge is correct about Hume’s system: there are no relations. Hume holds that the sensible entities of which we are aware are all independent of one another, in this sense that if a and b are two such entities, then a could continue to exist unchanged even if b ceased to exist. This is what he means when he argues that entities that are distinct are separable. But now suppose that a and b are related: a is R to b Then a has the property of being R to b. But in that case, if b ceases to exist then a ceases to have the property of being R to b. If Jack is kicking the football, and the football ceases to exist, then Jack ceases to have the property of kicking the football. Hence, given that the sensible entities are independent, they stand in no relations to another. As Coleridge put it at one point, “How opposite to nature and to fact to talk of the ‘one moment’ of Hume, of our whole being an aggregate of successive single sensations! Whoever felt a single sensation? Is not every one at the same moment conscious that there co-exist a thousand others, a darker shade or less light, even as when I fix my attention on a white house or a grey bare hill or rather long ridge that runs out of sight each way...”.19 The world is indeed a seeming chaos upon the Humean – and Hartleian – view, a mere aggregate of things, a pile without shape or form, or if it does have a shape or form then that shape or form is purely accidental, one of which it is more proper to say that the world is imagined to have than to say that it does have.20 It is clear that this view simply will not do. Structure is there, objectively in the world, and in our ontology we must account for that structure. This is the point that Coleridge is insisting upon against the Hume-Hartley view of things. In order to grasp the way things are we must

283 “contemplate, not things only, but likewise relations of things....”21 But this, Coleridge argues, shows not only the insufficiency of the HumeHartley view of the world but also the insufficiency of their associationist account of the human mind. ...that which unites, and makes many things one in the Mind of Man, must be an act of the Mind itself, a manifestation of intellect, and not a spontaneous and uncertain production of circumstances.22

He states that the “relations of things” ...are of two sorts, according as they present themselves to the Human Mind as necessary, or merely as the result of observation....The relations of things cannot be united by accident: they are united by an Idea either definite or instinctive. Their union, in proportion as it is clear, is also progressive.23

That which unites, however, is not simply the human mind; it is an objective Law or Idea. When [the relation of things] is absolutely necessary, we have called it a relation of Law; and as by Law we mean the laying down the rule, so the rule laid down we call, in the ancient and proper sense of the word, an Idea.24

Hume had recognized that his argument that power could not be attributed to God could not be sustained if it was allowed that there are innate ideas, and in particular, innate ideas of the powers of things. Coleridge notes this basic premise of Hume’s case; Hume was a follower of Locke and someone who “no less than [Locke] had exploded the doctrine of innate ideas….”25 Coleridge therefore argues contrary to this that the Ideas that unite things are innate. He agrees with Hume: “If in ourselves there be no such faculties as those of ... the scientific reason, we must either have an innate idea of them, which would overthrow the whole [Hartley-Hume] system; or we can have no idea at all.”26 Locke, of course, had argued that if ideas are innate then we must always have a clear and distinct knowledge of them; since we do not have such knowledge of the ideas claimed to be innate, Locke concluded we have no innate ideas. Coleridge replies that innate ideas need not always be fully present to our consciousness. As he puts it, “The Idea may exist in a clear, distinct, definite form, ... or it may be a mere instinct, a vague appetency towards something which the Mind incessantly hunts for, but cannot find....”27 In the Biographia Literaria he argues that the unity that the mind

284 seeks and finds has its ultimate source in the simple unity of God. We are to seek ... for some absolute truth capable of communicating to other positions a certainty, which it has not itself borrowed; a truth self-grounded, unconditional and known by its own light. In short, we have to find a somewhat which is simply because it is.28

There can only be one such principle. If there were two, then they would be related to each other, and that relation would itself demand a higher unity that encompasses them. ...there can be but one such principle...; for were there two or more, each must refer to some other, by which its equality is affirmed; consequently neither would be self-established, as the hypothesis demands.29

This principle is not a thing. Nor is it simply a subjective consciousness. For subject and object are correlative; they are related to one another. Consequently, the required principle “must be found in that which is neither subject nor object exclusively, but which is the identity of both.”30 This principle, as so characterised, manifests itself in the SUM or I AM; which I shall hereafter indiscriminately express by the words spirit, self, and selfconsciousness.31

We become aware of these higher realms in our own activity and in our feeling the activity of others, including in the end the activity of the great I AM. This is the realm not of the understanding but of Reason,32 where we actively grasp the reasons for things, the structuring activities of the source of all unity in the universe. What Coleridge had realized was what Hume had realized, that if one begins with the associationist account of the world and of human being, then there is no way in which we can form the idea of the active exercise of an objective power, and that if we cannot do this then we cannot conceive of God as an infinite power, In effect, what Coleridge came to recognize was that if we accept volume one of Hartley’s Observations on Man, then we cannot get from there to volume two. The associationist account of human reason is one that does restrict one to the realm of the senses, and if there is a realm beyond that, as volume two of the Observations maintains, then we need to revise the account of human being and of human reason that one finds in volume one. What Coleridge came to recognize was that if he was to maintain the metaphysics of volume two then it was volume one

285 that had to go. Things are in fact not separate: this view of the world is false. Things are in fact united with other things: it is their being unified that is their truth. Since a is R to b it is part of the truth of a that it is R to b: the being of b is involved inescapably in the being of a. Thus, to grasp the truth about a and b one must grasp the unifying relation. One must go beyond a and b as isolated individual things to the unifier that ties them together. By sense one grasps a and b as separate. But that is not to grasp the full truth about them, what they are in reality. To do that one must rise above sense experience and the understanding to Reason, the active consciousness that can grasp the unifying and substantive relation that gives to a and b that connectedness and structured being that forever eludes sense. Coleridge’s aim was to provide the metaphysics that could replace the Hume-Hartley view of the world and of the human mind which occupied volume one of the Observations and which could support the view of God as the infinite power animating and structuring the universe that occupied volume two. It is this which his doctrine of relations as constituted by Ideas is supposed to do. On this view, the relational fact a is R to b is a substantial whole, the activity of which is productive of the pair (a, b).33 R is the manner of this production.34 a and b we may grasp in sensory experience. But R is not given in sense. This form or Idea that does the structuring is grasped rather by Reason. Since R is an activity, and since the Human Mind, too, is activity, for one to grasp R is for consciousness to discover itself in things. In fact, knowledge of R as a form of activity is innate in the active consciousness, the Reason, of the human mind. There is not just one such relational whole, however; there are many. This substantial whole consisting of a being R to b is itself related to other such wholes. In fact, everything is, ultimately, related to everything else. The Idea or form that provides the ultimate unity, tying everything together into a single whole, is of course God, the great I AM. Upon the view of Hume and Hartley, or at least the Hartley of volume one, this great unifying principle is forever inaccessible to us. It is mysterious precisely because it cannot be grasped by reason, that is, the reason of the empiricists and the associationists. Upon Coleridge’s metaphysics of relations and his account of our knowledge of such relations the ultimate unifying principle still remains mysterious. But it is

286 no longer inaccessible. The mind can grasp relations, and can grasp ever larger unifying structures. The knowledge of these structures, the capacity to recognize them in things, is innate in the human mind. This innate knowledge of structures extends to the knowledge of the ultimate structure, the great I AM, the grasping of which is the ultimate end of all attempts at coming to know the world and ourselves in the world. The relation of LAW in its absolute perfection is conceivable only of GOD, that Supreme Light, and Living Law ... But yet the Human Mind is capable of viewing some relations of things as necessarily existent; that is to say, as predetermined by a truth in the Mind itself, pregnant with the consequence of other truths in an indefinite progression.35

The ultimate principle, however, is an absolute unity. Propositional thought involves a subject and a predicate, and this in turn involves a distinction. But where there is a distinction, there is no unity. Unity requires a subject and predicate where there is an identity, where these are not distinct. Thus, our knowledge of the ultimate unifying principle is not to be expressed in propositions of the ordinary sort; it cannot be the content of assertions. All one can do is name the ultimate principle; one can identify it, but one cannot say or assert anything about it. It simply IS. Moreover, the knowledge of this great I AM is direct, as direct as any knowledge could possibly be. In such knowledge the relation of subject to object is itself dissolved in a higher unity. In knowing the great I AM one becomes that I AM, identical with it, or, rather, losing one’s identity in unity with it – that is, one’s identity as a finite thing, one’s shallow and incomplete identity as a finite thing. The part of the human mind explored by Hartley was denoted by Coleridge as the understanding. This he contrasted to Reason, which is the capacity to grasp the reasons of things, the forms or powers that structure the world, and, of course, ultimately to grasp God as the infinite power that structures all lesser things. It is through the understanding that we reflect on the experience of sense and generalize,36 while Reason, he tells us, “is the power of universal and necessary convictions, the source and substance of truth above sense, and having their evidence in themselves.”37 Where the understanding is discursive, Reason is fixed; whereas the understanding in all its judgments refers to some other faculty as its ultimate authority, Reason in all its decisions appeals to itself as the ground and substance of their truth; whereas the understanding is the faculty of reflection, Reason is the faculty of contemplation.38 Reason as a faculty of the human mind is a

287 grasping of the intelligible Reasons for things, the active forms that provide the relational structures of things. These are present to consciousness in themselves as the objects of sense are present to consciousness in sense experience. In fact, there is a sense in which Reason is much nearer to sense than to the understanding: “for Reason ... is a direct aspect of truth, an inward beholding, having a similar relation to the intelligible or spiritual, as sense has to the material or phenomenal.”39 And as he noted in the Logic, the ‘is’ by which we express at once the connections of things and our knowledge of the connections of things represents “the perfect identity of being and knowing and therefore the ground and source of both.”40 3. Bradley and Russell against the Tradition With regard to the monadistic account of relations to which Coleridge objected, Bertrand Russell was later to raise a fatal objection to that traditional account of relations. Referring to the Leibnizian discussion, quoted above,41 he points out that what we have is not just L is greater Rather, what we have is L is (greater than M); That is, what we are concerned with is not simply the fact that L is greater but the fact that L is greater than a certain definite other individual, namely, M. Objectively in the fact with which we are concerned there is a reference from one individual to a second. This reference is not something that is simply added by the mind as it compares the two individuals in their state of being. In fact, since M is also presumably greater – i.e., greater than something else – we also have M is greater so that there is nothing in the being of either L or M to distinguish the one from the other and to ensure that there is a reference from the former to the latter. In terms of our symbols, what Russell is saying is that it won’t do to rely upon r1(a); what we need is (!) a is (R to b) As Russell puts it, “The supposed adjective of L involves some reference to M; but what can be meant by a reference the theory leaves unintelligible.”42 The theory does of course allude to mental acts of

288 comparison; the point is that objectively there is nothing in the relational state of affairs (#) a is r1 and b is r2 that would imply that there is a “reference” from one relatum to the other. But once we add something, as in (!), which makes such a reference, then it is no longer true that one relatum can cease to exist while the other remains unchanged. For, if b ceases to exist, then so does the property (R to b) If we make the reference objective, then, contrary to what the monadistic theory requires, the relata cannot exist independently of each other: if one ceases to exist, then that changes the being of the other. Bradley makes the much same case as Russell. He argues that there are relational wholes in which the relata are not independent in the way required by the monadistic account. He argues that genuine relations are incompatible with the independence that is a consequence of monadistic view. “... [A] mode of togetherness such as we can verify in feeling,” he tells us, “destroys the independence of our reals.”43 Conversely, if we do make the relata independent or absolute, then we destroy their relatedness: “Relations are unmeaning except within and on the basis of a substantial whole, and related terms, if made absolute, are forthwith destroyed” (A & R, p. 125). Bradley’s point is that, if (@) is genuinely relational, then what is predicated of a is the property “R to b”: (@') a is (R to b) and what is predicated of b is the property “R-ed by a”: (@") b is (R-ed by a) This being so, if a ceases to exist, then perforce b ceases to have the property of being “R-ed by a”. That is, if a ceases to exist then, when one has genuine relations, it is not the case that b remains unchanged. In this sense, genuine relations always affect the being of things. As Bradley puts it, “If the relations in which the reals somehow stand are viewed as essential, that, as soon as we understand it, involves at once the internal relativity of the reals. And any attempt to maintain the relations as merely external must fail” (A & R, p. 125). Thus, if (@) is genuinely relational then a and b are not independent as required by the monadistic account of relations. Nor will it do for the empiricist to suggest that he or she does insist upon there being a connection between the relata, save that it is subjective rather than objective. What is needed is an objective relation. As Bradley puts this

289 point, ...to understand a complex AB, I must begin with A or B. And beginning, say, with A, if I then find B, I have either lost A or I have got beside A something else, and in neither case have I understood. For my intellect cannot simply unite a diversity, nor has it in itself any form or way of togetherness, and you gain nothing if beside A and B you offer me their conjunction in fact (A & R, p. 509).

So much the worse, then, for the monadistic account of relations: since the empiricist account of relations makes it unintelligible how any two things can be genuinely related, that position is just wrong. 4. Russell’s Second Argument against the Tradition Bradley and Russell agree that there are genuine relations and that if there are such relations then their relata are not independent of each other. They further agree that this is a conclusive argument against the traditional monadistic account. But Russell has a second argument. We have for the relation of greater than L is greater and also presumably M is less At least, this is what the monadistic account requires when it asserts that there is no objective reference from L to M. In that case there is then no specific difference between L and M. Yet we have to have asymmetrical relations, as Burgersdijck indicated we must have when he distinguished between relations of aequiparency, such as those of friend or rival, i.e., symmetrical relations, and relations of disquiparancy, such as those of father and son and of master and servant, i.e., asymmetrical relations. If there are no specific differences in L and M to ground the relation, then one will have a symmetry between L and M. This will exclude the possibility of asymmetrical relations. Thus, if we admit that there are facts in the world involving asymmetrical relations then we must, on the monadistic account, conclude that these relations are after all symmetrical. But if there are instances of a relation and we hold that it is both asymmetrical and symmetrical, then we have arrived at a contradiction. As Russell puts it, Thus we should be forced, in all cases of asymmetrical relations, to admit a

290 specific difference between the related terms, although no analysis of either singly will reveal any relevant property which it possesses and the other lacks. For the monadistic account of relations, this constitutes a contradiction; and it is a contradiction which condemns the theory from which it springs (Pr. of Math., pp. 222-3).

The basic point that Russell wishes to emphasize against the monadistic account of relations is that relations not only imply that their relata are not in their being independent of each other but more strongly imply that there is an order in that connection. When a relation relates two individuals, it does so with, as Russell says, a sense: the relation goes from one relatum to the other. ...it is characteristic of a relation of two terms that it proceeds, so to speak, from one to the other. This is what may be called the sense of the relation, and is...the source of order and series. It must be held as an axiom that aRb implies and is implied by a relational proposition bR'a, in which the relation R' proceeds from b to a, and may or may not be the same relation as R. But even when aRb implies and is implied by bRa, it must be strictly maintained that these are different propositions....The sense of a relation is a fundamental notion, which is not capable of definition (Pr. of Math., pp. 95-6; italics added).

It is easy to show that the monadistic account is incapable of capturing this element of order.44 The two propositions (p1) (x)(Эy)(Rxy) and (p2) (Эy)(x)(Rxy) are not logically equivalent. (p2) entails (p1), but not conversely. For, if ‘R’ is taken to be loves, then (p1) asserts that for every thing there is something that it loves while (p2) asserts that there is a thing such that every thing loves it It is clear that if there is at least one thing that everyone loves, then everything does love something; but if every thing loves something or other then one person can love one thing and another another thing, meaning that it can still be false that there is at least one thing which everyone loves. In contrast, consider (q1) (x)(Эy)[r1(x) & r2(y)] and (q2) (Эy)(x)[r1(x) & r2(y)]

291 into which (p1) and (p2) respectively must be analyzed upon the monadistic account of relations. It is a simple matter to prove that these two propositions are in fact logically equivalent. Since (p1) and (p2) are not logically equivalent where their supposed analyses (q1) and (q2) are, it follows that the monadistic account of relations is no account at all: it cannot capture what must be captured if one is to have an ontology that includes relations as well as accidents, that is, relations as well as monadic properties. The element of order which is inescapably introduced by genuine relations inevitably escapes the monadistic account of relations, and condemns the latter as inadequate. As it turns out, the second argument of Russell against the traditional monadistic account of relations is equally telling against the account of relations that Bradley develops as an alternative to the traditional account. 5. Bradley’s Alternative to the Tradition Bradley joins Russell in rejecting the tradition, that is, the long tradition that goes back to Plato that construes relations monadistically. He goes on to offer an alternative account of relations. He appeals in effect to something like the empiricist’s Principle of Acquaintance (PA) in order to justify the claim that genuine relations exist. “Relational experience ... is ... unavoidable and is fully justified in its own place as a way of life and knowledge”.45 We do experience genuine structure, genuine relations, and these are there objectively, within the complex entity that is presented to us, not just in our thought about it; and we therefore need an ontology of relations which can, unlike that of the tradition, account for the existence of such genuine relational structures. His account of relations derives from that Coleridge, though Bradley’s is worked out more fully: Coleridge, for all his genuine philosophical insight on these issues, remains something of an amateur; not so Bradley, who is never an amateur. But in spite of his own position on relations being more philosophically sophisticated than that of Coleridge, nonetheless, it is the monism of Coleridge’s account of relations that in the end is what attracts Bradley. Bradley proposes such an alternative to the monadistic account of relations, one designed to meet its defects. There are in fact two conditions which it must fulfil. In the first place, this account must, unlike the monadistic, allow that there are relations that are genuine in the sense that their relata are not independent, or, equivalently, in the sense that the being of one relatum is not separable from the being of the other relatum.

292

If it [a relation] is to be real, it must be so at the expense of the terms, or, at least, must be something which appears in them or to which they belong. A relation between A and B implies really a substantial foundation within them. This foundation, if we say that A is like to B, is the identity X which holds these differences together. And so with space and time − everywhere there must be a whole embracing what is related, or there would be no differences and no relation. It seems as if a reality possessed differences A and B, incompatible with one another and also with itself. And so in order, without contradiction, to retain its various properties, this whole consents to wear the form of relation between them (A & R, p. 18).

And this alternative account of relations must, second, construe the relation as a particular. While Bradley rejected, on grounds that we have seen, the monadistic account of relations that was accepted by the empiricists, he nonetheless continued to accept the nominalistic assumptions of the empiricist tradition. The traditional account of the use of words, still present in Locke, depends upon a particular account of abstract ideas. On this account, things resemble each other by virtue of the properties that are present in, or characterize those things. Two particular things, two pieces of candy, say, or a piece of candy and a mental image of such a piece of candy, resemble each other in, for example, being red by virtue of the property red that characterizes both particulars. Abstract ideas consist of properties that have in the mind become separated from the things they characterize by a process of abstraction. The idea in the mind is the same as the property that characterizes various things in the world. For example, the abstract idea of red is the property red separated from any particular thing. The various things which are red, that is, which are characterized by the property red, are said to fall under the abstract idea of red. Words become general by virtue of becoming associated with abstract ideas; through such an association, words apply to, or mean, any thing that falls under the abstract idea, i.e., is characterized by the property which the abstract idea is. Berkeley, in the “Introduction” to his Principles of Human Knowledge,46 attacked this account of abstract ideas. It required the affirmation of three propositions which are in fact, when taken together, inconsistent: i) Whatever exists must be particular ii) What is conceivable (possible in thought) is possible (in reality) iii) Abstract ideas are created by separating properties from

293 particular things If abstract ideas are properties separate from things, then their separateness is possible in thought; but then their separateness is possible in reality; in which case it is possible that not every existent is particular. Philosophers such as Descartes and Locke, according to Berkeley, and quite correctly, accept all three propositions. Something must go. Berkeley, and his empiricist successors, accept the first two, agreeing here with their predecessors, but, in disagreement with them, reject the third. The sameness of things does not consist in the presence in them of a separable property. Rather, Berkeley insists, ...universality, so far as I can comprehend, not consisting in the absolutely positive nature or conception of anything, but in the relation it bears to the particulars signified or represented by it; by virtue whereof it is that things, names or notions, being in their own nature particular, are rendered universal (Principles, Intro., p. 15).

He uses the example (p. 16) of a mathematical demonstration in which a triangle drawn on paper is used to represent indifferently all triangles. It does this by virtue of the relations of resemblance in which it stands to all other triangles. Not only the things but the characteristics of things are particular: they are as particular as the things they characterize. What is given in sense, things and the characteristics of things, are localized in space and time. Generality appears through relations of resemblance. This is the nominalism of Berkeley and Hume. Bradley47 accepts this nominalism; “Nothing in the end is real but the individual...” (“Relations,” p. 663). He applies it in particular to relations: his account of relations, he lays it down as a metaphysical principle, must fulfil the condition of construing them as particulars. “A relation, to be experienced and to be actual, must be more than a mere abstraction. It must be an individual or particular fact, and, if less than this, it cannot be taken as itself” (“Relations,” pp. 635-6). Bradley concludes against the monadistic account that “there must be a whole embracing what is related”; the first of the conditions mentioned above requires that the parts be inseparable in this whole. Thus, “Your individual term is an abstraction always. It implies what we may call a selection from the concrete fact of the whole and entire experience” (“Relations,” p. 646). The second of the mentioned conditions requires furthermore that this relational whole must be itself a substantial particular. From these two conditions Bradley infers his own account of the nature of

294 this whole: The relata A and B are different things within a whole (A, B), which itself is an individual and which has those two terms as aspects. This whole then “consents to wear the form of a relation”; thus, if A and B stand in the relation R, then the correct representation of this fact consists in attributing a property corresponding to R, − a monadic property −, say r, to (A, B). Thus, according to Bradley’s account, the correct way to represent the fact reported by (@) a is R to b is given by (*) (a, b) is r or, r(a, b) where the whole (a, b) is itself a particular thing, of which the two terms a and b are but aspects, and where the arrangement r characterizes the particular or individual thing which the whole is. But this whole consists of the relata as parts. Thus, the relation holds of the relata, not separately as in the monadistic account, but jointly:48 “where the whole, relaxing its unity, takes the form of an arrangement, there is co-existence with concord” (A & R, p. 19). This is the monistic account of relations, as Russell called it. 6. Russell’s Criticism of Bradley Bradley applies his monistic account to all relations. He mentions something being like something else, which is to say, he applies it to resemblance. He applies it to space and time. He applies it to quality orders (A & R, p. 19), and to causation (A & R, p. 46ff). He applies it to contrariety among qualities, so that the general facts of the sort (!) (x)[Fx e ~Gx] which record such contrariety become relational facts among qualities (A & R, p. 19). He applies the account to all predication (A & R, pp. 16-17). At each stage newer, and more embracing unities appear to support the relations that structure the world. In the end all the apparently separable particulars, and all the apparently separable qualities or properties of things that we are aware of in ordinary experience disappear into one allembracing unity. The ultimately real thing is one substantiality (A & R, p. 124-6). This is the Absolute (A & R, p. 151) – Coleridge’s the great I AM. Thus, Russell’s characterization of Bradley’s account of relations as “monistic” is more than apt. Not only do relations characterize unities, but in the end it requires all reality to be a unity of seamlessly inseparable

295 parts. We should also note that Bradley’s account of relations introduces particulars that are not given in ordinary experience. In experiencing, say, a being to the left of b, we experience two particular things, a and b. We also experience the unity of a being to the left of b, the sort of unity represented by (@). But Bradley asserts that there is a further entity, a further particular thing, present in the situation. That is the whole (a, b) that is the subject of the arrangement. This further whole is not given in ordinary experience. It is, rather, something discovered by the faculty that Bradley calls “thought” − Coleridge’s Reason as opposed to the understanding. The object of thought is not something given in sense experience: “That [the object of thought] is not mere sense-experience should be a commonplace.”49 Instead, “judgement, on our view, transcends and must transcend that immediate unity of feeling upon which it cannot cease to depend.”50 Thought “grows from, and still it consists in, processes not dependent on itself. And the result may be summed up thus; certainly all relations are ideal, and as certainly not all relations are the product of thinking” (A & R, p. 426). And since relations are in effect not known by sense experience our knowledge of them is a priori; and since they are all ideal, transcending the entities of sense experience, the hold necessarily. In short, Bradley’s monistic account of relations requires the introduction of transcendent entities and the creation of a species of cognition through which we are supposed to be aware of these entities. And since these entities are necessary and are known a priori, they in effect re-introduce the necessary connections against which Hume and the empiricists argued. But Bradley’s case for necessary connections goes beyond dogmatism: it is based on his account of relations which responds to real difficulties in the monadistic account that was accepted by the empiricists. This account of relations was designed to solve problems with which the monadistic account could not cope. One of them was the fact that genuine relations imply that the relata are inseparable in their being from each other. This problem was successfully solved by the monistic account. But there was another problem. This was the problem of order. As we saw, Burgersdijck recognized the difference between symmetrical and asymmetrical relations. Any account of relations must not collapse this distinction. It was Russell's objection to the monadistic theory that it did precisely this. As it turns out, the monistic account also fails to pass this test. Thus, though it solves one problem faced by the monadistic account, it

296 does not solve this other problem. On the monistic view, relational states of affairs consist of a relational property being predicated of a complex individual whole (a, b). The relation of a to the whole (a, b) is the same as the relation of b to that whole. That is, the role of a in that whole is symmetrical with the role of b. Thus, (*) r(a, b) represents indifferently both the fact that (@) a is R to b and its converse, the fact that b is R to a (@c) As Russell says, using greater than as illustrating R, the whole (a, b), that individual of which the relation is predicated, “is symmetrical with regard to a and b and thus the property of the whole will be exactly the same in the case where a is greater than b as in the case where b is greater than” (Pr. of Math., p. 225). Where R is a symmetrical relation, one for which we have (x)(y)[R(x, y) e R(y, x)] then we have no problem: if (@) obtains so does (@c), and (*) can represent the two indifferently. But the same does not hold for asymmetrical relations (Pr. of Math., p. 221). Where R is asymmetrical, we have (+) (x)(y)[R(x, y) e ~R(y, x)] In this case (@) obtains while (@c) does not. “Thus,” Russell concludes, “the distinction of sense, i.e. the distinction between an asymmetrical relation and its converse, is one which the monistic theory of relations is wholly unable to explain” (Pr. of Math., p. 225). In the case of an asymmetrical relation, there is a difference – an ontological difference – between a relational fact and its converse that is not captured in any account, like the monistic, that requires both facts to be represented indifferently by the same notation. But the monistic account was introduced to solve the problem of relations. Since it cannot do that, it must be rejected. Russell offers two further criticisms of Bradley’s view of relations. It is worth noting them for later reference, when we discover that similar objections can be raised against the new ontology of relations has been proposed by Bergmann. One. Upon the monistic account of relations, as we saw, all relational statements require the introduction of a new particular not given

297 in ordinary experience, and since all particulars are in fact related to each other it follows that any relational statement can only be, ultimately, about a single transcendent particular, a single individual whole, the Absolute, that contains all reality within itself. Hence, in order to know any truth, one must grasp this individual totality, the Absolute itself. But we are finite; it is impossible for us to fully grasp this one subject of predication. It is worse than this, however. Bradley applies his account of relations to predication itself. The judgement that c is H, i.e., that c is qualified by H, becomes a judgment in which the property of “qualifying” is now predicated of the whole (c, H). But this leads to an infinite regress, one which is vicious in the sense that in order to make one judgment one must make an infinite number of further judgments. To avoid such a regress, one must insist in the end that one arrives at a whole in which the difference between it and its predicates disappears. But any judgment, any proposition, involves a distinction between subject(s) and predicate. So, on Bradley’s view, no judgment about the Absolute can ever be wholly true. In fact, any judgment must be self-contradictory. For, on the one hand, it purports to be true. On the other hand, it purports to ultimately be about reality, that is, the Absolute since it is the only reality, and this reality is such that it must be false. Any judgment must therefore claim of itself to be both true and false. This, moreover, must be equally true of the judgment that no statement about the Absolute – Coleridge’s great I AM – is ever fully true: this statement of Bradley’s view is itself contradictory upon that view. This, surely, is to condemn it. As Russell puts it, ...we find monists driven to the view that the only true whole, the Absolute, has no parts at all, and that no propositions in regard to it or anything else are quite true – a view which, in the mere statement, unavoidably contradicts itself. And surely an opinion which holds all propositions to be in the end selfcontradictory is sufficiently condemned by the fact that, if it be accepted, it also must be self-contradictory (Pr. of Math., p. 226).51

Two. Bradley’s monistic account of relations introduces transcendent entities. Any introduction of such entities, since it requires the mind to somehow transcend the world of ordinary experience, inevitably leads to a radical scepticism: the transcendent entities turn out to be unknowable.52 It is the introduction of transcendent entities that generates the radical scepticism we have found to be implicit in Bradley’s metaphysics. One can eliminate the scepticism by eliminating the transcendent entities upon which it is parasitic. Russell eliminates the transcendent entities by appeal

298 to the empiricist’s Principle of Acquaintance (PA). The appeal to PA can be either positive or negative. If one is acquainted with an entity, then, positively, PA can be used to justify the introduction of such an entity into one's ontology. Bradley and Russell both appeal to PA thus used positively, to justify the introduction of genuine relations, contrary to the claims of the monadistic view. But PA can also be used negatively, to exclude certain entities from one’s ontology on the grounds that one is not acquainted with such entities. It was just this sort of appeal that Hume made when he excluded objective necessary connections from any empiricist ontology. Russell also makes this negative use of PA to reject the monistic account of relations on the grounds that it requires the introduction of entities which are to be excluded on the grounds that we are not acquainted with them. If Bradley is correct, then one can identify a quality of a thing only if one goes beyond that quality and relates it to other qualities. For, qualities of things stand in the relation of contrariety one to another, that is, for qualities we have laws of the sort (!) Whatever is F is not G If this regularity is to be more than a matter of chance, it must be the expression of a real relation, or, what amounts to the same, a necessary connection, that obtains between the qualities F and G. Thus, in order to identify the quality of a thing as F, one must also refer as a matter of necessity to the property G, to which it is necessarily tied. So Bradley’s account of relations requires the introduction of a third particular, the Whole, over and above the two particulars that stand in relation to each other. This relation is such that the one quality cannot be identified independently of its necessary connections to other qualities. But, Russell argues, qualities can in fact be identified as themselves without reference to the relations in which they stand to other qualities and other things. As he puts it: To say that two terms which are different if they were not related, is to say something perfectly barren; for if they were different, they would be other, and it would not be the terms in question, but a different pair, that would be unrelated. The notion that a term can be modified arises from neglect to observe the eternal self-identity of all terms and all logical concepts, which alone form the constituents of propositions. What is called modification consists merely in having at one time, but not at another, some specific relation to some specific term; but the term which sometimes has and sometimes has not the relation in question must be unchanged, otherwise it would not be that term which has ceased to have the relation (Pr. of Math., p. 448).53

299

Note that here Russell is allowing Bradley’s point against the monadistic account of relations. On the latter, the predications of one term of a relation would not change if the other relatum ceased to exist. This was Bradley’s criticism of that account, and Russell accepts this point. What Russell is denying is the implication of the monistic account of relations that there is something about properties or qualities as presented that requires us when we are identifying them to refer as a matter of necessity to other properties, those to which they are necessarily tied. Russell is holding that properties are presented to us as logically selfcontained rather than as necessarily tied to one another; he concludes that there are no such necessary connections. But such connections are required by the monistic account of relations. The falsity of the latter view follows. Russell’s rejection of Bradley’s monistic account of relations on the basis of an appeal to the empiricist’s Principle of Acquaintance is clear. 7. Russell’s Account of Relations Russell offered another alternative to the monadistic account of relations, one which, like the monistic account, accepted that there are genuine relations which imply that their relata are inseparable in their being from each other, but which also, unlike the monistic account, recognized that order is irreducible and fundamental. Moreover, Russell’s account avoided the problems introduced into ontology by the monistic account of relations when it introduced entities – unexperienced particular relational wholes and objective necessary connections – that violate the empiricist’s PA. In fact, so successful has it been in solving the problems in the ontology of relations that Russell's proposal has become the commonplace. It does not therefore require much development. Consider once again the relational statement that (@) a is R to b or, in symbols, R(a,b) Russell rejects both the monadistic account in which (@) is to be analyzed into (#) a is r1 and b is r2 or, in symbols, r1(a) & r2(b) and also the monistic account in which (@) is to be analyzed into (*)

300 (a, b) is r or, r(a, b) where the whole (a, b) is taken itself to be a particular thing. Both the monadistic and the monistic accounts of relations assume that predication always involves only one term. This is the “common opinion...that all propositions, ultimately, consist of a subject and a predicate” (Pr. of Math. p. 221). We saw that Leibniz made this assumption when he said that about the proposal that greater than cannot be predicated of both L and M since accidents cannot be as it were bipedal: as Russell translated Leibnitz’s point, “It cannot be said that both of them, L and M together, are the subject of such an accident; for if so, we should have an accident in two subjects, with one leg in one, and the other in the other; which is contrary to the notion of accidents. Therefore we must say that this relation, in this third way of considering it, is indeed out of the subjects; but being neither a substance nor an accident, it must be a mere ideal thing, the consideration of which is nevertheless useful” (Pr. of Math., p. 222; n. 6 supra). Russell rejects this common assumption. Unlike the monistic and the monadistic accounts, Russell’s account of relations takes the grammatical form of (@) to perspicuously represent its logical form. The objective fact represented by (@) does not dissolve into a pair of facts about individuals – (#) – as on the monadistic account. Rather, as on the monistic account, a and b are located in a genuine unity such that, if one of a or b were not to exist, that unity could not exist. But on the other hand, this unity is not a whole of which the relation is predicated, as on the monistic account. To the contrary, the relation is predicated of the terms jointly. It is a and b being related that is the unified whole, rather than a and b being constituted into a whole of which the relation is then predicated. As for the nominalistic requirement that Bradley takes over from the empiricist tradition, the condition that the relation must be predicated of a (single) particular, Russell simply gives that up when he allows that the relation is an entity that can predicated of two particulars.54 Moreover, both the monadistic and monistic accounts of relations, insist that the order of the terms in the expression (@) a is R to b is irrelevant. Though ‘a’ and ‘b’ occur asymmetrically in (@), the individuals that they represent occur symmetrically in the facts in the world: ontologically the order that appears in the statement (@) is an

301 accident of grammar, or of the movement of the mind making the judgment that the use of (@) expresses, but in any case represents nothing real, nothing objectively in the entities said to be related. For Russell, to the contrary, it is precisely the order in the statement (@) that represents the order objectively there in the relational fact. Russell’s account meets the two crucial objections to the monadistic account. It allows that the relata in (@) are, given the relation, inseparable in their being: if b ceases to exist, so does the relational state of affairs (@), and if that happens then a changes by losing the property of being R to b. Moreover, unlike the monadistic account, it allows that there is in the world an element of order. Russell’s, unlike the monadistic account, thus provides an ontological ground for the difference between symmetrical and asymmetrical relations. Russell’s account also meets the objections that he raised against the monistic account of relations. The monistic account, Russell argued, is defective because it does not provide a ground for order; but we have just noted that Russell’s account specifically introduces order. The monistic account is defective, next, because it introduces transcendent particulars that require one to say that in the end all relational judgements are false; but upon Russell's account, there are no transcendent entities, no transcendent particulars, and no premises that would force one to the conclusion that all relational judgements are false. And finally, the monistic account is defective because it leads to an unacceptable holism; Russell rejected that holism by an appeal to PA that established that properties are self-contained, self-identical unities, and there is nothing in Russell's account of relations that would imply otherwise. But Bradley, as we all know, had a general argument against the reality of relations. ...how the relation can stand to the qualities is ... unintelligible. If it is nothing to the qualities, then they are not related at all; and, if so, ... they have ceased to be qualities, and their relation is a nonentity. But if it is to be something to them, then clearly we now shall require a new connecting relation. For the relation hardly can be the mere adjective of one or both of its terms; or, at least, as such it seems indefensible. And, being something itself, if it does not itself bear a relation to the terms, in what intelligible way will it succeed in being anything to them? But here again we are hurried off into the eddy of a hopeless process, since we are forced to go on finding new relations without end. The links are united by a link, and this bond of unition is a link which also has two ends, and these require each a fresh link to connect them with the old (A & R, pp. 27-8).

Bradley puts the point in terms of the relation that binds qualities into the

302 whole of which they are predicated, but his point is clearly generalizable to all relations. Either the relation R connects A and B or it does not. If it does not, then A and B cannot be said to be related. If R does connect A and B, then it must itself be related to A and B, let us say by the relation R’. R actually relates only by virtue of another relation R’. However, R’ will in turn have to be related to R, A, and B, by, say, R”. And R” relates only if there is another relation R”’. And so on. Thus, if R really does connect A and B, then we are driven to the conclusion that there is an infinity of relations, and, moreover, this regress is vicious, for we never arrive at a point where we find a relation actually relating. We therefore cannot have relations actually relating: all relations are unreal. This is a strong argument. It clearly eliminates the monadistic account of relations. That view holds that the relata A and B are independent. That means that there is objectively no genuine relation holding between them. But if the proponent of the view attempts to remedy this by adding the relation as a third entity, then, upon the view, it must be an entity separate from the other two. We have therefore achieved no relation. If we now try to achieve this relation by adding yet another, we again have failed. And so on. The monadistic view cannot escape the vicious regress that Bradley creates. But the regress does not affect the view that Russell defends. For, on Russell’s view, the relation is not separate from the relata, it is not a third entity that is as it were alongside the two which it relates. A “relation” which is separate from its relata will of course need a further relation to relate it to what is supposedly relates. But Russell’s relations are not in this sense separate from their relata: they need no further relation to relate them to what they relate. As Russell says, “it is part of the very meaning of a relational proposition that the relation involved should have to the terms the relation expressed in saying that it relates them...” (Pr. of Math. pp. 99-100). Bradley’s argument gains its force from the assumption that wherever any two entities are related there is a relation that relates them. Russell’s account of relations is not refuted by the argument because it rejects this assumption: to the contrary, relations are in their nature such that they need no relations to relate them to the entities that they relate.55 And indeed, Bradley gives no reason for supposing the assumption to be true. To be sure, it is an assumption that Bradley shares with those who defend the monadistic view of relations. But Russell, like Bradley, rejects that view. Russell accepts only the weaker assumption, that where ever any two non-relational entities are related there is a relation that relates them.

303

8. Bergmann’s New Account of Relations Among philosophers there has been a continuing reluctance to accept Russell’s case that order must be taken as primitive. There is a hankering after the notion that what is not structured can be made to generate structure, that order can be created out of non-order. At least some of this involves the wish that two- and n-place predicates can be replaced by oneplace predicates, the wish of the monadistic tradition that there are no, to use Leibniz's metaphor, bipedal properties. This no doubt lies behind the attraction that the Wiener-Kuratowski device56 has for certain philosophers. For many purposes, especially those of the mathematicians, the content of a relation is unimportant; all that is needed for these purposes is the fact that the relation imparts an order to the pairs that it relates. What is important is, as one says, the relation considered in extension. Wiener and Kuratowski aimed to simplify some of the mathematics involved in relations in set theory by finding a simplified way of dealing with relations in extension. If we have two individuals a and b, and the sets {a} and {b} that have these as their only members, then they ask us to consider first the sets {a, b} and then the two further sets {{a}, {a, b}} {{b}, {a, b}} which they then abbreviate respectively as < a, b > < b, a > and refer to as “ordered pairs.” They are able to prove that ordered pairs in this sense satisfy such axioms as < a, b > = < b, a > . = . a = b This establishes that ordered pairs in this sense satisfy the same formal axioms as pairs ordered by relations. That means that there is a one-one relation, or isomorphism, between Wiener-Kuratowski ordered pairs and pairs ordered by relations. This isomorphism permits the mathematicians to treat the two as equivalent for their purposes, and to substitute sets of Wiener-Kuratowski pairs for the sets of pairs ordered by relations. And it replaces the complication of two-place (and multi-place) predicates with the simplicity of sets, that is, one-place predicates – though, to be sure,

304 there is a price to be paid in that one must use not only individuals and sets of individuals but also sets of sets. But however useful this device is to the mathematicians, it is hardly the stuff of philosophy. Isomorphism does not secure identity, and one cannot therefore appeal to this device as a solution to the ontological problem of order. The device to the contrary presupposes the very order that its supporters hope to supplant: Wiener-Kuratowski sets are designed to reflect the order that relations establish among things, and those sets cannot do that unless there is an order there to be represented. The device may allow the mathematicians to ignore that order, but it does not eliminate it. As Reinhardt Grossmann once put it, “ordered couples are no more identical with their coordinated classes than people in a theatre are identical with the chairs on which they sit.”57 One finds the attempt to generate order out of non-order pursued in a different way in the later work of Gustav Bergmann. Bergmann long defended Russell's account of relations as a significant advance in philosophy,58 and emphasized the point that genuine relations need no relations to relate them to what they relate.59 To be sure, some ordinary “descriptive” relations, such as those of space and time, do need relations to relate them to what they relate. This job of relating is done by the “tie” or “nexus” of relational exemplification. This, though, is not a significant difference from Russell. However, in his posthumously published New Foundations of Ontology,60 we find that Bergmann came to revise his estimation of Russell’s account, and to offer yet another. In particular, he argued for an account of order that is significantly different from that of Russell. More strongly, he goes so far as to say that he cannot “credit Russell with a contribution to the ontology of order” (New Foundations, p. 120). Bergmann in fact now accuses Russell of simply “papering over” the absence of order in his ontology. Russell introduces relations in extension into Principia Mathematica in Section *21.61 He defines a predicate (i)

ûŵ[Ruw](x, y)

This is a predicate expression in just the sense in which ‘R(x, y)’ is a predicate expression. (i) is so defined that it holds of x and y if and only if (ii)

R(x, y)

305 (i) is a class of ordered pairs; it can be proved that the members of (i) satisfy the axioms that must be satisfied by ordered pairs. Specifically, (i) is the class of pairs ordered by the relation R; it is, in other words, the relation R considered in extension. The order in the pairs is that which is established by the relation R. Thus, the concept (i) does not introduce the notion of order but rather presupposes it. Russell does not make use of the notion of an ordered pair, but his apparatus enables one to introduce such a notion. The ordered pair < a, b > can be defined in the way just suggested as (iii)

ûŵ[u = a & w = b](x, y)

(iii) will hold of x and y just in case that we have (iv)

x=a&y=b

It is thus analytically true that (iii) holds of the pair a and b and analytically false that it holds of any other pair. In particular, it is analytically false that (iii) holds of the pair b and a. But here again, (iii) does not introduce order but rather presupposes it. The order which it presupposes is represented by the typographical order of the conjuncts in (iii), or, what amounts to the same, in (iv). If it is Gustav who has selected the arrangement (iv), then the order among the individuals a and b that is represented by the typographical order is given by the relational fact: (v)

a and b stand in the relation of prior to posterior in Gustav's selection

(v) is of course an objective relational fact, though it is, to be sure, a fact that exists by virtue of an arbitrary choice made by Gustav. In this sense, the arrangement is indeed arbitrary, but it is still an arrangement, a relational fact about the individuals a and b. It is the relation (v) that orders the pair consisting of a and b, and it is this order that is represented by the typographical order in (iii). Here, as in the case of (i), we see that the defined predicate (iii) does not introduce the notion of order but rather presupposes it. Specifically, then, Bergmann is wrong, or at best misleading, when he asserts that “the order in the existents Russell called relations in exstenso rests on the order of the marks on paper and on nothing else”

306 (New Foundations, p. 120). What he implies is that the only order that Russell introduced into the world was the order of marks on paper, that Russell did not make order a fundamental ingredient in his notion of a relational fact. In fact, as he continues, “That is why I didn’t credit Russell with a contribution to the ontology of order” (ibid.). But in fact, as we have seen, Russell made it central to his ontology of relations that they all involved a sense. Order was central to Russell’s concerns, and the fact that the monadistic and monistic views could not account for order, where his could, and indeed made it basic, was for Russell a central argument against those views and for his own. As for the issue of typographical order, it is not true to say that the order of the existents rests on the typographical order and nothing else, as if that typographical order were nothing more than a linguistic fact which did not represent any extra-linguistic structure, that is, as if the extra-linguistic structure was created by rather than represented by the linguistic order. Bergmann in fact is accusing Russell of falling into a radical nominalism with regard to order (reminiscent of the nominalism of Sellars), a nominalism that implies that there is no objective order and what order there is an artefact of language. But this is not a correct reading of Russell. In fact, the linguistic order in (iii) does represent an extra-linguistic structure, namely, the structure more fully expressed by (v). It is of course true that Gustav, in the act of writing down (iii) in the order that he did, created, brought into existence, the fact (v) which that order represents. The writing of the letters in that order makes palpable the fact (v). But nonetheless there is a distinction to be drawn between the order of the marks in (iii) and the relational fact (v) which the act of writing those marks in that order brought into existence. Once we have that distinction, it is clear that the order of the marks represents, rather than provides the ontological ground of, the order (v), contrary to what Bergmann suggests. It is just wrong, then, to say that for Russell the order in the ordered pair “rests on the order of the marks on paper and on nothing else.” However, for Bergmann, since Russell did not succeed in introducing order into his ontology, a new account is needed. In his new account, Bergmann introduces ordered pairs: < a, b >. The relational fact (@) a is R to b is given a very complex analysis (New Foundations, pp. 120-1). On this new analysis, the relation R is exemplified by the ordered pair < a, b >.

307 Bergmann represents relational exemplification by “η”. Thus the relational fact (@) is to be more perspicuously represented by η(R, < a, b > ) Exemplification η connects R and the ordered pair but is not itself a relation. Or at least, it is not the same sort of relation. For that would lead to a Bradley-type regress: we would have to have some further exemplification connection, say η’, connecting η to the ordered pair < R, < a, b > > with the regress both obvious and vicious. So relational exemplification η relates but needs no relation to relate it to what it relates. And rather than referring to η as a “relation”, Bergmann calls it a “function”. As for the ordered pair < a, b >, Bergmann construes this in terms of a more basic sort of entity. Bergmann here begins with the notion of a “diad” (New Foundations, p. 101f). If a and b are two entities, then the “diad” for a and b, represented by (n) (a, b) 62 is the fact that a is different from b The grammatical form representing the ordered pair, that is, the expression “< a, b >”, is a shorthand representation of the fact, or, rather, diad (nn) (a, (a, b)) which is the fact that a is different from the fact that a is different from b; while the expression for the opposite ordered pair, namely, “< b, a >”, is a shorthand representation of the diad (b, (a, b)) Thus, the relational fact (@) is, according to Bergmann, still more perspicuously represented by (g) η(R, (a, (a, b))) Now, whatever we might say about the status of Bergmann’s special category of fact (n) that he calls diads, we can already see that there are problems with what he says concerning relations. As Bergmann admits (New Foundations, p. 120), he puts order into relational facts by means of the Wiener-Kuratowski device, save that he uses diads rather than sets. But surely the objection is the same. What Bergmann establishes is an isomorphism between the set of diads (nn) and the pairs that are ordered by the relation R. As we said, however, isomorphism does not secure identity, and one cannot therefore appeal to this device as a solution to the ontological problem of order. The device to the contrary presupposes the very order that its supporters hope to supplant: Like Wiener-Kuratowski

308 sets, Bergmann’s special diads are designed to reflect the order that relations establish among things, and those diads cannot do that unless there is an order there to be represented. What we need is an order between a and b as related by R. Russell’s account secures this order by building it into the relation itself: relations, upon Russell’s account, invariably have a sense. What Bergmann provides us with is not this order but another set of entities, which is such that the order among these entities is isomorphic to the order established by the relation. In fact, Bergmann on the one hand ignores this order established by the relation R substituting for it the order among his special diads; while on the other hand he implicitly presupposes it as just that order which has to be reflected isomorphically in the structure of these diads. Bergmann’s device may allow him to ignore the order that R imposes upon its relata, but it does not eliminate that order. The order that Bergmann here ignores is, of course, precisely that order that he accused Russell of introducing only linguistically. In both cases, Bergmann fails to see the order that is present with the relation itself, given Russell’s account of relations. Failing to see it, he feels forced to introduce a substitute to do the job. There is another way of making roughly the same point. In our experience of relational situations, what we experience are particular things related to each other. Consider the sort of experienced situation that would be represented by (@), a is R to b, say the situation in which a red square a is wholly within – R – a green circle b. What we experience here is a being related in a certain specific way to b. As Bradley put it, a relation is not only not its terms but “is between them” (“Relations,” p. 636), between them, not between other things, and not exemplified by other things. What we have in the situation represented by (@) is a relation R relating a to b and not a property R exemplified by the fact that the individual a is different from the fact that a is different from b which is what we would be presented with if Bergmann were correct in holding that (@) is most perspicuously represented by (g). Bergmann’s (g) simply does not represent what is given to us in experience, the facts as we are acquainted with them. Bergmann, like Bradley, is forced by his dialectic to introduce into his account of relations entities that are excluded

309 form one’s ontology by the empiricist’s Principle of Acquaintance. Anyone who adopts that Principle is therefore forced to reject Bergmann’s claim about the ontological adequacy of his account of what we are presented with when we are presented with the fact that a is R to b. For such a one as Russell, that is conclusive. Bergmann can avoid this only by claiming that ordinary experience, that to which appeal is made in the empiricist’s PA, does not reveal the deep structure of the relational fact. For that we need other sorts of thought, other sorts of analysis. In fact, Bergmann suggests just such an appeal. He of course allows that the individuals a and b are given in ordinary experience, but he refuses to allow such a appeal to establish that the diad, a is different from b, is given in experience: he does not here turn to PA. Rather, he turns to what he calls the “Principle of Presentation,” which attributes existence to any entity that can be thought. We do sometimes think that a is different from b, and so, by the Principle of Presentation, this diad must exist. Hence, while we may not be acquainted in ordinary experience with the real structure (g) of the relational fact (@), we can know that real, deeper structure when we think with Bergmann what is really involved; and when we so think that structure, we know that it exists: for, then it has been Presented to us. This is ever the way with metaphysics.63 First an entity is invented that is not given in ordinary experience. It is introduced in order to solve some problem supposedly insolvable in any other way. Then a new way of experience is invented in order to convince us that it is really there and can be known by us. Bergmann wishes to account for the order of the world. This can be solved only by introducing the complex account (g) of relational facts. The complexities this account requires are not presented to us in ordinary experience; we are not acquainted with them. Bergmann therefore introduces another mode of experience, which he calls “presentation,” and according to which the new entities are in fact and after all experienced. We saw this pattern in Bradley, and we saw that Russell raised important objections to it. Specifically, it leads to an intolerable scepticism. The problem with this new sort of experience is that, simply, it is problematic. Either ‘presentation’ covers what we are thinking about, in which case it allows too much into one’s ontology – after all, Bradley thought quite a bit about the Absolute –, or it covers some restricted part of what we are thinking about, namely, just those things that one wants to allow into one’s ontology. But in the latter case, it cannot itself be a

310 principle that can determine what is or is not part of one’s ontology: one needs a further principle that determines whether what is thought is ontologically permissible, and therefore “presented” in the special sense. Not surprisingly, Bergmann (New Foundations, Ch. 2) spends a lot of time arguing that certain things can, while other things cannot, be presented. That means that he cannot after all appeal to his “Principle of Presentation” as a way of justifying the claim that his deeper relational structures (g) are real simply because they are “presented,” that is, thought of. Of course they are, but then so is Bradley’s Absolute. In neither case can the appeal to “presentation” by itself establish the ontological claim.64 More strongly, once transcendent entities are introduced, whether they be those of Bradley or of Bergmann, there is no clear ground in experience for ontological claims. Without such a ground, everything and nothing may be claimed about these entities. In the absence of a conclusive criterion, a disastrous scepticism results, as we saw Russell arguing concerning Bradley’s transcendent Absolute. Bergmann’s ontology of relations is therefore subject to much the same sort of criticism that Russell used against Bradley. What is at the root of Bergmann’s problem of order would seem to be a horror of it, the same sort of horror which we can see in the defenders of the monadistic and monistic accounts of relations, and which we can see motivating those who would replace relations by means of the WienerKuratowski device. In Bergmann’s case, the horror locates itself at a different point, but it is nonetheless there. Bergmann ignores the order, the sense of the relation, that is there at the core of Russell's account of relations. But order does exist, and he must therefore construct a substitute. What resources does he permit himself to use in this construction? Where he begins is with the diad, that is, (n) (a, b) which is the fact that a is different from b (n), clearly, is relational. These facts of diversity are the only relational facts which Bergmann will allow as irreducible. All other order is build up in the manner we have described from these basic facts of diversity. Now, these basic facts of diversity, while relational, are all symmetrical. Where a relation is symmetrical, if it holds of a pair of individuals, then so does its converse. The order or sense of the relation can therefore be neglected. But in the case of an asymmetrical relation, if it holds between a pair of

311 individuals, then its converse does not. That means that in the case of asymmetrical relations, we cannot neglect the order or sense of the relation. Since in the case of asymmetrical relations, the obtaining of the relation guarantees that the converse does not obtain, it follows that asymmetrical relations cannot be defined in terms of symmetrical relations, as Russell noted when he pointed out that “some asymmetrical relations must be ultimate, and ... at least one such ultimate asymmetrical relation must be a component in any asymmetrical relation that may be suggested” (Pr. of Math., p. 224). Bergmann thus, when he limits himself to symmetrical relations in his relational basis, does not provide himself with a starting point sufficient to generate asymmetrical relations. He must therefore resort to the Wiener-Kuratowski device, or its close analogue, to generate the order necessary for asymmetrical relations. But why does Bergmann restrict himself to a set of basic relations that are only symmetrical? Because, I suggest, with these relations, order can be neglected. It is precisely in restricting himself to symmetrical relations that Bergmann indicates that he shares with the monadists, monists, and lovers of Wiener-Kuratowski the horror of order that Russell rightly insisted is basic to any ontolgoically adequate account of relations.

Conclusion: Russell’s arguments against both the monadistic and monistic accounts of relations seem conclusive. Bergamnn has suggested that Russell has not provided an adequate account, and has suggested an alternative. We have argued, however, that Bergmann is wrong, both in his evaluation of Russell’s contribution and in the suggestion that he has achieved an alternative that survives the sorts of criticisms that Russell made of Bradley. 9. The Russell-Hochberg Emendation Russell was later to question what he had earlier argued with respect to the ontology of relations. This was not to withdraw from the thesis that order is there in the world, and which therefore needs some account in one’s ontology: but it became Russell’s view that his earlier position (of the Principles of Mathematics) that the sense of a relation, that which grounded its capacity to order individuals, was a primitive feature of relations, incapable of further analysis, failed to provide the ontological

312 account of order that would be required for any account to be adequate. Bergmann’s student, H. Hochberg accepts the criticism of Bergmann, arguing that it amounts to an attempt, inadequate, to use the WeinerKuratowski trick to replace relations with non-relations; but has more recently come to agree with Russell, that the earlier Russellian analysis will not do. The problem of order returned to Russell’s consideration through a problem in the philosophy of mind. The problem was that of false belief. The statement that Jones believes that p apparrently states a relational state of affairs, one in which there is Jones, on the one hand, the fact p, on the other hand, and the relation of believing holding between these two entities. The difficulty is this. Russell takes the example of Othello’s believing that Desdemona loves Cassio. Here we have Othello who believes, on the one hand, the object of the belief, the state of affairs of Desdemona loving Cassio, and the relation of believing that is supposed to hold between them. Now, ordinarily for a relation to exist in the world as we experience it, both relata must exist. If we have Peter kicks Paul then Peter has kicked nothing, and indeed has failed to kick, if there is no Paul there, to be kicked: no kicker without kickee. But belief is different. Suppose that it is an ordinary relation. It will then require two relata, both of which exist. Now, Othello believes that Desdemona loves Cassio. The object of this belief is the state of affairs that Desdemona loves Cassio: this is one of the relata. As one relatum of the relation of believing, it must exist, as with a relatum of any relation. However, Othello’s belief is false: the state of affairs that Desdemona loves Cassio does not exist. That is what makes Othello’s belief a false belief. One thus seems to end up with the position that here is no false belief. For surely, the point about a false belief is precisely that the object, the state of affairs that is its object, does not exxist. One alternative, then, for those who would follow Russell in looking upon belief as a relation is that no belief has an object that does not exist: no belief, in other words, would on this alternative have a nonexistent state of affairs as its object – no belief would be false. The alternative is to allow that the object of the believing is an existent entity, but that existents divide into those that have the “property” of being actual, while the others have the “property” of being (merely) “potential.”65 The problem is the “aboutness” of thought: more specifically, our thoughts can be about what does not exist. What is the connection that

313 relates, or perhaps “relates”, our thoughts that are to objects of thought that are not? Russell does insist that the aboutness be analyzed in terms of his ontology of relations. So the relata of whatever relation or relations are involved must exist. The straight-forward relational account of the intentionality or “aboutness” of thought won’t do. So at least Russell argued in his unpublished 1913 manuscript on the theory of knowledge.66 Russell therefore suggested that the connection of thought to its object was not a simple two-place relation.67 To be sure, simple acquaintance, as in sensory experience, remained, as Russell proposed, a two-place relation. This was satisfactory, since no one argued that simple acquaintance was non-veridical. This was so, whether one was acquainted with a sense impression or with a property or universal such as (the colour) red or (the property) square: with simple acquaintance the question of the non-existence of the object of belief simply did not arise.68 Where the problem arose was in those cases where the object was some complex entity, a fact. One might well in the awareness of such an entity be simply conscious of the parts, so that the existence of the complex, the fact, would not be an issue. The problem of the non-existent object of thought arises with respect to complexes in the sense that the parts may not exist in reality combined in the way in which they are presented as combined: the world may be structured in some way other than it is presented as structured. For Othello, the world presents itself as one in which Desdemona loves Cassio, but in reality the world is otherwise structured, the entities are combined in other ways, Desdemona does not love Cassio, Othello’s judgment is false.69 Russell allows that Othello is acquainted with the entities loves, Desdemona, Cassio This acquaintance is unproblematic: there is no issue as to the existence of these entities as logically and ontologically separate, so acquaintance with them can be construed as a simple relation.70 Russell then proposes that we can analyze (i) Othello judges that Desdemona loves Cassio into (j) U(o, d, L, c) where U is a many-term relation (“understands”) connecting Othello (“o”), Desdemona (“d”), loves (“L”) and Cassio (“c”). There is a difficulty here,71 however, which Russell did not at first notice. Suppose, contrary to fact, that (k) Othello judges that Cassio loves Desdemona

314 If we use the same model for analyzing this judgment as we used to analyze (i), then we have (l) U(o, c, L, d) On the one hand, we might suppose that the order of d, L, c in (j) and (o) makes no difference. But in that case, (j) and (l) do not represent different complexes. In that case, (j) and (l) fail to capture the difference between (i) and (k). So this won’t do. On the other hand, we might suppose that the order of d, L, c counts. In that case there is a difference between (j) and (l), with the former representing (i) and the latter (k). But in this case the order should be explicit as among the objects thought. (j) and (l) are therefore by themselves inadequate: they ought to show how U has this form among the objects or entities to which the mind o is related. So Russell introduced the notion of the form of a fact. L or Lxy is a relation and Ldc is different from Lcd; in o’s thought it is d which is in the first place of the relational fact and c which is in the second place. Russell therefore suggests that there are two further relations, “x is in the first place of the relational state of affairs involving Lxy” and “y is in the second place of the relational state of affairs involving Lxy”. Call the former L’ and the latter as L”. L’ and L” relate an individual to a fact.72 We need to refer to this fact. A proper name won’t do, since if the belief is false the fact does not exist and therefore cannot be named. Russell therefore suggests we use an “incomplete symbol”73 to refer to the fact, that is, in effect a definite description. We have something like this: (E) (E! p)[ L’dp & L”cp & CLp & F(rxy, p) ]74 where “CLp” represents that the relation L is contained in (“C”) p, and “F(rxy, p)” represents that the form rxy of a relational fact is the form (“F”) of p. The “E!” has the usual connotation of existence and uniqueness (“there is one and only one p such that ...”). (E) permmits the introduction of the definite description (D) (ιp)[ L’dp & L”cp & CLp & F(rxy, p) ] “the fact p such that ...”. As a piece of notation let us use the expression [Lcd] as an abbreviation for (D). Then what Russell suggests is that we can represents Othello’s state of believing as (m1) U (o, [Lcd]) where “[Lcd]” is the fact that Othello believes.75 This solves the problem

315 of false belief. That problem arose when we took Othello’s belief to be represented by something like (m2) U (o, Ldc) with Othello “o” related to the fact Ldc. But it could not be so related since there is no such fact; it is false that Desdemona loves Cassio. Since there is no relatum as the second term of U in (m2), it follows that (m2) is in effect meaningless, and we have therefore not succeeded in representing Othello’s believing as he does. But with (m1) we do not have this problem. For, ‘[Lcd]’ is functioning as a definite description. Consider Peter kicks Paul If Paul does not exist then the term ‘Paul’ does not refer and in fact is a meaningless term. So this sentence is meaningless. In contrast Peter kicks the present King of France remains meaningful even if there is no present King of France. Or consider 3 comes before the greatest prime number This, too, is meaningful even if there is no greatest prime number. This is because, as we have learned from Russell, ‘the greatest prime number’ and ‘the present King of France’ are not genuine referring expressions, and so they remain meaningful even when they do not refer.76 The point is that ‘[Ldc]’ in (m1) is a definite description, so, even if it does not refer, that is, even if, as one says, it is not successful, sentences which contain it remain meaningful, just as sentences containing the definite descriptions ‘the present King of France’ and ‘the greatest prime number’ remain meaningful even though these definite descriptions are not successful. The problem that arises for (n), that it is meaningless when ‘Ldc’ is false, does not arise for (m). Since the problem with (n) does not arise for (m), we can use the latter to represent Othello’s state of believing that Desdemona love Cassio. This ingenious, and even has a further point in its support. This has to do with the account to be given of truth (and falsity).77 There is not just a problem of false belief, there is also a problem with a sentence like Ldc or like “a is red” Fa when it is false. In that case, we have both ~Ldc and ~Fa

316 as true. What do these facts represent? At the time of the 1913 manuscript Russell did not recognize or, at least, did not wish to recognize, the existence of negative facts, or, what amounts to the same, the existence of false facts – a realm of the merely possible alongside the real, the actual. It amounts to the same since if there is no fact for the false proposition to reefer to or describe, then it refers to or describes nothing and is therefore meaningless; it must therefore refer to something, a fact characterized by non-existence, or, in other words, a negative fact. Russell’s approach to false belief provides him with a way of avoiding a commitment to false facts or negative facts. We can do for ‘Fa’ what we did for ‘Ldc’, namely make its form explicit in a reference to the fact that makes ‘Fa’ true, if it is true. (E! p)[ C(F, p) & C(a, p) & I(fx, p) ] where we record that a and F are in p and that the form of p is fx. We can then form the definite description (ι p)[ C(F, p) & C(a, p) & I(fx, p) ] which we abbreviate as [Fa] We have (t) ‘Fa’ is true = Fa = E![Fa] and (t’) ‘Ldc’ is true = Ldc = E![Ldc] Thus, ‘Fa’ is not true does not require the existence of a false fact or a negative fact; to say it is not true is to say that the definite description ‘[Fa]’ is not successful. A proposition being true is thus grounded in a fact, an existing or real fact – are there any other kinds of fact? – but a proposition being false does not require us to say that the right hand side of ‘p’ is true = p is meaningless when ‘p’ is false because the term ‘p’ is not in that case a referring expression. Rather, we understand “‘p’ is true” as saying ‘p’ is true = E! [ p ] where we do for ‘p’, whatever ‘p’ is, what we have done for ‘Fa’ and ‘Ldc’. Since ‘[p]’ is a definite description, it does not become meaningless when it does not refer, that is, when ‘p’ is false.78 Russell later came to accept the existence of negative facts (in his “Logical Atomism” lectures79), and he also later came (in “On Propositions: What They Are and How They Mean”80) to reject the account

317 of belief that he devised in the 1913 manuscript. We need not here deal further with the issue of the existence of negative facts or of negation.81 As for the problem of belief that he was struggling with in 1913, one must surely recognize that when Othello expresses his belief as Desdemona loves Cassio what is before his mind is the state of affairs represented by Ldc and not the rather more complicated state of affairs represented by [Ldc]82 Russell has his problem because he has in U(o, L, d, c) the list of constituents L, d, c and he must reconstruct the state of affairs Ldc from this list – which cannot done: the order required for the state of affairs is not there in the list.83 So Russell must include something else among the relata of U that will enable his to as it were reconstruct the state of affairs believed in from the list of its constituents. These are the relations L’ (“in the first place of Lxy”) and L” (“in the second place of Lxy”). And instead of the simple state of affairs represented by Ldc we need something like, to begin with, L’dp & L”cp & I(Lxy, p) but ending up with the more complicated state of affairs denoted by [Ldc] But, while Othello is believing a state of affairs for which all this more complicated statement is true, nonetheless it is simply the state of affairs itself which is before the mind and not all the details of how constituents all fit together to form that state of affairs. Russell’s construction of the judgment is therefore implausible because it fails to capture what is in fact the real intention of the thought. Russell came to recognize the point by the time he wrote the essay on “Propositions.” Of course, this returned him to the problem of thought intending non-existent states of affairs. But by that time he was on his way to developing an account of intentionality based on language, in which “means” in the sense of intentionality is construed in the same as one construes “means” in the sense of the meaning of sentences in language, that is, language as overt linguistic behaviour. This he worked out in greater detail in his Analysis of Mind.84

318 In that same work, Russell worked out an account of negation that obviated the need for ontologizing, construing as existing entities, negative facts, or, what is much the same, false facts. So there is no need for one to turn to the complications of ‘[Fa]’ and other such expressions to handle these problems. One can return to treating ‘p’ is true = p as a linguistic truth on the model first suggested by Wittgenstein’s Tractatus. What, then, of the remaining use of the relations “x is in the nth place of the relational fact p”, that is, relational facts like L’ap and L”bp to provide, as Hochberg argues,85 an account of the order in relational facts? We had argued above that the relational structure of the sentence ‘Ldc’ represents the order of entities present as constituents of the state of affairs Ldc What we need is an order between d and c as related by L. What we proposed is that Russell’s account of relations as laid in the Principles of Mathematics secures this order by building it into the relation itself: relations, upon Russell’s account, invariably have a sense. And we argued against Bergmann that this sense is represented in language describing these states of affairs by the relational order of the terms in the sentence representing the relational state of affairs.86 Hochberg argues that, while this does admit, against the monistic and monadistic accounts of relations that relations as actually relating things into facts have an intrinsic sense, it does not suffice to recognize the existence of that order, but one must provide entities, beyond the relation relating which account for that order. It is, Hochberg argues,87 the task of the “x is in the nth place of p” relations to provide that account of order. From the list R, a, b one cannot reconstruct the state of affairs that exists, is it Rab or Rba which is actual? which is the fact or existing states of affairs? Suppose that it is the former which is the existing state of affairs: the point is that one cannot tell that this is the way the world is from the list of objects in the

319 world. Knowing that a is in the first place and b in the second place of p, i.e. that R’ap and R”bp then we can reconstruct the fact p (that is, Rab) from the list of its constituents, that is, reconstruct it given that p has a relational form I(rxy,p) and that a and b are the particulars in p and R is the relation in p C(a, p), C(b, p), C(R, p) Now, it is certainly true that if we take p to be Rab then we have as two additional relational facts R’ap , R”bp But it does not end there: there is an order here that also exists. If p’ is the first of these facts and p” the second then we have something like (R’)’ap’, (R’)”bp”, (R”)’bp”, (R”)”bp” There is a regress here – certainly all these facts in this infinity of facts do exist if Rab exists. Indeed, if our aim is to reconstruct the latter from the list R, a, b then the regress is vicious, since we never end up with a set of facts from which we can construct Rab. The regress is not vicious, however, if we start with from Russell’s account of relations, and admit that the fact Rab is irreducible, and as it part of the irreducibility admit that the sense of R as it does the relating in this fact is simply given as being the way the world is. If it is admitted that Rab, including the sense of R as does its relating of a and b, is irreducible, then the facts appearing in the regress built on the “is in the nth place of” relations are all there but are there a simple linguistic variations on the basic sentence “Rab”: they are not additional facts, but only different ways of saying what is already said by “Rab”. What we are saying here is that this regress is analogous to the Bradley-style regress R(a, b) R’(R, a, b) R”(R’, R, a, b) etc. Russell, in the Principles of Mathematics, took the relational fact from which this regress begins to be irreducible, contrary to what Bradley held,

320 and argued that the further facts of this Bradley-like regress were simply linguistic variations on the sentence “R(a, b)” which represents that irreducible fact. These variations can all be generated from the sentence from which the regress starts and all of them are true, but the regress is not vicious – it would be vicious if knowing the later terms of the regress to be true before we could add that the relational state of affairs R(a, b) to our description of the world – but it is after all on Russell’s account of relations not vicious because we take as already given the fact R(a, b) from which the regress starts: we do not need the regress to reconstruct the fact, since that fact is taken as a given. Similarly, Hochberg’s facts based on the “is in the nth place of” generate a vicious regress if we take the point of these statements to describe an ordering that is not taken as given, but these facts towards which Hochberg directs our attention do not form a vicious regress provided that we take the starting point “Rab” as Russell takes it, and its sense, as irreducible, something that we must take as simply given. Hochberg suggests that taking the fact Rab and the sense of the relation in that fact as irreducible won’t do, that that cannot be “the end of the matter”, and that one’s ontology requires further entities that can account for the order. What sort of “thing” or “feature” it [the order, the sense of the relation] taken to be, specifically what it is, is another matter, and one that [our Russellian account of relations] fails to take up. One cannot merely speak of order being there or of directions, for the question is about how the order is ontologically grounded in facts.88

One constructs one’s picture of the world by reflecting in language the facts as one is acquainted with them. One writes down a sentence – not a list of terms from which one is expected to re-create the picture of the world, to pick from among all possible sentences those which are true. Asked for a description of the world, one writes down a sentence to picture the fact Rab that is given to one in experience. In our experience, what we are acquainted with is a fact Rab in which R goes from a to b, and one records in one’s picture of the world that fact. One is given the fact in which the relation R relates a to b, with the a in the first place and b in the second place, but one is not given the further fact that a is related to the fact by the relation of being in the first place in the fact and the further fact that b is related to the fact by the relation of being in the second place in

321 the fact. The fact itself is that which “grounds” the order: there is no need for something further to “ground” the order, some further entity or entities to give the order and which one must include if one is to give a complete description of the world as we experience it. But what we are given in experience is the fact, the relational fact and as part of the relational fact that the relation goes in this direction rather than that, from a to b and not from b to a, and this direction or sense is represented in our picture of the world by the order of the terms in the picturing sentence. The order of the relation as it relates is grounded in the relational fact itself, there is no need for some further entity or entities to do the grounding. Hochberg is suggesting that the ontologist needs something to ground the order beyond the fact itself, some further entity, but the fact does the job itself, and there is no such need. His proposed emendation of the account of relations developed by Russell in criticism of the monistic and monadistic accounts of relations is thus not needed. Hochberg slides away from what our acquaintance with world gives to us. We are presented with facts, and, among those facts, relational facts, and in being given those relational facts we are given the sense or direction of the relating relation. Not taking what one is acquainted with as basic, Hochberg slides into thinking that some entities, additional to the fact itself, are needed to ground the order. But, to repeat, that order is given when the fact is given, and the further entities are not needed. Russell introduced the special ordering relations “is in the nth place of” to solve a problem in the philosophy of mind. Hochberg ingeniously adapts this idea to solve what he sees as the more general problem of grounding the orderings which relations as it were impose upon the world. But such entities are not needed: the order the relations impose is given when the facts are given. That is all there is to it.

322 Endnotes to Study Seven

1.See J. Weinberg, “The Concept of Relation,”in his Abstraction, Relation and Induction (Madison, WI: University of Wisconsin Press, 1965). The term ‘monadistic’ is due to Russell; see his Principles of Mathematics, Second Edition (London: Allen and Unwin, 1936), Ch. XXVI. 2.Franco Burgersdijck, Monitio Logica, or An Abstract and translation of Burgersdiscius His Logick, by a Gentleman (London: Richard Cumberland, 1697). 3.Burgersdijck does not get it quite right: the converse of paternity is not son-ship but father’s-child -- he fails to note that daughters too have fathers! 4.John Locke, Essay concerning Human Understanding, Fourth Edition, ed. P. H. Nidditch (Oxford: Oxford University Press, 1979).. 5.G. F. W. von Leibniz, New Essays concerning Human Understanding, trans. Alfred Gideon Langley (LaSalle, IL: Open Court Pub. Co., 1949), Bk. II, Ch. 25. 6.Letter to des Bosses, 21 April 1714, Die Philosophischen Schriften von Gottfried Wilhelm Leibniz, ed. K. I. Gerhardt, reprint edition, vol. II, p. 486; translated in Philosophical Papers and Letters, second edition, ed. L. E. Loemker (Dordrecht, The Netherlands: D. Reidel, 1979), p. 609. 7.In Gerhardt, vol. VII, p. 401. Russell translates this as follows: The ratio or proportion between two lines L and M may be conceived three several ways; as a ratio of the greater L to the lesser M; as a ratio of the lesser M to the greater L; and lastly, as something abstracted from both, that is, as the ratio between L and M, without considering which is the antecedent, or which the consequent; which the subject, and which the object .... In the first way of considering them, L the greater, in the second M the lesser, is the subject of that accident which philosophers call relation. But which of them will be subject in the third way of considering them? It cannot be said that both of them, L and M together, are the subject of such an accident; for if so, we should have an accident in two subjects, with one leg in one, and the other in the other; which is contrary to the notion of accidents. Therefore we must say that this relation, in this third way of considering it is indeed out of the subjects; but being neither a substance nor an accident, it must be a mere ideal thing, the consideration of which is nevertheless useful (Principles of Mathematics, p. 222).

323 The passage also appears in English translation in Loemker, p. 704. 8.Cf. D. Hume, Treatise of Human Nature, ed. L. A. Selby-Bigge, rev. P. H. Nidditch (Oxford: Oxford University Press, 1978), p. 14. 9.Treatise, Bk. I, Part iii; the conclusion occurs on p. 169ff. See Wilson, “Hume’s Theory of Mental Activity,” in D. F. Norton et al., McGill Hume Studies (San Diego: Austin Hill Press, 1979), pp. 101-120; and also F. Wilson, Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1997). For an extended defence of the Humean account of causation, see F. Wilson, Laws and Other Worlds (Dordrecht, The Netherlands: D. Reidel, 1986). 10.Treatise, pp. 14-15. 11.W. Wordsworth and S. T. Coleridge, Lyrical Ballads, ed. W. Owen (London: Oxford University Press, 1971). 12.David Hartley, Observations on Man (orig. 1749; reprinted Gainesville, Fla.: Scholars’ Facsimiles & Reprints, 1966). 13.For a discussion of the way Hartley conceived the connection between associationism and mysticism by means of his – not altogether clear – notion of “coalescence,” see the interesting discussion in Stephen H. Ford, “Coalescence: David Hartley’s ‘Great Apparatus’,” in Christopher Fox, ed., Psychology and Literature in the Eighteenth Century (New York: AMS Press, 1987), pp. 199-223. See also Richard Haven, “Coleridge, Hartley, and the Mystics”, Journal of the History of Ideas, 20 (1959), pp. 477-89. 14.S. T. Coleridge, Biographia Literaria, 2 vols., ed. J. Shawcross (Oxford: Oxford University Press, 1907), vol. 2, p. 39. 15.S. T. Coleridge, Treatise on Method, (London: Constable, 1934), p. 3 16.S. T. Coleridge, Logic, ed. J. R. de J. Jackson [The Collected Works of Samuel Taylor Coleridge, vol. 13], (Princeton, N. J.: Princeton University Press, 1981), p. 189. 17.Ibid., p. 189. 18.To J. P. Estin, 13 Feb. 1798, in S. T. Coleridge, Collected Letters, ed. E. L. Griggs, 6 vols. (Oxford: Oxford University Press, 1956-71), vol. 1, pp. 385-6. 19.S. T. Coleridge, Anima Poetae (London: Heinemann, 1895), p. 102. 20.Dorothy Emmet has noted Coleridge’s concern for structure; see her “Coleridge on the Growth of the Mind.” Bulletin of the John Rylands Library, 34 (1951-52), pp. 27695.

324

21.Coleridge, Treatise on Method, p. 3. 22.Ibid., p. 2. 23.Ibid., p. 11. 24.Ibid., p. 6. 25.Coleridge, Logic, p. 183. 26.Coleridge, Biographia Literaria, vol. 1, p. 83. 27.Ibid., p. 6. 28.Biographia Literaria, vol. 1, p. 181 29.Ibid., p. 181. 30.Ibid., p. 182. 31.Ibid., p. 183. 32.Ibid., p. 109. 33.Berkeley had preceded Coleridge in arguing that mind is responsible for the structuring of things; see F. Wilson, “On the Hausmans’ ‘New Approach to Berkeley’s Ideal Reality’,” in this volume, above. 34.This view of relations was to be developed with greater metaphysical sophistication by T. H. Green, F. H. Bradley and the other British Idealists of the latter part of the 19th century and the early part of the 20th. We are, of course, examining the pattern in the present study. See also F. Wilson, “Moore’s Refutation of Idealism,” in this volume, above. Green, of course, was a product of Rugby, where Dr. Arnold was a strong defender and advocate of Coleridgean principles. 35.Coleridge, Treatise on Method, p. 4. Reason, in Coleridge’s sense, is not seriously to be distinguished from Faith: his metaphysical argument is therefore also a justification of his religious faith. Reason and faith were to remain united in his thought. See his “Essay on Faith,” in S. T. Coleridge, Shorter Words and Fragments, ed. H. J. Jackson and J. R. de J. Jackson, 2 vols. (vols. 10 and 11 of The Collected Works of Samuel Taylor Coleridge [London: Routledge, 1995]). See also his Stateman’s Manual Appendix B, in The Complete Works of Samuel Taylor Coleridge, ed. W. G. T. Shedd, 7 vols. (New York: Harper & Bros., 1884), vol. 1, pp. 456-72. 36.S. T. Coleridge, Aids to Reflection, ed. H. N. Coleridge, 4th edition (London:

325 William Pickering, 1839), p. 164 37.Ibid., p. 157. 38.Ibid., p. 163. 39.Ibid., p. 163. 40.Coleridge, Logic, p. 82. 41.See the passage cited by Note 7, above. 42.Principles of Mathematics, p. 222. 43.F. H. Bradley, Appearance and Reality, Second Edition, p. 125; hereafter cited as “A & R”. 44.Cf. C. I. Lewis and C. H. Langford, Symbolic Logic, pp. 387-88. 45.F. H. Bradley, “Relations,” in his Collected Essays, 2 vols. (Oxford: Oxford University Press, 1935), vol. 2, p. 635. 46.George Berkeley, Principles of Human Knowledge, in his Works, ed. A. A. Luce and T. E. Jessop (seven vols., London: Nelson, 1948-1954), vol. 2, “Introduction,” ¶¶ 7-10. 47.Cf. Bradley, “Relations.” 48.Ibid., p. 636. 49.F. H. Bradley, “Association and Thought,” in his Collected Essays, vol. 1, p. 208. 50.F. H. Bradley, Essays on Truth and Reality, p. 231. 51.Russell elaborates this case in detail in his essay on “The Monistic Theory of Truth,” in his Philosophical Essays (London: Allen & Unwin, 1966). 52.Cf. F. Wilson, “Acquaintance, Ontology and Knowledge,” in this volume, above. 53.Cf. G. Bergmann, “The Revolt against Logical Atomism,”in his Meaning and Existence (Madison, WI: University of Wisconsin Press, 1960), pp39-72. 54.Russell more strongly rejects the nominalism of Berkeley and Hume, as well as that of Bradley; see his Problems of Philosophy (London: Williams & Norgate, 1912), pp. 94-97. 55.Cf. G. E. Moore, Some Main Problems of Philosophy (London: Allen and Unwin, 1962), pp. 389-90; G. Bergmann, Realism: A Critique of Brentano and Meinong

326 (Madison, WI: University of Wisconsin Press, 1967), p. 9, p. 43; R. Grossmann, The Categorial Structure of the World (Bloomington, IN: Indiana University Press, 1983), p. 167ff. 56.Cf. N. Wiener, “A Simplification of the Logic of Relations,” originally 1914, in Jean van Heijenoort, ed., From Frege to Gödel: A Source Book in Mathematical Logic, 1879-1931. (Cambridge, MA: Harvard University Press), pp. 224-27. 57.The Categorial Structure of the World, p. 164; see also his Ontological Reduction (Bloomington, IN: University of Indiana Press, 1973). 58.Cf. Bergmann, “Russell's Examination of Leibniz Examined,” in his Meaning and Existence. 59.See references in fn. 21. 60.G. Bergmann, New Foundations of Ontology, ed. W. Heald, with a Forward by E. B. Allaire (Madison, WI.: University of Wisconsin Press, 1992). 61.B. Russell and A. N. Whitehead, Principia Mathematica, Second Edition, 3 vols. (Cambridge: Cambridge University Press, 1927), vol. 1, p. 200. 62.Or rather, to use Begmann’s technical terminology, (n) is the circumstance that a is different from b. Bergmann distinguishes circumstances from facts. But for what we are about this difference makes no difference. 63.Cf. Wilson, “Acquaintance, Ontology and Knowledge,” in this volume, above. 64.On the issue of acquaintance vs. presentation, see also F. Wilson, “Effability, Ontology and Method,” in this volume, below. 65.This is Bergmann’s terminology. Some, e.g., Meinong, speak of things which merely are or have being, where things that are divide into the existent (or actual) and subsistent (or merely potential. These differences in terminology at times mark differences in ontology. For our purposes these differences make no difference. 66.It was later edited by E. Eames and K. Blackwell as B. Russell, Theory of Knowledge: The 1913 Manuscript (London: Routledge and Kegan Paul, 1992). The story of the lost manuscript and its reconstruction is given in the admirable Introduction by E. Eames, as is the account therein of its place in both Russell’s life and his thought. Russell rejects the simple relational view when he discusses Meinong on intentionality, pp. 107-108. There is an admirable and careful analysis of Russell’s discussion of the problem of belief in H. Hochberg, The Positivist and the Ontologist: Bergmann, Carnap and Logical Realism (Amsterdam: Rodopi, 2001), Ch. VI.

327

67.Russell, Theory of Knowledge, p. 109ff. 68.Ibid., p. 108. 69.The example here is adapted from B. Russell, Problems of Philosophy. 70.Russell, Theory of Knowledge, p. 110. 71.Ibid., p. 111ff. 72.Ibid., p. 111., p. 148. 73.Ibid., p. 108. 74.Ibid., p. 115. 75.Ibid., p. 142. 76.Ibid. 77.Ibid., p. 144ff. 78.Cf. Hochberg, The Positivist and the Philosopher, p. 124ff, p. 214f. 79.The Logical Atomism lectures are reprinted in Bertrand Russell, Logic and Knowledge, ed. R. C. Marsh (London: Allen and Unwin, 1956). 80.This essay, “On Propositions: What They are and How They Mean,” also is reprinted in Russell’s Logic and Knowledge, ed. R. C. Marsh. 81.See below, the essays on “The World and Reality in the Tractatus,” and “Grossmann on the Categorial Structure of the World.” 82.It is even more complicated than this suggests; see the diagram, ibid., p. 118, of how judgment works and note the complexity: surely the aboutness of thoughts is less complex than this! 83.Russell, Theory of Knowledge, p. 145. 84.B. Russell, Analysis of Mind (London: Allen and Unwin, 1921). On these issues, see the essays, below, on “The Aboutness of Thought” and “Effability, Ontology and Method.” 85.Hochberg, The Positivist and the Ontologist, pp. 214-216. 86. Compare Wittgenstein’s remark that

328 Instead of, ‘The complex sign “aRb” says that a stands to b in the relation R’, we ought to put ‘That “a” stands to “b” in a certain relation says that aRb.’ (Tractatus Logico-Philosophicus, trans. D. Pears and B. F. McGuiness, with Intro. by Bertrand Russell [London: Routledge and Kegan Paul, 1961], ¶ 3.1432.) 87.H. Hochberg, The Positivist and the Ontologist, p. 178, p. 215f. 88.Hochberg, The Positivist and the Ontologist, p. 178.

Eight Bareness, as in “Bare” Particular: Its Ubiquity* Many philosophers have argued that ordinary things are bundles of properties, where these properties are universals, entities able to be properties of more than thing. Consider, for the sake of simplicity, two red spots or images. The red in the one spot is, let us also suppose, the same as the red in the other spot. Thus, the two spots share a common property. This would seem to imply that they are the same entity. But they are two. It is therefore concluded that there must be other entities present, two of them, one in each spot. This accounts for there being two different things.1 This further entity is a particular, and, since in itself it has no properties, it is said to be, in itself, in its own being, “bare”: in so far as it is anything, that is, anything other than itself, it is so by virtue of its being with the properties that, together with it, constitute the ordinary things. In itself, it never ceases to be bare, but at the same time it never is naked – it always comes clothed, if you wish, by properties. Now, many philosophers have objected to bare particulars. Russell, for example, once argued that “One is tempted to regard ‘This is red’ as a subject-predicate proposition, but if one does so, one finds that ‘this’ becomes a substance, an unknowable something in which predicates inhere...”.2 How, Russell and others ask, could a good empiricist ever admit into his or her ontology these horrid little things? How could one actually believe that these little things populate our world? After all, you can’t even see them! What I wish to argue is that, after all, a bare particular is not such a horrid thing – that particulars are there in things, that they are bare but that such bareness both is to be expected and is innocuous, that such bareness is in fact ubiquitous, and that it not only harmless, but a central feature of the world of the empiricist. * This first appeared as “Bareness, as in ‘Bare Particular’: Its Ubiquity,” in H. Hochberg and K. Mulligan, eds., Relations and Predicates (Frankfurt: Ontos Verlag, 2004), pp. 81-112.

330 I

Begin with ordinary sensible things, red images, for example, and the properties of and relations among these things. William James was characteristically perceptive on these things. He carefully distinguished the properties of things and the relations among them. With an apt metaphor he likened the world of which we are conscious to the world of a bird’s life. “Like a bird’s life, it [the world as experienced] seems to be made of an alternation of flights and perching,” where the resting-places are “usually occupied by sensorial imaginations of some sort.”3 The flights are relations, the perchings are properties – relations among and properties of sensible things. As for the resting-places, James observes that “In the sensations of hearing, touch, sight, and pain, we are accustomed to distinguish from among the other elements the element of voluminousness.”4 He refers to the discussion found in James Ward, who refers to this element as “extensity.”5 James notes that “this element [extensity] [is] discernible in each and every sensation”; and comments that “extensity, being an entirely peculiar kind of feeling indescribable except in terms of itself, and inseparable in actual experience from some sensational quality which it must accompany, can itself receive no other name than that of sensational element.”6 Extensity is, it is clear, a distinguishable part of the things we experience. Each ordinary thing has, as an element within it, its extensity. It is there, upon the extensities, that perchings perch; and it is among these elements that flights take off and come to rest. Let us refer to the extensity of a thing like a red image as its “area.” The quality of redness as a property of the thing is a perching upon the area in the thing. And if one red image is to the left of another, then the relation of being to the left of is a flight that takes off from the area of the one thing and comes to rest on the area of the other. And all these elements are presented in our ordinary sensory awareness or experience of the world, the areas or extensities, the properties or perchings, and the relations or flights: all are admissible into any ontology based upon the empiricist’s Principle of Acquaintance.

331 II

Ontology is not, or ought not to be, at least for the empiricist, all dialectical. As Locke and Hume and Russell and William James argued, it ought to be rooted in ordinary concrete things, the sensible things with which we are acquainted in ordinary experience. But it is often stated, even by those with empiricist leanings, that bare particulars are introduced for dialectical reasons, by way of argument and not because they are presented in experience. Bergmann, who says he accepts the Principle of Acquaintance, once wrote that “I, of course, have convinced myself that I am actually presented with two things [two particulars in two images]. Yet I am loath to rest the case on this conviction, for I am convinced that a very major part of it is dialectical.”7 Just how has he convinced himself? If it is by looking, by virtue of his being aware of them in experience, then ‘convince’ is surely not correct: one accepts that red exists because it is given in experience, and for the same reason, it would seem, one should accept that (bare) particulars exist because they are given in experience. Being convinced consists of being given an argument that moves one from ignorance to justified belief. Of the obvious one need not be convinced. If you are confronted with one who does not know these entities, one who is not acquainted with them, then, one does not offer an argument but rather, as William James puts it, all I can do is “...say to my friends, Go to certain places and act in certain ways and these objects will probably come.”8 Berrgmann’s way of putting his point suggests that (bare) particulars are introduced into one’s ontology on dialectical grounds rather than the fact of acquaintance. But that is not what is demanded by the empiricist stance. Consider again our two red concrete objects, the two red images. We have this fact: the red in this image is indistinguishable from the red in that image. In this sense, there are two references to, two definite descriptions for, the same entity, that is, an entity which indistinguishably itself is in two things. It is for this reason that we can refer to this property and properties in general as “universals”. That the property in the one image is indistinguishable from the property in the other image is what is meant when we speak of them as the same property. That properties in things are in this sense the same accounts for why we apply the same predicate, namely ‘red’, to the two things. Given the traditional usage, this implies that properties are universals. As G. E. Moore once put it (as usual, in his somewhat convoluted way),

332 In the case of two sense-data, A and B, both of which appear to me to be red, I often cannot tell that the most specific shade of red which A presents to me is not exactly the same as the most specific shade which B presents to me. I also cannot tell that the most specific shade which A presents to me is not an absolutely specific shade. And I think I can see quite clearly that it is logically possible both that it is an absolutely specific shade, and that it does in fact characterize A and B.9

There is no argument to the effect that we need to construe properties as universals in order to account for why we apply the same predicate to different things. To the contrary, we do apply the same predicate to different things, and we do so on the basis of the commonsense fact that the property in the one is the same as the property in the other. It is this commonsense fact that leads us to say that properties are universals. Again, as Moore puts it, ... it is quite certain that many characters of concrete things are common characters, and also that many are not. And if ... we use the phrase ‘is a universal’ in a sense which logically implies ‘is a common character’, it follows, of course, that ... we shall have to say that many of the characters of concrete things are universals... 10

At the same time, the colour property of a green spot is clearly distinguishable from the red which is the property of another spot. The property in this case in the one spot is different from the property in the other spot. In this sense of ‘different,’ the area upon which the property red perches in the one image is distinguishable from and therefore different from the property red which is perched upon the other image. Moore notes the role of areas in determining the differing of things. ... there are cases in which I can distinguish between two concrete things, A and B (as, for instance, when I distinguish between two different parts of a sheet of white paper), although I cannot perceive that A is qualitatively unlike B in any respect whatsoever – either in shape, or size or colour.11

As we saw James, following Ward, making the point, that every sensible thing comes as a piece as it were of extension; there is an area which defines each thing. Thus, we have one image - one area, or one concrete thing - one area. Now, ordinary concrete things, images for example, are individual

333 things, we have this image and that. A concrete thing, a this or a that, is something complex. It has properties and these properties are with each other. An ordinary thing is thus a group of properties that are with one another. But it is not just a group of properties that are with one another: there is also the area that is in the thing. An ordinary concrete thing is thus a group of properties together with an area; and these entities are with one another forming the thing. Furthermore, an ordinary thing is not just a thing: as a group of entities that are with each other, the ordinary thing is a fact. An ordinary thing is a particular, but it is a particular fact. The qualities in the fact do not make it a particular fact, it is not by virtue of the qualities in that fact that it is distinguished from other facts. For, after all, the qualities in the fact, the properties of the thing, are universals. That which distinguishes the fact from other such facts is the area in the fact. It is the area which is the entity which makes an ordinary thing a particular. In that sense, the area itself may be called the particular which is in the particular fact which is the ordinary thing. Since things are wholes of which areas and properties are parts, and the properties are universal, the only entity that is unique in each concrete thing is the area. Areas are particulars, and as such they individuate concrete things.12 The case is not dialectical. The case is made in terms of that with which we are acquainted. Areas are there and these are the reasons why we take the two concrete things to be different, different particular things. E. B. Allaire turns it around: we make two references and therefore the particulars must be there – at least, so he argues. Allaire’s way of putting it makes it seem as if the dialectics are central. He asks us to consider two red discs. He then argues that To claim that both discs are but collections of literarily the same universals does not account for the thisness and thatness which are implicitly referred to in speaking of them as two collections. That is, the two collections of characters ... are, as presented, numerically different. Clearly, therefore, something other than a character must be presented.13

Not: something other is presented but: something other must be presented. But for one who accepts a Principle of Acquaintance what counts is what is presented. The dialectics are not there to convince one that one must be presented with certain things, but rather to convince one that entities which are in fact there, entities which are in fact presented to us, provide a solution to the traditional ontological problem of individuation.

334 They are presented, and these entities do in fact, we subsequently argue, solve/dissolve the traditional problem. Their role relative to the traditional problems is a matter of dialectics: they are presented in our sensible experience of the world, and because they do in fact exist we can appeal to them to solve/dissolve the traditional problems: they are not there because they must be.14 But areas, particulars, are never naked: they are always presented as with some quality or other. Here, we clearly have to distinguish an entity from facts involving that entity. In stating a fact about an entity one is saying something about that entity: one is stating what that entity is like, how it is characterized. These facts about the entity are things that can be said. However, the area really is just an area. We can say things about it; specifically, we can state what qualities are with it. But in itself it has no characteristics, and nothing can therefore be said about it, that is, said about it as it is in itself. In this sense, the particular is bare: it cannot be described, since there is about it, as it is in itself, nothing to describe. In itself, it cannot be described, it can only be named. If “to say something” is taken to mean “to assert a proposition”, then nothing can be said about the areas in things; that is why they are said to be “bare.” They are presented to us in our experience of things, and they can be referred to, but there is nothing sayable about them. To make the point again, however: while areas are bare, in the sense just explained, they do not come to us in experience as unclothed: they are not naked. They all occur as parts of ordinary things, that is, as having qualities and as standing in relations. As James put it, In minds able to speak at all there is, it is true, some knowledge about everything. Things can at least be classed, and the times of their appearance told. But in general, the less we analyze a thing, and the fewer of the relations we perceive, the less we know about it ... 15

Areas always occur as, and are always presented as, parts of facts.

335 III The empiricist admits entities into his or her ontology provided that they conform to the Principle of Acquaintance: admit no entity unless one is acquainted with it (or with entities of that kind). What we must recognize about the basic entities of the world is that they are in themselves wholly, or logically, or ontologically, self-contained. Acquaintance with them is thus mere acquaintance. Acquaintance with a quality or a relation or an area is thus not knowledge about. To be sure, we are acquainted with facts, with the bundles that are ordinary things. This provides us with knowledge about the entities in the facts that are thus presented. But mere acquaintance with the basic entities is dumb. James put the point in his usual telling way: I know the color blue when I see it, and the flavor of a pear when I taste it; I know an inch when I move my finger through it; a second of time, when I feel it pass; an effort of attention when I make it; a difference between two things when I notice it; but about the inner nature of these facts or what makes them what they are, I can say nothing at all.16

But there are philosophers who argue that our experience of sensible things is not in this way dumb. The point can be made in a simple way. It is argued that in order to know what the quality red is we must also know that it is not the quality green, and, more strongly, that red’s qualifying something is incompatible with green’s qualifying that thing. Thus, in knowing red one also, it is said, knows that (I) (x)[ red (x) e ~green (x)] So, on this view, when we know red in itself, we also know something about red, namely, that being red is incompatible with being green. (I) describes (part of) the being of red, and so is part of the meaning of ‘red’.: it is a metaphysical necessity. Acquaintance, then, always is, or involves, knowledge about. The pattern goes back to Aristotle. His metaphysical scheme is designed to provide a way of explaining sensible events. On his view, an ordinary thing is a substance. A substance has qualities present in it. Sensible events are the being in a substance of a sensible quality. Change consists in one quality ceasing to be in a substance followed by the coming to be in that substance of a different, and incompatible sensible quality. A substance is an individual, and, more particularly, an individual

336 that endures through change. Upon the metaphysics of explanation that Aristotle proposes, every substance, that is, every ordinary object, has a nature. This nature is metaphysically necessary to the being of the object; it defines what it is in its essence. This nature is a power, an active disposition, that moves the object in certain defined ways.17 Thus, for example, it is the nature of a stone to gravitate. To be grave is an active power. In exercising this power, the object moves itself.18 This power is such that if the object is unsupported then it moves towards the centre of the universe More generally, let “N” be the nature, “F” the occasion of its exercise, and “G” the end of its exercise. The nature relates F and G: (*) N(F, G) and this necesssary connection guarantees – so the argument goes – that All F are G i.e., that (@) (x)(Fx e Gx) Note that (*) can guarantee the truth of (@) only if it is also a necessary truth that (#) N(F, G) e (x)(Fx e Gx) Among the recent philosophers who have defended the view that laws are to be construed as relations among universals, D. M. Armstrong alone has faced up to this problem, or, what amounts to the same, the problem why the structural relation (*) should somehow imply that the regularity (@) among particulars should also hold.19 Armstrong considers the properties F and G to be causally related; this relation is R. We therefore have N(F,G) Armstrong is concerned to understand how it is that this fact about universals can lead to there being a regularity All F’s are G’s among the particulars that exemplify F and G. He proposes that ...N(F,G) is [itself] a universal, instantiated in the positive instances of the law...

If we accept this view, he suggests, then

337 ...it will be much easier to accept the primitive nature of N. It will be possible to see clearly that if N holds between F and G, then this involves a uniformity at the level of first-order particulars (p. 88).

We have, therefore, certain facts about particulars exemplifying the (funny) universal “N(F,G)”. These facts Armstrong understands (p. 91) to be perspicuously represented by something like (N(F,G)) (a’s being F, a’s being G) which we might equally well write as (&) (N(F,G)) (Fa, Gb) which makes clear that what we have is a relation holding between states of affairs. It is the “causal relation”, but because it holds between states of affairs rather than particulars, its status is rather more like that of a connective than it is like an ordinary first order relation among ordinary particulars. This is the first feature of the causal relation that we should notice. Second, we should notice this connective is not truth-functional; the sentences ‘Fa’ and ‘Gb’ that occur in (&) do not by themselves determine the truth-value of that sentence. In particular, third, the complex sentence (&) will be true only if the simple sentences appearing in it mention kinds that are causally related to each other. Finally, fourth, it is worth noticing that the nature of the causal relatedness will vary from case to case. It may be the simple kind involved in a regularity like All F’s are G’s or it may be a more involved connection such as All F’s that R an H are either G’s or K’s In the former case we will have N(F, G) while in the latter case we will have N[(F,R,H), (G, K)] or something like it. (Armstrong does not give details.) This will make “N” an odd relation indeed, a relation with a varying number of terms. Or perhaps there will be a variety of relations all sharing a common property, that of being “causal necessitation relations.” How these themes are worked out by Armstrong or other defenders of the view we may leave to one side. (In general they aren’t worked out.) It suffices to note only the requirement that the nature of the causal relation among the states of affairs will in fact have to vary from case to case depending upon the relevant law. These are the four crucial features of the nomological connective.20 These four crucial features give rise to two philosophical problems. The first of these is the fact that when one introduces a nomological con-

338 nective such as that which appears in (&), one challenges the traditional analytic/synthetic distinction, which presupposes a language in which the only connectives are truth-functional. This problem amounts to saying in another way that there is a real question of the entailment relation that is supposed to hold between N(F,G) and All F’s are G’s or, what is the same, the notion of logical form that is supposed to make (#) a necessary truth. The second philosophical problem concerns how it is that we know that the non-truth-functional nomological connective actually holds between two facts or states of affairs. Here we have the statement about the nature “N” of things, which states that if something is in situation F, then, if this thing has nature N then, and only then, it is also G, where F and G are properties given in ordinary experience, either sensible experience or inner awareness: (D) (x)[Fx e (Nx = Gx)] We explain the behaviour of an object by appeal to its nature. This nature is active: the model is that of human volition. Thus, for Aristotle, all objects are active in the sense in which human beings are active, though some, e.g., human beings or dogs, are more active than others, e.g., stones. To say that they are more active is to say that they have more powers, more complex natures. Since the powers are active, modelled on human activities, they are powers the exercise of which is towards an end. The prescientific explanations of Aristotle and his successors such as Ptolemy are thus purposive; every explanation is a teleological explanation. In the case of stones, the purpose or end at which the stone’s activity is aimed at achieving is being at the centre of the universe. The activity is as it were constant. But it is not always exercised. The stone is constantly striving to be at the centre of the universe. But sometimes it is prevented from moving towards that end. Thus, if I hold the stone up at the top of the tower, I am preventing it from moving towards the centre of the universe. That tendency I feel as the weight of the stone. If the impediment is removed, if the stone becomes unsupported, then the tendency will manifest itself in the properties of the stone, it will in fact change places as it moves itself systematically towards the centre of the universe. In an Aristotelian world the patterns among sensible appearances that derive from the underlying natures of things are not universal: they are

339 gappy. The nature, that is, the “N” of (D), is not given to us in sense experience. It is rather, Aristotle argued, given to us in a rational intuition. For Aristotle, reason is what grasps the reasons for things, and the reasons for things behaving as they do are their natures. Reason, then, for Aristotle, provides us with special insight into the metaphysical structure of the world. This notion of “reason” is very different from that of the empiricists, according to whom reason aims to discover genuine matter of fact regularities, universal and exceptionless patterns of behaviour. Reason, on this empiricist alternative, does not aim at insight into metaphysical structures but was a human instrument that restricted itself to the world of sense experience, endeavouring to discover exceptionless patterns of behaviour of objects.21 In (D), the “F” and “G” are features of the object known in sense experience. Since (D) relates the nature N to these features of sense experience, where N is not given in sense experience, it follows that (D) is not itself an empirical truth, something the truth of which can be discovered in sense experience. We discover its truth not by observation but by reason, that is, the reason that grasps the natures or reasons of things. A statement such as (D) which relates a nature to the empirically observable occasion and end of its exercise is metaphysically necessary. As for understanding the natures of things, this is done, according to Aristotle, by giving a real definition of the nature. The nature is a species, and the species is defined by giving its genus and specific difference. Thus, in the case of human beings, the nature is “human” and the real definition is given by “rational animal”, where “animal” is the genus and “rational” is the specific difference. The real definition is exhibited in a syllogism: All M are P All S are M All S are P “S” and “P” are the subject and predicate of the conclusion, and “M” is the middle term that joins them in the premises. When the syllogism exhibits a real definition, “S” is the species, “P” is the genus, and “M” is the specific difference. Thus, the real definition human is rational animal is exhibited in the syllogism All rational are animal All human are rational All human are animal

340 In the case of stones we would have All centre loving objects are material All stones are centre loving objects All stones are material Syllogism is thus not only a form of argument but also a logical structure that exhibits the metaphysical structure of the world. It reveals the complex structure of the active dispositions or nature of an object. It reveals, in the genus, those dispositions which the nature shares with other objects, and, in the specific difference, it reveals those dispositions which distinguish it from other sorts or species of object. Thus, for Aristotle and his successors, understanding the natures of things consists in grasping the ways in which they are similar to and differ from other sorts of things. Explanation consists in grasping similarities and differences among things.22 That is not, however, the point that here needs to be emphasized. When we know an Aristotelian nature we not only know it as it is in itself but in knowing it as it is itself we have knowledge about it: (D) gives the meaning of the term ‘N’. Knowledge is always knowledge about: substances are not bare entities. IV Aristotle makes the ontological structure of the world a matter of necessity. This was taken further by idealists such as F. H. Bradley. According to Bradley all knowledge is knowledge about. Like the empiricists, Bradley argues that knowledge is rooted in experience. But, because all knowledge is knowledge about, the role of feeling in Bradley’s philosophy, specifically the role of feeling in Bradley’s ontology/epistemology, has a very different status and role from that of the feeling = sensation of the empiricists.23 The latter is indeed “mere” feeling, from Bradley’s point of view, and from the empiricist position too: such knowledge of the properties in things is dumb, it involves no knowledge about those things, nothing that can be said. However, that feeling which plays a central role in Bradley’s philosophy is anything but mere. On Bradley’s view, a content is ideal if it falls short of perfect reality. Now, the real is the fully individual or particular; as he puts it, “Nothing in the end is real but the individual...”.24 This doctrine, that in order to be real an entity must be individual or particular, is applied in particular to relations: his account of relations must fulfil the condition of construing them as particulars. “A relation, to be experienced and to be actual, must

341 be more than a mere abstraction. It must be an individual or particular fact, and, if less than this, it cannot be taken as itself.”25 Thus, an ideal content that falls short of full reality falls short of individuality or particularity. It is therefore abstract rather than concrete, general rather than singular or individual or particular. Further, the particularity of a thing derives from its relations to other things. The this – this physical object, this sensation, this red, this colour – is what it is only because it not that. This thing itself is a particular only to the extent, then, that is an aspect of a larger relational whole. In itself it has less particularity, less reality, than the relational whole of which it is an aspect. The fully real is the relational whole that includes all other things as aspects of its own reality. The judgment that “This is a such” brings together the subject “this” and the predicate “such”. This “this” is isolated from other things, but when the “such” is brought over against it and affirmed of it, that is, said to be part of the whole which is the “this”, we in fact particularize the thing by bringing it into relations with other things: the “such” carries within itself relations, at least those of similarity and dissimilarity. And, with those relations, the judgment points to other things, other “thats” which are also such “suches”. Bradley’s account of judgment is not terribly different from that of the Aristotelian. A judgment of the form S is P can be justified, according to Bradley, by forming an argument or, rather, inference M is P S is M S is P S implying M, implies P.26 The middle term M links together the S on the one hand and the P on the other. It as it were fills out the copula in “S is P”. The judgment itself refers to a reality that links S and P, but in the judgment taken alone that reality is ignored. In the inference that background context becomes explicit: the conditions that were previously external to the judgment are internalized. The connection between S and P which was external to the judgment is internal to the inference. Where S and P were unconnected, they are now connected: the being of the one becomes implicated with the being of the other. They are now no longer simply external to one another; they are connected in their very being, internally. In this internalization, the ideality of the judgment is decreased. At the same time, the contingency of the judgment is decreased. In the judgment the terms are separate, their connection (or, rather, “connection”) is contingent. As the inference fills out the judgment the separateness of the terms is decreased, and therefore the contingency. In the inference we be-

342 gin to grasp the structure that constitutes the necessary ties that link the terms of the judgment into a unified Whole. As ideality is decreased in the inference, so is the contingency; or, conversely, as the inference more closely approaches reality, so does it approach a complete necessity. In perception we isolate a portion of reality: “Lo! an S”.27 In judgment we locate the perceived S as a P. In such a judgment we separate the P from the S. In inference we proceed to fill in the context in which the S which is a P is located. The result is the location of the S which is P in the larger part of reality constituted by the M which links them. Now, one of the criticisms of the claim that coherence is the criterion of truth is that coherence, like consistency, is as compatible with falsehood as it is with truth. This is so even if one begins with perception, which must be an isolation of part of the total reality. Bradley avoids this problem by insisting that beyond perception there is primitive experience or feeling: in feeling we encounter Reality, or, rather, in feeling Reality is fully present, not present to us, but including us within the whole – in feeling or sensation, primitive awareness of the way things are, the distinction between sensing and the sensed, the act and the object, disappears. In the mere isolating sensation of the empiricists, we separate parts of this whole. As the empiricists see it, James among them, sensation, feeling, is indeed isolating, but, moreover, the entities known in such acquaintance are what they are, independently of any other entity. For Bradley, in contrast, sensation, in the sense in which it distinguishes the object sensed from other objects and also from the sensing, is indeed isolating, but that is not the end of the story. What is separated is also in itself connected to other entities. Thought moves from sensation through inference, into perception and then into judgment, and in so moving, it moves from the full reality present in feeling to ideality. In inference we gradually move to restore the lost unity. As we fill in the structures in inference, we gradually lessen the separateness of the things that are first given to us in sensation, perception and judgment. And in the ultimate judgment, or, what amounts to the same, the ultimate inference, we discover the whole truth that we previously felt but lost in sensation, perception and judgment. Only, it is not “previously”: the feeling is there, with us, all the while. All the while in feeling we encounter the reality that includes us and to which we are in thought striving to return. Thus, “judgement, on our view, transcends and must transcend that immediate unity of feeling upon which it cannot cease to depend.”28 Reality is thus both the origin of the movement of thought – reality

343 as feeling – and the goal towards which thought moves – reality as selfconscious awareness of the manifold of structures which are implicit within itself. And more: reality is the structure that guides thought as it moves from feeling to the total self-conscious awareness which is its telos. We accept the idealizations in judgment not because they are true – for they are not wholly true – but because we have a sense that they can be made true, and, indeed, that they can ultimately be made true. In feeling consciousness already implicitly recognizes its goal, the complete structured unity of which it is a part and which is at once the end and the guide towards that end. If at any point there were a genuine separation of knower from known or of entity from entity, or of this from such, of this from that, then there an ultimate re-union could not be achieved: no re-union without union. Bradley’s argument for this position consists in the claim that it can, where empiricism cannot, account for the soundness of inference. Upon the empiricist account of inference as defended by Locke, Hume, the Mills, and James, what we know is what is given in sensation, and what is given in sensation are entities that are intrinsically separate and isolated, in their being not related to other things. Or rather, insofar as they are related, it is only psychologically. What unity that is there is provided by the mind that judges them; objectively, however, in the entities themselves there are no connections.29 This is what Russell was later to refer to as the monadistic account of relations.30 James characterizes it as “sensationalism”. These philosophers “deny the reality of relations,” and “the upshot of this view” is that what we experience is a world consisting of ...sensations and their copies and derivatives, juxtaposed like dominoes in a game, but really separate, everything else verbal illusion.31

Bradley proposes that genuine relations are incompatible with the independence that is a consequence of monadistic view. “...a mode of togetherness such as we can verify in feeling,” he tells us, “destroys the independence of our reals.”32 Conversely, if we do make the relata independent or absolute, then we destroy their relatedness: “Relations are unmeaning except within and on the basis of a substantial whole, and related terms, if made absolute, are forthwith destroyed.”33 The point that Bradley makes is that in the absence of any objective connections among entities there is no objective ground for the soundness of inference. Judgments are justified by inferences, and the latter can do their job only if they are grounded in objective necessary connections among things.

344 Judgments are clearly not themselves primitive feeling. They are not even the feeling that initially, in the growth of knowledge, isolates from the whole sensible parts. Perception unifies these sensible entities into larger wholes, and judgment develops that process further. There is a continuity of thought from primitive feeling, through isolating sensation, through perception, through judgment, to the cognitive end where the Whole is wholly conscious of itself as a unity of diversity. For our purposes, the point is that Bradleyan judgments, the inferences that trace out necessary connection, are not sensations, or, at least, not just sensations in the empiricist sense. Nor, according to Bradley, are relations given in sensible experience. Bradley is thus among those whom James characterizes as “intellectualists”. These philosophers are ...unable to give up reality of relations extra mentem, but equally unable to point to any distinct substantive feelings in which they were known, have made the same admission [as the sensationalists] that the feelings do not exist. The relations must be known, they say, in something that is no feeling, no mental modification, continuous and consubstantial with the subjective tissue out of which sensations and other substantive states are made. They are known, these relations, by something that lies on an entirely different plane, by an actus purus of Thought, Intellect, or Reason, all written with capitals and considered to mean something unutterably superior to any fact of sensibility whatever.34

James agrees with the intellectualists that the sensationalists are wrong in holding that reality consists of unrelated sensible elements. Besides perchings, there are also flights. James disagrees with the empiricists, the sensationalists, in holding contrary to the latter, that reality consists of sensible elements which are related one to another in sensible experience. These relations are given in our ordinary experience of things, rather than being all of them necessary connections that are known only by acts of thought – Thought – that is a form of knowing higher than, and different from, our ordinary sensible experience of things. V

Bradley was not the first so to argue that the structure of things is given in non-empirical judgments of necessary connection.. The pattern is Aristotelian. Thus, the 17th century English Aristotelian John Sergeant argued, in

345 his Method to Science,35 that science, understood in empiricist fashion, as based in sensation, cannot achieve a genuine unity, and therefore leaves things unexplained. ...Matter of Fact shows evidently, that this Method [that of experiment], alone, and Unassisted by Principles, is utterly Incompetent or Unable to beget Science. For, what one Universal conclusion in Natural Philosophy, (in knowing which kind of Truths Science consists) has been demonstrated by Experiments. ...it is ... merely Historical, and Narrative of Particular Observations; from which to deduce Universal Conclusions is against plain Logick, and Common Sense (unpaginated, d4).

Genuine science, in contrast, requires the grasp of objective necessities that tie things together into wholes. In order genuinely to understand things, this objective structure must be grasped. ...’tis Connexion of Terms which I onlely esteem as Proper to advance Science. Where I find not such Connexion, and the Discourse grounded on Self-evident Principles, or (which is the same) on the metaphysical Verity of the Subject, which engages the Nature of the Thing, I neither expect Science can be gain’d, nor Method to Science Estalbish’d (ibid.).

In fact, Sergeant, like Bradley, argues that judgment ultimately refers to a reality implicitly mentioned in the copula. Sergeant argued that “There is but one onely Notion that is perfectly Absolute, viz. that of Existence, and all the rest are in some manner or other, Respective...” (p. 15). We begin with the notion of being or existence and subdivide it according to species and difference, as Porphyry showed. Differences are successively added to genera to create ever more inferior species. The species most inferior to the supreme genus are individual things. ... every individual Man is but One Ens or thing; since he descends Lineally from that Common Head by intrinsecal Differences of more or less, which constitute him truly One in that Line; that is, one Ens, or one Thing (p. 32).

At the other end of the scale, the supreme genus is that of being, which admits of no definition in terms of genus and difference. ...the Notion of is, or Actual Being, is impossible to admit any Explication... (p. 120).

But if being is the supreme genus it is also that which contains within itself

346 as the source the being of all inferiors. If it is the supreme genus it is also the most determinate being, the most “fixed”. As the source of all being, of all reality, being is that which links its own determinations into determinate wholes. The Notion of is is the Determinate of its own Nature, and so most Fixt of it’s self; and, therefore, most proper to fix the Judgment (p. 120).

Being “fixes” judgments by providing the linkage represented by the copula: ...the meaning of the word is which is the Copula, is this, that those Words are Fundamentally Connected in the same Thing and Identify’d with it Materially; however those Notions themselves be Formally different, provided they be not Incompossible....As when we say a Stone is Hard the Truth of that Proposition consists in this, that the Nature of hard is found in that Thing or Suppositum call’d a Stone, and is in part Identify’d with it; however the Notions of Stone and Hard be Formally Distinct. Or, (which is the same) it is as much as to say, that that Thing which is Stone is the same thing that is Hard (p. 119).

Thus, “This Proposition Self-Existence is Self-existence is, of it self, most Supremely Self-Evident.” (p. 133). This proposition, which is the same as the propositions that “what is is” and “existence is existence”, contains within itself all other predications: “...not only the Notion of the Copula, but of the Subject and Predicate too, is Existence” (p. 134). Being, of course, is God Himself. As Sergeant puts it, “...God himself has expressed his own Supreme Essence by this Identical Proposition, Ego Sum qui Sum...” (p. 145). Our primary awareness is an awareness of being: “...the Notion of Existence is imprinted in the Soul before any other in priority of Nature” (p. 15). But this being of which one is aware is the being which constitutes the objective order of things. Thus, the connection between things is on the one hand an act of judgment while, on the other hand, is an objective connection in things. There being ... a Real Relation between those Notions which are the Subject and Predicate, the latter being really in the understanding and That which is said of the Former, and the Former that of which ’tis said; and Relation being necessarily compleated and actually such, but the Act of a Comparing Power; it follows, that every Judgment is a Referring or Comparing one of those Notions to the other, and (by means of the Copula) of both of them to the same Stock of Being on which they are engrafted, or the same Ends; where they are Entatatively

347 Connected (or the same Materially) before they are Seen or Judg’d to be so by our understanding (p. 121).

This awareness of being is, of course, much of a piece with the primitive feeling of Bradley’s metapyhysics, the primitive feeling which has incorporated within itself Reality.

VI Locke, in his Essay concerning Human Understanding,36 argued against Sergeant’s account of knowledge. The necessary connections that Sergeant supposed to be there are in fact simply not to be seen. It is evident, Locke says, that we do not know the necessary connections required for an Aristotelian understanding of why parts of things cohere (Bk. IV, Ch. iii, sec. 26, p. 526ff). But even if we knew why the parts cohere, we still would not know everything necessary for a grasp of the notion of the thing in Sergeant’s sense. For the notion must account for all the causal activities of the substance of which it is the notion, insofar as these activities are not merely occasional. Now, the regular activities of external substances include the production of the ideas of the secondary qualities, that is, the production of the simple ideas red, sweet, and so on. For these activities to be knowable scientifically, in Sergeant’s Aristotelian sense, regularities revealed by sense about such activities must be demonstrable by syllogisms grounded in notions. But for that to be possible, there must be necessary connections between red, sweet, etc., and the notions or natures of the substances that cause these qualities to appear. These necessary connections must be both ontological, in the entities themselves, and epistemological, giving us, when in the mind, scientific knowledge of those entities. But, Locke argues, we grasp no such connections: ’Tis evident that the bulk, figure, and motion of several Bodies about us, produce in us several Sensations, as of Colours, Sounds, Tastes, Smells, Pleasure and Pain, etc. These mechanical Affections of Bodies, having no affinity at all with those Ideas, they produce in us, (there being no conceivable connexion between any impulse of any sort of Body, and any perception of a Colour, or Smell, which we find in our Minds) we can have no distinct knowledge of such Operations beyond our Experience; and can reason no otherwise about them, than as effects produced by the appointment of an infinitely Wise Agent, which

348 perfectly surpasses our Comprehensions.... (IV, iii, 28, pp. 558-9; see also IV, vi, 10, pp. 384-5).

Properties are perceived to be just as they are, in themselves; to know them as they are we need not know any of the relations in which they stand to other entities. ... the immediate perception of the agreement or disagreement of identity being founded in the mind’s having distinct ideas ... affords us as many self-evident propositions, as we have distinct ideas. Every one that has any knowledge at all, has as the foundation of it, various and distinct ideas: And it is the first act of the mind (without which it can never be capable of any knowledge) to know every one of its ideas by itself, and distinguish it from others. Every one finds in himself, that he knows the ideas he has; that he knows also, when any one is in his understanding, and what it is; and that when more than one are there, he knows them distinctly and unconfusedly one from another (IV, viii, 2).

Locke’s appeal to an empiricist’s Principle of Acquaintance is clear.37 The conclusion that Locke draws is that account of knowledge and of syllogism that Sergeant developed is not sound: we cannot erect the edifice of knowledge on the proposition that “what is, is”: ... all purely identical propositions.... obviously, and at first blush, appear to contain no instruction in them. For when we affirm the said term of itself, whether it be barely verbal, or whether it contains any clear and real idea, it shows us nothing but what we must certainly know before, whether such a proposition be either made by or proposed to us. Indeed that most general one, “what is, is,” may serve sometimes to show a man the absurdity he is guilty of, when by circumlocution, or equivocal terms, he would, in particular instances, deny the same thing of itself; because nobody will so openly bid defiance to common sense, as to affirm visible and direct contradictions in plain words; or if he does, a man is excused if he breaks off any farther discourse with him. But yet, I think, I may say, that neither that received maxim, nor any other identical proposition teaches us any thing: And though in such kind of propositions, this great and magnified maxim, boasted to be the foundation of demonstration, may be and often is made use of to confirm them; yet all it proves amounts to no more than this, that the same word may with great certainty be affirmed of itself, without any doubt of the truth of any such proposition; and let me add also, without any real knowledge (IV, vii, 4).

So much the worse for the sort of reason that Sergeant defends: the world in which we live is simply not one in which there are any of the objective necessities that that account of reason supposes are there.38

349 VII

Russell made the same point against Bradley as Locke made against Sergeant. Bradley’s account of relations requires the introduction of a third particular, the Whole, over and above the two entities that stand in relation to each other.39 This relation is such that the one entity so related cannot be distinguished as itself independently of its necessary connections to other entities – connections which are necessary because they define the very being of the entities related. But Russell argues, with Locke and James, that entities – “thises” and “suches” – can in fact be identified as themselves without reference to the relations in which they stand to other qualities and other things. As Russell puts it: To say that two terms which are different if they were not related, is to say something perfectly barren; for if they were different, they would be other, and it would not be the terms in question, but a different pair, that would be unrelated. The notion that a term can be modified arises from neglect to observe the eternal self-identity of all terms and all logical concepts, which alone form the constituents of propositions. What is called modification consists merely in having at one time, but not at another, some specific relation to some specific term; but the term which sometimes has and sometimes has not the relation in question must be unchanged, otherwise it would not be that term which has ceased to have the relation.40

Note that here Russell is allowing Bradley’s point against the monadistic account of relations. On the latter, the predication of one term of a relation would not change if the other relatum ceased to exist.41 Russell accepts this criticism; he accepts that the monadistic account of relations is mistaken, and that there are, objectively, genuine relational unities. What he is denying is the implication of Bradley’s own account of relations that there is something about properties or qualities as presented that requires us when we are identifying them to refer as a matter of necessity to other properties, those to which they are necessarily tied. In order to know the property red it is not necessary to know the principle (I) that red differs from and excludes green. Russell is holding that properties are presented to us as logically and ontologically self-contained rather than as necessarily tied to one another; he concludes that there are no such necessary connections. But such connections are required by Bradley’s account of relations. The falsity of the latter view follows. Russell’s rejection of Bradley’s account of relations on the basis of an appeal to Locke’s empiricist Princi-

350 ple of Acquaintance is evident. James makes much the same point as Russell. He argues that All the elementary natures of the world, its highest genera, the simple qualities of matter and mind, together with the kinds of relation that subsist between them, must either be not known at all, or known in this dumb way of acquaintance without knowledge-about.42

The basic entities are what they are independently of their relations to other things: all knowledge about presupposes knowledge by acquaintance. Michael J. Loux43 is among those who have objected to the doctrine that there are among the constituents of things, entities whose only role in one’s ontology is that of individuating, grounding the particularity of ordinary things. This, he suggests, is what it means to say that these entities are bare. Loux objects to such entities: “in themselves, they have no properties at all, so that they cannot be the object of any kind of cognitive act,”44 and elsewhere he says that “the notion of a bare particular is epistemologically suspect”: Since bare particulars ... are essentially unknowable, since they are lacking in all characteristics, they cannot be experienced, nor can they even be conceived.45

On this doctrine, an entity can be the object of a cognitive act only if we cognize it through its properties. This is the doctrine of Sergeant, that to know a thing is to know its definition. For Sergeant, this is to know its species, and to know that in turn requires us to know the genus and specific difference. To know its genus and specific difference is to know how it is the same and different from other entities. Bradley argues the same thesis as Sergeant: to know a thing one must know its relations to other things, and in particular the relations of sameness and difference. Locke and Russell and James argue otherwise: when we are presented with a thing we thereby know it as it is, and in particular to know it we do not need to know its relations to other things. Thus, in order to know we do not need to know its species or its genus or any other property that it might have or to which it might be tied. An entity for which this is true is, as Loux says, bare. Locke and Russell and James are thus arguing on the basis of the empiricist’s Principle of Acquaintance that all presented entities are bare. In other words, it is not just particulars, individuators, that are bare. So are properties. And so are rela-

351 tions. For the empiricist, all basic entities are bare: bareness is ubiquitous. The same point can be put another way. If, as we have suggested, to say something is taken to mean to assert a proposition, then with regard to the basic entities of the world, be they particulars in things or the qualities of things or the relations among things, we cannot say what they are. Their being, what they are in themselves, cannot be expressed in a proposition. They can only be named, not said. Or, rather, as Locke saw, if it be said, as in, for example, “this is this”, the proposition in which it is said is trivial and verbal.46 Russell could make the same point. So could James. Since the being of the basic entities, what they are in themselves, can only be grasped in perception and not said, it is evident that such entities are ineffable. Bradley, too, has ineffable entities, or, rather, an ineffable entity. This is Reality as such, the Whole or the Absolute. To say something is to express a judgment, and a judgment S-P is always ideal, partially false, at least insofar as it requires us to separate the subject S and the predicate P. We achieve the truth, the whole truth, when we abolish the distinction between subject and predicate, when we grasp the ultimate unity which, precisely because it is a unity, cannot be said but only felt or experienced. It is the ineffable. The difference between the ineffable in Locke (or Russell or James) and the ineffable in Bradley is that for Locke (and Russell and James) the ineffable is located in ordinary experience, whereas for Bradley it is located either as it were below ordinary experience, in mere feeling, or above ordinary experience, ordinary perception, in Absolute consciousness, the consciousness which the Whole, the Absolute, has of itself. Furthermore, even though for Locke and Russell and James the basic entities that constitute ordinary things are ineffable, it does not follow that nothing can be said about them. To the contrary: to say that the entities are bare and to say that they are ineffable is to make the same point. But to say that they are bare is not to say that they are presented devoid of properties, and devoid of relations. It is clear from Locke and Russell and James, and from acquaintance itself, that we are always presented with complexes, with facts, and not with solitary entities, entities somehow in total isolation from each other. To the extent that these entities do stand in various relations to other entities, things can be said about them, namely, such things as that this is next to that or that this has such and such a property. Bradley’s ineffable entity, however, stands in relation to nothing: all other entities lose their own being within its enfolding totality, its smothering wholeness. For Bradley, nothing can be said that is wholly true. For Locke

352 and Russell and James, in contrast, there are many things that can be said that are not just true but wholly true. What can be said, and truly said, is that things stand in various relations to each other. It is just that the intrinsic being of these entities, what they are in themselves, is not constituted by those relations to other entities. As we saw, Russell and James allow, with Bradley and against Hume and the Mills, that there are objective relational structures. What they reject is that these objective connections are necessary to the intrinsic being of the entities that they relate. To put it another way, what Russell and James are arguing is that there are connections in the world of the empiricist but these are not essential. In this sense, the entities of Locke's world are all separable, though not in fact separate. This is in contrast to the monadistic account of relations on the one hand and Bradley’s account on the other. On the former account, things are not only separable but separate. On Bradley’s account, things are not only not separate but also not separable: the relations that structure them into unities are necessary, defining the intrinsic being of the entities related. For Russell, however, while entities are indeed structured by relations into unities, the related entities are separable in the sense that the relations are not necessary, not essential to the being of the things related. It follows that for empiricism, and specifically empiricism as developed by Russell and James, because none of the relations in which things stand are essential, reason cannot consist in the grasp of essential truths. In this respect, then, Russell agrees with Locke and the other empiricists such as James that the soundness of inference does not consist in the grasp of objective necessary connections. Thus, understanding is no longer the grasping an entity that provides an underlying unity to the apparently separable. It is, rather, the recognition of things as falling under certain general patterns, universal and exceptionless but contingent regularities, that hold among the logically and ontologically separable entities of experience.47 And reason, reason that grasps the reasons of things, is no longer the grasping of an entity that unifies things understood within itself, but is rather the judging that certain universal but contingent patterns obtain among things.

353 VIII

Having just argued that all basic entities are bare, it needs to be qualified. Bare they may be, but they are not quite naked. Thus, in experience qualities are qualities and not relations, while relations are relations and not qualities. These are two different forms of being. Those who do not begin clearly with the empiricist Principle of Acquaintance are sometimes inclined to deny this fact. Such a one was Frege. Properties, he argues, are indeed among the objects (“things”) in the world.48 But he also construes predication on the model of functions in mathematics.49 His basic model for predication is given by mathematical formulae like (1) 22 = 4 On this model, the sentence a is red that is, (2) red(a) is not in itself complete. “a is red” is an instance of the function red (x) just as (3) 22 is an instance of the function (4) x2 This has two difficulties. First, if sentences like (1) are basic, then, as we said, expressions like (2) are not complete, no more than expressions like (3) are complete sentences. Sentences like (1) represent a particular mapping by the function “x2” of the number 2 onto the number 4. On this model, expressions like (2) are incomplete in the sense of representing a mapping of one thing, a, onto something, without indicating what that something is onto which the object a is mapped. The complete sentence would have a form similar to the form of (1): red(a) = ... But what is it that the function “red(x)” maps the thing a onto? Frege argues that it is the True:50 red(a) = T Or, perhaps, it is the False. The problem here is that the True and the False are two monsters, at least from the empiricist perspective: they are certainly not given in any way in any sensible experience of the world. This is

354 one difficulty of Frege’s position. The other is the fact that a mapping is a relation. The function (4), for example, represents a relation that connects the number 2 on the one side to the number 4 on the other side. As a function it is a relation with particular properties. Specifically, it is one-one or bi-unique, and is therefore a definite description, or, rather, expressions such as (3) are definite descriptions. But for all that a function is still a relation. This makes qualities like red into relations. In our experience of things, however, we clearly distinguish qualities like red, on the one hand, and relations like, for example, next to on the other hand. Any language which would perspicuously represent differences in the world would therefore represent relations in one way and qualities in another way: the different objective forms of these entities would be represented by different logical or grammatical forms in language. In this way, if we take Frege to be constructing a perspicuous language – and what else could a begriffschrift be? – then to the extent that he assimilates qualities to relations, ignoring the difference of these things in the world, – to that extent his proposed language fails to be perspicuous – fails, in other words, to be adequate as a begriffschrift. If our argument is correct, then any (basic) relation is bare, but it always has the property of being a relation. This is a property shared by all relations: it is the highest genus among relations. As the highest genus it is represented in a perspicuous language by the grammatical or logical form of the expressions used to refer to specific relations. For that reason each relation is said to have the logical form of being a relation. Now, the same point applies to areas. Areas, that is, the entities that we have decided are particulars which, since the rule is one area - one image, individuate concrete things. Each area is a particular or individual, and has the property of being an area. In a perspicuous language we customarily represent the presence of a particular in a fact by labelling it with the subject term of the sentence expressing that fact. Names of individuals or particulars share the grammatical or logical form of being subject terms. This is usually represented in a perspicuous language by having a common form, e.g., lower case letters from the beginning of the alphabet. This grammatical or logical form of the name of an area represents that the thing so named has the property of being an area. Since an area is a particular, this grammatical or logical form is said to represent that each particular has the logical from of particularity. G. Bergmann is such a one. On the one hand, he specifically identifies areas as particulars – bare particulars.51 On the other hand, he argues that particu-

355 lars have the logical form of particularity.52 What we are arguing, then, is that there is nothing particularly mysterious about the notion of particularity: it is simply the property of being an area. Areas – particulars – are in facts, ordinary things, together with the properties that are with them. These properties come in various genera – red is a colour, B-flat is a tone, etc. They are all, however, to be contrasted to particulars: they can occur in more than one concrete thing. Since each property is a universal, Bergmann refers to the common property that picks them out as universality.53 This is represented in language by making the terms which refer to properties have the grammatical or logical form of occurring in the predicate spot – pictorially, these names of properties are taken from the set, say, F, G, H, ... There is indeed such a common property. It is not, however, a property parallel to the property of particularity. The latter is an affirmative concept. In contrast, that which all properties have in common is that each is not a particular. Red is a colour, but what makes it a universal is that fact that it is not a particular, that is, not an area: colour is a positive concept, universality is negative.54 Bergmann makes particularity and universality as logical forms with much the same status – he misses the point that one is positive and the other negative. It has also been claimed that universals have the property of being recognizable or re-identifiable and that this property is lacking for particulars. Thus, Allaire has suggested that “individuals [bare particulars] are merely numerically different from each other and thus not re-identifiable as such.”55 This is supposed to mark a difference in kind between universals and particulars. “The fundamental difference in kind between particulars and characters is that the former are bare, the latter are not. That is, particulars cannot be recognized (‘re-recognized’ would be better perhaps), characters can be. This is brought out that (at least some) characters are reidentifiable without criteria, things [particulars] are not.”56 Allaire speculates that the fact that particulars are not and characters are re-identifiable “explains why they [particulars] have been overlooked so often.”57 Let us leave the latter as it may be, and ask ourselves whether Allaire’s way of distinguishing characters, i.e., universals, from (bare) particulars is one that makes sense. Certainly, given that particulars and universals are equally bare, it cannot be a way of distinguishing bare entities from those that are somehow not bare. Yet this way of separating particulars and universals is not without its point. Only, it does not point to an intrinsic difference between the two kinds of entities.

356 The point is that to speak of things being “re-identifiable” is to make a comment more about our cognitive capacities than it is about the nature of the things cognized. To say that things are re-identifiable is to say that we can recognize not only difference but also sameness among characters. In contrast, to say that things are not re-identifiable is to say that we can recognize difference but not sameness. Now, we have agreed that for areas the rule is: one image - one area, or one concrete thing - one area. Particulars do not as it were repeat themselves in more than one thing. It follows that what is significant about them for our getting on in the world is that we recognize difference. But since there is no repetition, there is no need for us to recognize sameness. This is not to say that there is no sameness – that, surely, is there – but we have no occasion to notice it. Characters, in contrast, do repeat themselves – that is why they turn out to be universals. Being in more than one thing, they are locally separate from themselves, as Moore put it: “... with this sense of ‘locally separate’ [that is, that something can ‘be in two different places at the same time’] it seems to me perfectly obvious that a quality can be ‘locally separate’ from itself: one and the same quality can be in two different places at the same time.”58 Since qualities or characteristics of things can be in two different places at the same time while they are the same quality, if we are to get on in the world, if we are to find out about it and amongst the things in it, then we have need not only to recognize difference among characters but also on many an occasion to recognize sameness, recognize that this is the same characteristic here as over there. Thus, characters are indeed re-identifiable, particulars are not. But this is not an intrinsic difference, one that is built into the natures of the things. It is rather a reflection of, on the one hand, the fact that each ordinary concrete thing such as an image has within it one particular and that that particular is unique to it, and, on the other hand, the cognitive ends that we have as creatures trying to make our way about in the world. Conclusion: Bareness is often cited as an objection to a category of entities – particulars – whose ontological role is to individuate. This is what makes them such horrid little creatures. But in fact, it ought not to be shocking. Certainly, it ought not to be thought by an empiricist to be an objection to particulars. For, bareness turns out to be ubiquitous in the empiricist’s world: when the latter is clearly thought through it becomes evident that

357 the properties of things, which no one seems to find horrid, just as much as particulars are bare. So, just as the empiricist can admit universals as licensed by the Principle of Acquaintance, so he or she can also admit particulars as licensed by that Principle.

358 Endnotes to Study Eight

1.Cf. E. B. Allaire, “Bare Particulars,” in M. J. Loux, ed., Universals and Particulars; Readings in Ontology (Notre Dame, Ill.: University of Notre Dame Press, 1976), pp. 281-290. For discussion of particulars, objecting to them on account of their bareness, see H. Hochberg, The Positivist and the Ontologist: Bergmann, Carnap, and Logical Realism (Amsterdam, The Netherlands, and Atlanta, GA: Rudopi, 2001), Ch. 2. 2.Bertrand Russell, An Inquiry into Meaning and Truth (London: Allen and Unwin, 1948), p. 97. 3.William James, Principles of Psychology, 2 vols. (New York: Henry Holt, 1890), vol. 1, p. 243. 4.Ibid., vol. 1, 134. 5.Encyclopedia Britannica, 9th edition, article “Psychology,” p. 46, p. 53. 6.James, Principles of Psychology, vol. 2, pp. 135-136. 7.G. Bergmann, “Strawson’s Ontology,” in his Logic and Reality (Madison, Wisc.: University of Wisconsin Press, 1964), pp. 171-192, at p. 185. 8.James, Principles, vol. 1, p. 221. 9.G. E. Moore, “Are the Characteristics of Particular Things Universal or Particular?” in his Philosophical Papers (London: George Allen and Unwin, 1959), pp. 17-29, at p. 24. 10.Ibid., p. 31. 11.Ibid., p. 28. 12.Cf. G. Bergmann, “Synthetic A Priori,” in his Logic and Reality, pp.272-301, at p. 288. 13.See E B. Allaire, “Bare Particulars,” p. 288. 14.H. Hochberg, The Positivist and the Ontologist, p. 50, suggests that the identification of the areas in things with particulars is a “desperate” attempt to convince oneself that particulars are presented to one in ordinary experience. Perhaps. Hochberg suggests that the move is wrong-headed, but in fact he does not say why the identification ought not be made. Hochberg does note that Bergmann, having once made the identification (see fn. 12, above), later more or less dropped the point and relied upon dialec-

359 tics to make the case for bare particulars. But that is not to establish that the earlier identification is wrong. For myself, I, like James (whom Hochberg does not mention) and Bergmann (on occasion), find the identification persuasive. 15.James, Principles, vol. 1, p.221. 16.James, Psychology, vol. 1, p. 221. 17.For greater detail, see F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies (Toronto: University of Toronto Press, 1999), Study One. Also F. Wilson, The Logic and Methodology of Science and Pseudoscience (Toronto: Canadian Scholars Press, 2000), Ch. 3. 18.Note the contrast to our, more recent and scientific, notion of gravity; in the latter there is no notion of self movement. 19.D. M. Armstrong, What Is a Law of Nature? (London: Cambridge University Press, 1983). 20.See F. Wilson, Explanation, Causation and Deduction (Dordrecht, The Netherlands: Reidel, 1985), sec. 3.6, for an extended discussion of another, rather different defence of the idea of a primitive nomological connective. 21.Cf. F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies, Study One. Also F. Wilson, Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1997). 22.It is worth noting that when one ascribes, in the Aristotelian system, a nature or essence to a substance, one is not merely describing it but also making a normative claim about how it ought to be. On this scheme the ontological structure of the universe is also a normative structure. See F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies, Study One. Also F. Wilson, Socrates, Lucretius, Camus – Two Philosophical Traditions on Death (Lewiston NY and Queenston ON: Edwin Mellen Press, 2001), Ch. 3, and passim. 23.Compare P. Ferreira, Bradley and the Structure of Knowledge (Albany, NY: State University of New York Press, 1999). 24.F. H. Bradley, “Relations,” in his Collected Essays (Oxford: Oxford University Press, 1935), pp. 635-6. 25.Ibid., pp. 635-6. 26.Cf. F. H. Bradley, “Terminal Essays: On Judgment,” p. 634ff; in his Principles of Logic, Second revised edition, 2 vol., (London: Oxford University Press, 1922), in vol. 2.

360

27.Cf. F. H. Bradley, “Terminal Essays: The ‘This’,” ibid. 28.F. H. Bradley, Essays on Truth and Reality (London: Oxford University Press, 1914), p. 231. 29.Cf. J. Weinberg, “Relation,” in his Abstraction, Relation and Induction (Madison, Wisc.: University of Wisconsin Press, 1965); and also F. Wilson, “Burgersdijck, Coleridge, Bradley, Russell, Bergmann, Hochberg: Six Philosophers on the Ontology of Relations,” this volume, above. For some criticism of the latter, see H. Hochberg, The Positivist and the Ontologist: Bergmann, Carnap, and Logical Realism (Amsterdam, The Netherlands, and Atlanta, GA: Rudopi, 2001), p. 176ff. There is a reply to this criticism in the just mentioned study in this volume. 30.Cf. B. Russell, Principles of Mathematics, Second Edition (London: Allen and Unwin, 1938), Ch. XXVI. 31.James, Principles, vol. 1, p. 244. 32.F. H. Bradley, Appearance and Reality, Second Edition (Oxford: Oxford University Press, 1897), p. 125. 33.Ibid., p. 125. 34.James, Principles, vol. 1, p. 245. 35.John Sergeant, Method to Science (London: W. Redmayne, 1696). 36.John Locke, Essay concerning Human Understanding, ed. P. H. Nidditch (Oxford: Clarendon Press, 1979). 37.Cf. F. Wilson, “Acquaintance, Ontology and Knowledge,” and also “Moore’s Refutation of Idealism,” both in this volume, above.. 38.Cf. F. Wilson, “The Lockean Revolution in the Theory of Science,” in S. Tweyman and G. Moyal, eds., Early Modern Philosophy: Epistemology, Metaphysics and Politics (New York: Caravan Press), pp. 65-97 39.Cf. F. Wilson, “Burgersdijck, Coleridge, Bradley, Russell, Bergmann, Hochberg: Four Philosophers on the Ontology of Relations,” this volume, above. See also F. Wilson, “The Ultimate Unifying Principle of Coleridge’s Metaphysics of Relations and Our Knowledge of Them,” Ultimate Reality and Meaning, 21 (1999), pp. 243-61. 40.Russell, Principles of Mathematics, p. 448. 41.Cf. F. Wilson, “Bradley’s Impact on Empiricism,” in this volume, above. Also F.

361 Wilson, “Bradley’s Critique of Associationism,” Bradley Studies, 4 (1998), pp. 5-60. 42.James, Principles, p. 221. 43.M. J. Loux, “Kinds and the Dilemma of Individuation,” Review of Metaphysics, 27 (1973-4), pp. 773-784. 44.Loux, “Kinds,” p. 771. 45.M. J. Loux, “Particulars and their Individuation,” in Loux, ed., Universals and Particulars: Readings in Ontology, pp. 235-249, at p. 239. 46.On this point, which is in effect about the nature or ontological status of logic, see. F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies, Study Two. 47.Cf. F. Wilson, “The Rationalist Response to Aristotle in Descartes and Arnauld,” in The Great Arnauld and Some of His Philosophical Contemporaries, ed. E. Kremer (Toronto: University of Toronto Press), pp. 28-68. 48.G. Frege, “On Concept and Object,” in his Translations from the Philosophical Writings of Gotlob Frege, P. Geach and M. Black, trans., (Oxford: Blackwell, 1952), pp. 42-51, at p. 51. 49.G. Frege, “Function and Concept,” in Geach and Black, Philosophical Writings of Gotlob Frege, pp. 21-41, at p. 31. 50.Frege, “Function and Concept,” p. 28, p. 30, p. 32. 51.G. Bergmann, “Realistic Postscript,” in his Logic and Reality, pp. 302-340, at p.288. Compare F. Wilson, “Effability, Ontology and Method,” this volume, below. 52.G. Bergmann, “Ineffability, Ontology, and Method,” in his Logic and Reality, pp 45-63. 53.Bergmann, “Ineffability, Ontology and Method,” passim. Compare Wilson, “Effability, Ontology and Method.” 54.Cf. F. Wilson, “Effability, Ontology and Method.” 55.Allaire, “Bare Particulars,” p. 289. Compare Bergmann, “Strawson’s Ontology,” in his Logic and Reality, pp. 171-192, at p. 174. 56.E. B. Allaire, “Another Look at Bare Particulars,” in M. J. Loux, Universals and Particulars, pp. 297-303, at p. 301. Allaire is responding to V. C. Chappell, “Particulars Re-clothed,” in Loux, Universals and Particulars, pp. 290-295. Chappell is commenting on Allaire’s “Bare Particulars.” Chappell argues that Allaire’s case in “Bare

362 Particulars” does not establish on phenomenological grounds that there are bare particulars, but in the end makes the case on dialectical grounds. Allaire’s “Another Look” responds. 57.Allaire, “Bare Particulars,” p. 289. Bergmann makes the same point, “Strawson’s Ontology,” p. 174. 58.Moore, “Are the Characteristics of Particular Things Universal or Particular?” p. 25.

NINE UNIVERSALS, BARE PARTICULARS AND TROPES: THE ROLE OF A PRINCIPLE OF ACQUAINTANCE IN ONTOLOGY * Some philosophers suggest that there are universals. Others suggest that there are bare particulars. The former are appealed to in order to resolve the problem of sameness. The latter are appealed to in order to resolve the problem of difference. Others argue that both problems can be solved by taking properties to be individualized property instances or tropes. In the discussions of these matters there is often an appeal to the so-called Principle of Acquaintance. It is the purpose of this paper to investigate exactly what role this principle can play in these discussions, and exactly how it might be appealed to in order to rule out certain alternatives. This means that this study is primarily concerned with structural questions. It is concerned with the structure of arguments that may be used by one who accepts a Principle of Acquaintance as playing a fundamental role in ontological analysis. Many who have discussed the issues of particulars and universals have in fact accepted such a principle. Thus, we have Stout, Moore, Russell, Bergmann. Recently some metaphysicians, for example Sellars, have attacked the idea that a Principle of Acquaintance has any place in ontological discussions. Since the main point of this paper is structural, such attacks can be put to one side.1 That is, this paper is an attempt to show how, supposing there is such a principle, it may be deployed in ontological discussions. It may well be that the whole idea of attempting to settle ontological questions by means of such a principle is futile. Still, though, it must be said that this latter issue remains unsettled, and for this reason a discussion of the role of such a principle remains worthwhile. And there is, moreover, the further interest of *

This is a revised version of an essay which originally appeared as “The Role of a Principle of Acquaintance in Ontology,” The Modern Schoolman, 47 (l969), pp.37-56.

364 trying to understand structurally the positions taken by certain philosophers. Specifically, about the last, I shall urge that some who have claimed to hold to such a principle have not done so firmly. For example, it is generally assumed that a certain version of nominalism and a certain version of realism are incompatible. This paper tries to show that this is not so if a Principle of Acquaintance is made the fundamental principle of one’s ontology. Let us, then, assume that a Principle of Acquaintance is at least a defensible notion. First I shall make some relatively non-controversial terminological points about the entities with which one could be said to be presented, or acquainted. These points would be acceptable to almost all who would appeal to a Principle of Acquaintance. Next I shall try to state the ontological issues which arise with respect to these entities in a way that is as neutral as possible among the various alternatives. These issues are the problems of sameness and difference. Then the idea of a Principle of Acquaintance is introduced, and certain basic points about its role in settling ontological issues are made. The paper proceeds to a discussion of its role in the case of, first, the problem of sameness, and, second, the problem of difference, and, in connection with the second, its role with respect to (one version of) nominalism on the one hand and bare particulars on the other. A. Some Terminology In English ‘is’ is often used in a predicative sense. We are all able to pick out vast numbers of cases where it is, or might be, used in that sense. When it is so used, it occurs in sentences of the form ‘. . . is (a) --- .’ Here ‘. . .’ can be replaced by a demonstrative pronoun (‘this,’ ‘that,’ etc.), or by a name (‘Socrates,’ etc.), or by a definite description (‘the tallest man in Toronto,’ etc.). To obtain a sentence ‘ --- ’ would then be replaced by a suitable substantive noun (‘person,’ etc.), or adjective (‘red,’ etc.). When such replacement occurs we have a subject-predicate sentence. ‘. . .’ can also be replaced by an adjective (‘red,’ etc.); in this case ‘---’ would be replaced by a suitable noun (‘[a] colour,’ etc.), or adjective (‘drab,’ etc.). Again, when such replacement occurs we have a subject-predicate sentence. However, I do not wish to consider these latter sorts of subjectpredicate sentences. I wish to consider only those- which were indicated in the preceding paragraph; when I use 'subject-predicate sentence' hereafter, I shall be referring to sentences of those sorts alone. I now state what I mean by a thing. A thing is what one refers to by means of the subject term when one uses a subject-predicate sentence.2 For example,

365 consider two red discs. We may truly say of the one “This is red” and of the other “That is red.” ‘This’ and ‘that’ in these two subject-predicate sentences are used to refer to things. Thus, each of the red discs is a thing. Notice that I am not examining how (the word) ‘thing’, is used. Nor am I merely giving a linguistic sense to ‘thing.’ I am specifying the meaning of ‘thing,’ the meaning which I wish to give it, by giving examples of the sorts of entities to which I intend to refer when I use ‘thing.’ What I am interested in, and what one ought to be interested in, are things, not ‘thing.’ The point is that, the fact that I use a linguistic device to make clear what entities things are, does not entail that I am talking about ‘thing.’ What it is our aim to discuss is the problem of difference. It will therefore pay to restrict our attention to only a certain class of things, namely, phenomenal things, that is, entities with which we are acquainted in sense experience. If we stick with phenomenal things, then we may ignore the realismphenomenalism issue, whether such entities are, in some sense, minddependent. Further, it will also pay to restrict ourselves to a non-temporal world, that is, to ignore time and change. This will enable us to avoid the issue whether things are or are not (do or do not contain simple) continuants.3 A property (of a phenomenal thing) is the entity to which one refers by means of the predicate term when one uses a subject-predicate sentence. Thus, when one says truly of the disc “This is red,” ‘red’ is used to refer to a property, namely a property of the (phenomenal) thing to which one is referring with the ‘this.’ Moreover, the sentence “This is red” is true by virtue of the fact that the thing has the property it does and that ‘red’ correctly refers to this property. Some may disagree with the explication of ‘property.’ It might be objected that predicate terms, as well as subject terms, are used to refer to things; the difference between the two would be that the latter are used to make a proper reference while the former are used to make a common reference.4 However, there is about the things a reason or ground why this common name is used rather than that, why (e.g.) ‘red’ is used with respect to the two discs and not ‘green.’ So I reply to the objection that it is the entity that provides this reason or ground in the thing that I call the property. The objection is based only on my use of ‘refer.’ Properties are, of course, entities which are presented in our sense experience of the world. One is acquainted with them. Just as things are presented, so are the properties of things. By virtue of the way the thing is, one has a reason for saying that the disc is red rather than (e.g.) green. That ground or reason is presented to one. When one is acquainted with a red disc then one is

366 acquainted with that entity (viz., a ‘property’) which makes it true to say of that disc that it is red. This is not to say, however, that a property is another thing. The property is not another thing in addition to, or separable from, the thing of which it is a property; there is no entity red which exists apart from red things. Rather, properties are entities which are in things. Or at least, this is the meaning which I stipulate that ‘in’ shall have: the “being-in relation” is precisely that relation which obtains between the property red and the thing which is the red disc.5 I shall be referring to the converse relation when I say that things have properties.6 B. Two Philosophical Problems Consider a phenomenal field consisting of two red spots. Call the one Plato and the other Socrates. We now have the two true sentences: (1) Plato is red. (2) Socrates is red. (1) asserts that the property referred to by ‘red’ as used in this sentence is in the thing referred to by ‘Plato,’ and (2) asserts that the property referred to by ‘red’ as used in this sentence is in the thing referred to by ‘Socrates.’ The qualification ‘as used in this sentence’ which occurs twice is inserted in order to avoid begging any ontological issues concerning whether the property in Socrates is or is not in some sense identical with the property in Plato. The point about what (1) and (2) assert is a linguistic point, not ontological. So the only point that can as yet be made about the sameness or difference of the properties is the purely linguistic one that the same word correctly refers to the property in Socrates and to the property in Plato. (The notion of the ‘same word’ is, of course, complex. But that does not make it philosophically problematic. The notion is commonsensical, and appealing to it prejudges no ontological issues.) Suppose we were to give Plato another name, say Nikita. Then we would have the sentence (3) Nikita is red. (3) also is true. However, it asserts nothing different than what (1) asserts. For the linguistic conventions are such that ‘Nikita’ and ‘Plato’ refer to the same thing. (1) and (2) on the other hand make two different assertions. This is a linguistic point. A necessary condition for this is that ‘Plato’ and ‘Socrates’ do not refer to the same thing, i.e., do refer to different things. Two assertions are made by (1) and (2) only if the things referred to are different. The

367 necessary condition states, with respect to the entities (1) and (2) are about, what must be the case if two assertions are to be made. Thus, it states a condition relevant to ontology. In the case of (1) and (2) this necessary condition is fulfilled; in the case of (1) and (3) the corresponding necessary condition is not fulfilled. We can now state the Problem of Difference. What is the meaning of ‘different’ when one says, “Two assertions are made by (1) and (2) only if the things referred to are different”? Although (1) and (2) make two different assertions (a linguistic point), it remains true that, what they assert to be true of the two things to which their subject terms refer, is the same (another linguistic point). This latter linguistic point can be true only if there is something with respect to the entities which the sentences (1) and (2) are about such that the same predicate term ‘red’ correctly refers to a property in Plato and also correctly refers to a property in Socrates. This necessary condition, that there be something about the entities such that ‘red’ correctly refers to both the property in Plato and the property in Socrates, is a condition relevant to ontology; it is not simply linguistic. We can now state the Problem of Sameness. What is it about the entities such that the same predicate correctly refers to a property in one thing and also a property in another diverse thing? The two problems have been explained linguistically. This has been done in order that the statement of them be as neutral between various solutions as possible. Whether complete neutrality has been attained I cannot say. However, it must be emphasized that, although the problems have been approached linguistically, the problems themselves are not linguistic: they are problems about the world, about the facts that language is about, and they can be solved only by turning to that world, away from language. C. How to Answer Philosophical Problems Philosophical (ontological) problems have been answered in two ways: (a) transcendentally, and (b) explicatively.7 To offer a transcendental answer to an ontological problem is to argue that, unless certain entities exist, certain “obvious” truths will be false. The entities in terms of which the problem is solved are said to exist, not because one is acquainted with them, but because they must exist if the problem is to be solved. To offer an explicative answer to an ontological problem is to stipulate (either by direct interpretation of the terms into entities, or by defining them in terms of directly interpreted terms) that philosophically crucial terms have a certain commonsensical meaning,

368 and to argue that, with the crucial terms defined in the way stipulated, the problems stated using the terms are resolved. PA excludes attempts to answer ontological problems transcendentally. This is clear. It also excludes explicatory answers in which the stipulations do not anchor the meanings of the crucial terms in the entities with which one is acquainted. The first to adhere more or less consciously to PA as a principle in philosophy was Berkeley.8 He employed it as an effective weapon against Locke’s resolution of the problem of perceptual error. A brief examination will serve to illustrate how PA may be deployed.9 Locke’s account of perceptual error is this.10 One is acquainted with phenomenal objects, which he calls ideas. One is acquainted with nothing else. Ideas represent perceptual objects to us. The latter, one is never acquainted with; they are transcendent entities. (Exactly how Locke characterizes these perceptual objects is irrelevant for what we are about.) Ideas may represent these transcendent perceptual objects correctly or incorrectly. If the idea which we have correctly represents a perceptual object, then no perceptual error occurs; if the idea incorrectly represents a perceptual object, then error occurs. This clearly solves the problem of perceptual error, but only at the cost of introducing a transcendent entity as the perceptual object. Berkeley simply applies PA to rule out this solution: whatever the perceptual object might be, it is not a transcendent entity.11 Berkeley goes on to offer an explicative resolution of the problem of perceptual error. The crucial term is ‘perceptual object.’ He stipulates that ‘perceptual object’ is to mean ‘pattern of ideas (phenomenal objects).’12 Thus, the only (kinds of) entities which occur in perceptual objects defined in this way are entities with which one is (or could be)13 acquainted. The pattern is one that is learned by experience; it is a lawful pattern and laws are analyzed in the Humean way.14 This stipulation does not violate PA.15 Berkeley goes on to argue that, if one thus defines the crucial term >perceptual object,= the problem of perceptual error is solved. Perceptual error occurs if one has a false belief about the pattern, that is, if one has not correctly learned the laws definitory of the various kinds of perceptual objects, and in terms of which one predicts which ideas will be presented to one.16 Just as Berkeley appealed to PA to exclude a transcendent object being the perceptual object as “unintelligible,”17 so others have appealed to PA to exclude certain attempts to answer the problems of sameness and difference.18 Consider the Platonistic answer to the problem of sameness. According to it,

369 ‘red’ correctly refers to the property in Plato and the property in Socrates because, first, ‘red’ names a transcendent form, The Red Itself, and, second, the property in Plato participates in The Red Itself. Platonists argue that, if transcendent forms did not exist, the obvious truth, that ‘red’ correctly refers both to the property in PLato and to the property in Socrates, would be false. PA immediately excludes Platonic forms.19 They are transcendent entities. One who accepts PA therefore also believes the arguments of the Platonists to be unsound. In other words, he or she believes that an explicative answer can be given to the problem of sameness. This explicative answer must be supplemented by diagnostic remarks indicating just where the Platonist arguments are unsound, and indicating the confusions which make them plausible. (PA requires the arguments to be unsound, but does not show the point where the unsoundness lies.) D. PA and the Problem of Sameness Consider Plato and Socrates. They are two and not one. That is to say, they are numerically different. And, just as the things are presented to one, so is the fact that they are numerically different. It is this numerical difference to which I refer when I say of the two things that they are diverse. Thus, Plato and Socrates are different, i.e., diverse, while Plato and Nikita are the same, i.e., not diverse. Consider a visual field containing a .red round disc. There are two properties in this thing, namely, the one referred to by ‘red’ and the one referred to by ‘round.’ The one property can be distinguished in and of itself from the other. Now suppose that one has a visual field containing a red round disc and one containing a blue round disc. Call the former Pat and the latter Mike. The property of Pat referred to by ‘red’ is distinguishable in and of itself from the property of Pat referred to by ‘round.’ The property of Pat referred to ‘red’ is distinguishable in and of itself from the property of Mike referred to by ‘blue.’ But the property of Pat referred to by ‘round’ is, in and of itself, indistinguishable from the property of Mike referred to by ‘round’; Pat and Mike are indistinguishable with respect to shape. Finally, consider Tom, who, like Pat, is a red round disc. The property of Tom referred to by ‘red’ is indistinguishable in and of itself from the property of Pat referred to by ‘red’; Tom and Pat are indistinguishable with respect to colour. In short, in and of themselves certain properties are distinguishable; others are indistinguishable in and of themselves.20

370 If the properties in (one or more diverse) things are distinguishable in and of themselves we may say that they are different. Conversely, if a property in one thing is indistinguishable in and of itself from a property of another diverse thing, then we may say that they are the same. This sameness and difference of properties is presented to one. That is, when one is acquainted with two diverse things, one is presented with their having different properties (properties distinguishable in and of themselves), or with their having the same property (the property in the one is indistinguishable in and of itself from the property where the unsoundness lies). I have, therefore, provided an interpretation of the word ‘different’ as applied to properties, just as I previously provided an interpretation of the word ‘diverse’ applied to things. These interpretations are in terms of what one is acquainted with, in terms of what is presented to one. One should note that diversity is not the same as distinguishability. They are both senses of ‘different’. That means they have the same grammar (axiomatics). But diversity is a relation among things, while distinguishability is a relation among properties. Specifically, diversity is not a relation among properties. To see this, consider a disc which is both red and round. Here are two properties which are distinguishable in and of themselves. And yet they are not diverse. Suppose one said diversity was a relation among properties. Now, ‘diverse’ is a sense of ‘different’. Then there will be a sense of ‘same’ corresponding to this sense of ‘different’ such that if two entities are not different then they are the same: this is the grammar of ‘same’ and ‘different’. In the red round disc the two properties referred to by ‘red’ and ‘round’ are not diverse. So they are the same. But I can attach no sense to the claim that the properties red and round are the same. So I conclude that diversity is not a relation which holds among properties, but only among things. Of course, one can define a sense of ‘different’ in terms of ‘diverse’ which holds among properties qua in diverse things. But the present point is that diversity, unlike distinguishability, is not a relation which holds between properties in and of themselves, and with no reference to any entity they might happen to be in. If two diverse things have the same property, I shall say of the things that they are qualitatively the same in respect of that property. If two diverse things have different properties (distinguishable in and of themselves), I shall say of the things that they are qualitatively different in respect of those two different properties. Qualitative sameness and difference of diverse things are not primitive notions; they are defined in terms of the sameness and difference of properties. Nor, of course, is qualitative difference of things in any or all respects what is meant when one speaks of the diversity of things.

371 The meanings of the following words have been given: ‘property,’ ‘thing,’ and ‘(being) in (a thing).’ They are labels for kinds of entities with which one is acquainted. Furthermore, ‘sameness’ and ‘difference’ among properties are anchored to facts which are presented to one; such sameness and difference is explained in terms of distinguishability and indistinguishability. Finally, in terms of these notions one can define ‘qualitative sameness’ and ‘qualitative difference’ among things. The words occurring in the definiens of each of these defined notions are interpreted directly in terms of what is presented. That means all these stipulations are acceptable if one accepts PA. This solves the problem of sameness. This was stated as: What is it about the entities such that the same predicate correctly refers to a property in one thing and also a property in another diverse thing? The answer to this question is: The same predicate correctly refers to both because the two diverse things are qualitatively the same in respect of that property; or, to say the same thing differently, the same predicate correctly refers to both because the property in the one thing is the same as the property in the other thing. This constitutes the ontologically relevant necessary condition for correctly applying the same predicate to two diverse things. And the point to be emphasized is this, that this solution to the problem of sameness, unlike the Platonist’s, does not violate PA. One property is the same as another if the one is indistinguishable in and of itself from the other. It follows that the same property may be, and indeed often is in several different (diverse) things. If one maintains that this notion of sameness and difference among properties is primitive, then one might reasonably be characterized as a realist: the same property can be in several different things; properties would then be, in a sense acceptable to PA, universals.21 E. PA and Nominalism Consider once again Plato and Socrates. These two things are diverse. However, they have the same property; the property in Plato referred to by ‘red’ is the same as the property in Socrates referred to by ‘red’; the property in Plato to which ‘red’ refers is indistinguishable in and of itself from the property in Socrates to which ‘red’ refers. Some philosophers wish to introduce a third sense of ‘same’ and ‘different.’ This sense can be verbally specified as follows. Consider the property in Plato and the property in Socrates to which ‘red’ refers. Now take the sentences ‘This is red’ and ‘That is red,’ and suppose that ‘this’ is used to refer to

372 the property in Plato referred to by ‘red’ in (1), and that ‘that’ is used to refer to the property in Socrates referred to by ‘red’ in (2). The words ‘this’ and ‘that’ are to refer to the properties alone. This and that, therefore, are (a) in the two diverse things, and are (b) the same. Those philosophers who wish to introduce the third sense of ‘different’ grant both (a) and (b), but go on to assert that ‘this’ and ‘that’ refer to two different entities. We shall say the philosopher asserts the properties to be “different3.” Correspondingly, there is “sameness3.” Such a philosopher holds, we might say, that the instances of properties are particular. And such a philosopher might reasonably be called a nominalist. What are his “instance properties” are often called tropes; they have also been called perfect particulars. It is important to note that nominalism, in this sense, if it can in fact be maintained, is not incompatible with realism as previously defined. One can be both a realist and a nominalist. Only, there will be two meanings for ‘different’ as applied to properties. There will be distinguishable and different3. And certain properties will be the same in the sense of indistinguishable and yet different in the nominalist’s sense of ‘different3.’22 One could believe realism and nominalism to be incompatible only if one believed there was but one meaning for ‘different.’ However, the nominalist has not yet made his case. What he or she has done thus far is only verbal; he or she has only indicated the grammar of ‘different3.’23 But, if he or she accepts PA, then he or she must provide, in terms of what is presented, an interpretation for ‘different3.’ as he or she applies it to the so-called particular instances. He or she must do this in the same way as the realist provided an interpretation above for ‘diverse’ as applied to things, and for ‘different’ as applied to properties. Now, so far as I can see, there is no presented kind of difference for ‘different3’ to mark. That means nominalism is either false or meaningless. It follows that, though nominalism is verbally compatible with realism, for one who accepts PA the latter is, while the former is not, defensible. The nominalist may interpret ‘different3’ to mean what ‘diverse’ means. In that case, it is false that the property in Plato is different from the property in Socrates; for diversity is a relation which holds among things and not among properties. Or ‘different3’ may be interpreted to mean distinguishable in and of themselves. In this case it is again false that the property in Plato is different from the property in Socrates; for they are (by hypothesis) indistinguishable in and of themselves. Nor is there a third alternative, so far as I can see, for interpreting ‘different3.’ So if the nominalist holds that ‘different3’ means neither diversity nor distinguishability, then his or her view is meaningless.

373 He or she has not provided us with an interpretation for the crucial ‘different3’; all he or she has done is specified its grammar: it has a syntax but no semantics; it has been given an implicit definition but has no referent. Thus, consider Stout’s statement: He writes that: “A character characterizing a concrete thing or individual is as particular as the thing or individual which it characterizes. Of two billiard balls, each has its own particular roundness separate and distinct from that of the other, just as the billiard balls themselves are distinct and separate ... What then do we mean when we say, for instance, that roundness is a character common to all billiard balls. I answer that phrase ‘common character’ is elliptical. It really signifies a certain general kind or class of characters. To say that particular things share in the common character is to say that each of them has a character which is a particular instance of this kind or class of characters.”24 The nominalist might try to provide the required interpretation as follows. He or she will say, “This (referring to the colour and only the colour of the one thing) is one instance of (this shade of) red, and that (referring to only the colour of the other diverse thing) is another different3 instance of red.” But, as Moore correctly pointed out,25 when the nominalist, attempting thusly to interpret ‘different3,’ uses ‘this’ to refer only to the colour of the one thing, then he or she would not refer to anything different from what he or she refers to with ‘that.’ For, ‘this’ and ‘that’ are used to refer to the same (indistinguishable) colour, and so far as what is presented is concerned, the first conjunct of the nominalist’s attempt to interpret ‘different3’ asserts the same as the second conjunct. The nominalist insists that there are two different3 instances of the colour, but all that he or she has in fact succeeded in doing is repeating him- or herself! He or she certainly has not succeeded in giving a meaning in the sense of reference to the word ‘different3.’ Moore puts the point this way: “The refutation perhaps is that: When I say ‘This sensibly appears to me to be of that shade’ and ‘That sensibly appears to me to be of that shade,’ I am saying the same thing of this & that.” 26 Indeed, Moore puts it quite strongly: of nominalism, Moore exclaims, “Could anything be more absurd!”27 The error of the nominalist is like the error of the person who argues as follows. There is (this person claims) a sense of ‘different’, namely ‘different4’, in which the same thing can be different from itself; an interpretation for this ‘different4’ can be given in terms of what is presented, by noticing (he or she continues) the sentences (1) and (3), and saying (following the nominalist’s pattern) “Plato is one instance of this thing, and Nikita is another different, instance of this same (non-diverse) thing.” To be sure, this person has twice referred to Plato. But this is not the same as indicating something about

374 what is presented which will provide an interpretation for the proposed use of ‘different4.’ It has been objected that Moore is simply begging the question. J. R. Jones asserts that “... I cannot see that Dr. Moore is doing anything here beyond reiterating the orthodox theory...”28.But this objection to Moore’s criticism of nominalism does not seem to me to be fair. Moore pointed out that the nominalist merely specified the grammar of his terms, and that he or she has not given them any content (interpretation) compatible with PA. It is not simply a matter of “theory,’ whatever that may be; it is a matter of phenomenology, of what is presented. Jones says that Moore simply rejects Cook Wilson’s proposed criterion for the phrase ‘case or instance of.’ This “criterion” proposes that “... a universal can have only those particulars for its instances whose nature it completely ‘covers’ or exhausts.”29 With this as our criterion it would follow, according to Jones, “... that a thing’s characters [properties] must be particular.”30 And he goes on to assert that “I cannot see that he [Moore] proves that the notion of qualities as ingredient instances of quality-universals is a meaningless notion.”31 I think that Moore would reply that he indeed failed to prove meaninglessness. What he did was point out that either what the nominalist was saying was false on phenomenological grounds, or his words had been given no clear meaning in terms of entities presented to one; and while this may not be a proof that nominalism is meaningless, it is certainly as much as one could reasonably expect in way of refutation. Jones, of course, supposes that the nominalist’s use of ‘instance’ can be given a clear meaning in terms of Cook Wilson’s “criterion.” But Moore’s reply here would be that either this is merely to push the issue back one step or else it is to lapse into the transcendentalism of the Platonist. For, though the meaning of ‘instance’ is, in one sense, specified by the “criterion,” in another sense it is not specified. Its “meaning” is given in terms of ‘universal’ and ‘exhausting.’ But this is only to give some syntactical relations, an implicit definition, but no reference to what is in the world as it is presented to us. These two terms, ‘universal’ and ‘exhausting,’ must be interpreted in a way compatible with PA. If they are not, then ‘instance’ has not been interpreted in a way compatible with PA. Either Jones’s ‘universal’ is a Platonic Form (in which case he has violated PA already), or else all he has done is specify the grammar of ‘universal,’ ‘exhausting,’ and ‘instance,’ without providing an interpretation for them in terms of what is presented (in which case they are not meaningful in the sense required by PA).

375 F. PA and the Problem of Difference Consider two ontologies, call them UP and UO.32 Both these ontologies purport to be explicative. Both are realistic; they resolve the problem of sameness in the same way. Neither is nominalistic. Moreover, both UP and UO resolve the problem of difference in the same way. Both, being explicative, stipulate a meaning for ‘different’ which enables them, their proponents argue, to resolve the problem of difference. Both stipulate that ‘different’ is to mean ‘diverse.’ This is compatible with PA. Then “Plato is red” and “Socrates is red” make two different assertions only if this necessary condition is fulfilled; that the thing to which ‘Plato’ is used to refer is diverse from the thing to which ‘Socrates’ is used to refer. And since the thing to which ‘Plato’ refers is not diverse from the thing to which ‘Nikita’ refers, the sentences (1) and (3) do not make different assertions. This resolves the problem of difference. The crucial difference between the ontologies UP and UO is that the former maintains there are certain entities called “bare particulars,” while the latter denies there are any things of this kind. (Hence, UP = universal-particular, and UO = universal only). UP says the following about bare particulars. (b1) These entities are in things, in the same way that properties are in things. (b2) A bare particular which is in a thing is different from (distinguishable in and of itself from) the properties which are also in the thing. Moreover, (b3) there is in every thing exactly one bare particular. Finally, (b4) the bare particular in one thing is different from (distinguishable in and of itself from) the bare particulars which are in the other diverse things. It follows that UP can say the following about the problem of difference. (But it cannot be said by UO since the latter denies there are any entities of the kind UP calls “bare particulars.”) UP assumes acquaintance with bare particulars. And it agrees with UO that neither the category of “thing” nor the relation of “being in a thing” violates PA. UP stipulates that ‘different’ is to mean ‘distinguishable.’ Then difference is, in this sense, a relation which holds among the entities, properties and bare particulars, which are in things. Finally, UP defines the word ‘diverse.’ According to UP we can define ‘A is diverse from B’ as short for ‘there is a bare particular in A and there is a bare particular in B and the former is distinguishable from the latter.’ And we can define ‘A is the same as B’ as short for ‘there is a bare particular in A and there is a bare particular in B and the former is not distinguishable from the latter.’

376 Does this definition successfully analyze “diversity”? Assume bare particulars satisfy PA, and assume the claims (b1) and (b2). If these assumptions are false, then, for that reason alone, the analysis cannot be a successful explicative analysis. On the other hand, even if these claims are true, the analysis may still be unacceptable, for it may still fail to account for all cases of diversity. Here is where one begins to argue. The defender of UP argues that the definition of ‘diverse’ which he or she proposes successfully analyses the notion of diversity in terms of which the problem of difference has already been resolved. The defender of UP considers the following proposition, that if one has two diverse things, in the pre-analytical sense of ‘diverse,’ then one also has two distinguishable bare particulars, one of which is in one of the things, and the other of which is in the other diverse thing. He or she notes that, if this proposition is true, then his or her definition of ‘diverse,’ in terms of distinguishable bare particulars to be found in the things said to be diverse, is a satisfactory definition. He or she then notes that the proposition is in fact true; it is true by virtue of the claims (b3) and (b4) which UP makes concerning bare particulars. If, therefore, the claims (b1) - (b4) can be sustained, the definition UP proposes for “diversity” yields a successful analysis.33 UP maintains that ‘diverse’ is the meaning of ‘different’ appropriate for the resolution of the problem of difference. But it also maintains that this is not a primitive notion of difference; that there is only one primitive notion of difference; and that this notion is that of distinguishability. It can maintain this by virtue of its claims about bare particulars. In an obvious sense, therefore, it is bare particulars which provide UP’s answer to the problem of difference. This points out the basic disagreements between the two ontologies UP and UO. (D1) UO maintains that there are two fundamental meanings of ‘different,’ viz., diverse and distinguishable. (D2) UP maintains that there is only one fundamental meaning of ‘different,’ viz., distinguishable; ‘diverse,’ it says, is definable because there is in each thing exactly one bare particular, entities of a kind which do not occur in UO. At the same time, UP and UO also agree, as we saw. They agree on their answers to the problems of sameness (they are both realistic) and difference (they both appeal to diversity). Moreover, neither is nominalistic.34 My main point has been to concentrate on senses of ‘different’; how do these tie in with PA? Answering this question illuminated the realismnominalism issue. It does not in the same way illuminate the disagreement between UO and UP. There, clearly, the crucial issue is whether bare particulars are compatible with PA. Equally clearly, this moves off on a different path than the one we have just been pursuing. Let me conclude the present

377 line of thought, therefore, simply by briefly illustrating how those who defend bare particulars on the basis of PA have not always been as clear about the issue we have been discussing as they might reasonably have been expected to be.35 A single articulate example will suffice to illustrate this. E. B. Allaire’s essay, “Bare Particulars,”36 provides such an example. 37 Allaire suggests that realism is incompatible with nominalism, in the sense in which I am using these terms. As I have already pointed out, this makes sense only if one believes there to be but one sense of ‘different,’ and that given dialectically. Consider two discs of the same (shade of) color, size, shape and so on. The objection to [the nominalist’s analysis] is that it cannot account for the sameness of the discs since the members of the two collections are all unanalyzable and different from each other.38

This is so, of course, only if the nominalist is not permitted to introduce two primitive relations of difference between his particular instances; particular instances can be the same in the sense of indistinguishable, but none the less different. Allaire tends to think of only one sense of ‘different.’ This in turn renders what he says about bare particulars and diversity obscure. Defending bare particulars, he, of course, believes that there is but one primitive sense of ‘different,’ viz., distinguishable. Furthermore, since he accepts PA, he realizes his words must be hooked to what is presented to him. Some things, our two red discs for example, are the same shade of color... They are the same in that they are distinguishable as such, or, more precisely, one cannot differentiate them by their color alone.39

Diversity, too, is presented: [A] Consider once more the two discs. When presented together, they are presented as numerically different. That difference is presented as is their sameness with respect to shape, (shade of) color, and so on.40

Allaire is here clearly speaking about things, in the sense in which we use that word. The phrase “difference which is presented” can only be diversity.41 Yet Allaire continues immediately with:

378 [B] What accounts for that difference is the numerically different [bare particulars]. No [property], or group of [properties] can do that. Thus, to say that there are [bare particulars] is to say that things may be merely numerically different.42

In quotation [A] it is things which are “numerically different.” So this is ‘different’ in the sense of diverse. And the final sentence of [B] reinforces this, for there it is clear that he is talking about things being diverse while being qualitatively the same. And it is true to say, as the second sentence of [B] does, that ‘diverse’ cannot be defined in terms of properties, if one holds that the only sense of ‘different’ appropriate to properties is distinguishable. But now re-read the first sentence of [B]. Here Allaire is talking about “numerical differences” among bare particulars. So ‘difference’ has now shifted its meaning from ‘diverse’ which refers to a relation among things only and not entities in things, to ‘distinguishable’ which refers to a relation among bare particulars, entities which are in things.43 Allaire blurs ‘distinguishable’ into ‘diverse’, as if there were only one meaning of ‘different.’ Of course, because he is defending UP he holds that only ‘distinguishable’ is primitive, and that ‘diverse’ can be defined in terms of it. Indeed, this I take to be the force of the word ‘account’ in the first sentence of [B].44 But the point is that the fact that on UP ‘diverse’ can be defined does not mean that this is not a sense of ‘different’ which must be distinguished from ‘distinguishable.’ It is precisely this distinction which Allaire blurs. Nor is the matter helped at all by the suggestion that subject-terms of ordinary subject-predicate sentences refer to bare particulars. Consider ‘this is red’ asserted truly of a colored disc. Some philosophers claim that the sentence refers to a fact consisting of an individual (bare particular) and a [property] (universal) referred to by ‘this’ and ‘red,’ respectively.45

Assertions are made, primarily, about things.46 Sentences involving ‘Plato’ and ‘Socrates’ make different assertions by virtue of (we are agreed) the fact that the thing referred to by ‘Plato’ is diverse from the thing referred to by ‘Socrates.’ It is diversity which solves the problem of difference. Because distinguishability among bare particulars can, on UP, be used to define diversity, it does not follow that diversity does not solve the problem of difference. All that follows is that, when using ‘Plato’ one refers to a thing, then one indirectly also refers to the bare particular which is in that thing. By suggesting that ‘Plato’ refers directly to the bare particular rather than the thing, Allaire obscures the fact that it is diversity which is crucial for the problem of difference.47

379 This, together with the suggestion that there can be but one sense of ‘different,’ covers up the fact UP and UO agree on the solution to the problem of difference. Both agree that diversity solves the problem. The disagreement between the two ontologies is, rather, with respect to (D1), the claim of UO, in contrast to that of UP, that there are two primitive meanings of ‘different’; and with respect to (D2), the claim of UP, in contrast to that of UO, that there are entities of the kind called bare particulars which are presented and which are in things. G. Tropes and Blobs G. E. Moore wonders whether there can be anything more absurd than the tropish nominalism of Stout. In fact, there can be something more absurd. Howard Robinson has remarked that “materialist theories are incompatible with realist theories of universals. The tie between nominalism and materialism is an ancient one.”48 Ancient or not, one of the positions thus tied, namely materialism, is, as Broad once remarked, just silly; or, as Bergmann once said, materialism is philosophy of mind for non-philosophers: in reply to the materialist one must simply say that there are so mental states, from pains and pleasures, to perceivings, to believings, supposings, lovings, hatings, and so on, we experience these states and none of them are physiological states.49 But, one presumes, the point would be the same if one substituted ‘naturalism’ for ‘materialism’, where naturalism could be taken in either of two senses, neither of which denies the existence of minds or mental states. In the first sense of ‘naturalism’ the world B nature B consists of entities that are localized in space and time. In the second sense of ‘naturalism’ the world consists of entities that are in or are parts of the world that we know by sense experience.50 Now, naturalism in the first sense is indeed popular. But there seems little reason to accept the principle that defines it, what can be called the “Principle of Localization,” which is the axiom that No entity can exist at different spatial locations at once and the same time nor at interrupted temporal intervals.51 For, it is evident that things have properties are in and of themselves indistinguishable and in that sense there are properties in diverse things that are not different, i.e., are the same. In this sense, properties are not localized: they are in fact in this sense universals. Of course, the nominalist argues that there is a sense of ‘different,’ namely, ‘different3’ in which properties, though indistinguishable, are differ-

380 ent: properties exist not as universals but as property instances or tropes. It is true, as Moore said, that it is difficult and indeed so far as one experiences these entities impossible to see in what sense of ‘different’ the supposed property instances are different, that is, what the sense is that is given to ‘different3’. That being so, it is easily seen that the real problem is the axiom, the Principle of Localization. Or rather, Dogma, since seldom are reasons given for accepting this axiom. But, once questioned, one can find no reason for not rejecting it. In fact, the Principle of Acquaintance gives one such a reason. PA admits properties into one=s ontology, and these can be the same, in the sense of indistinguishable, and yet at different places in the sense of being in itself in diverse things. Nominalists generally represent different3 property instances or tropes by complex names. We have Socrates is red1 and Plato is red2 where we can say that red1 is (a) Red and that red2 is (a) Red The Platonist would hold that the term ‘Red’ in these two sentences refers to the transcendent form “the Red Itself”. The more moderate nominalist, like Stout, would reject such forms, often on grounds of something like an appeal to PA or to common sense,52 and simply insist that the redi’s are, while all different ( = different3), are all the same in the sense of indistinguishable. The more radical nominalist argues that the sameness among redi’s is not a matter of objective sameness but rather something that is as it were induced on these quality instances by language: the redi’s are all the same, namely, red, simply by virtue of being called by the same name, namely, ‘red’. Sellars is such a one.53 Thus, at one point he tells us that ...for a predicate to stand for an attribute or relation is for it to be of a certain kind. Thus, to stand for Triangularity is to be a *triangular*. What is it to be a *triangular*? It is to be an item which does the job in the base language by ‘triangular’s. Specifically, it is to give a singular term concatenated with a counterpart character, T’. It is T’ individual constants which correctly picture triangular objects, provided that the individual constants are correlated ... with the objects.54

He has earlier explained that

381 The fundamental job of singular first-level matter-of-factual statements is to picture, and hence the fundamental job of referring expressions is to be correlated as simple linguistic objects with single non-linguistic objects.55

Names name objects in the sense of properties like triangularity or, one presumes, redness. Thus, we have the name red1 naming the property, that is, the property instance, red1, and we form the sentence red1 is (a) Red or the sentence triangular1 is (a) Triangular or T= is (a) Triangular by the appropriate concatenations of signs. The meaning, however, of the predicate ‘Red’ or the predicate ‘Triangular’ in these sentences is not given by the aspects as it were of the world to which they apparently refer: meaning is not reference. Rather, for Sellars, the (cognitive) meaning of a term is given by its intralinguistic connections to other terms: ...although T’ individual constants are correlated with T objects, the concept of this correlation is not the analysis of what it is for T’ individual constants to stand for triangularity, nor does it explain what it is for T’ individual constants to denote triangular objects. The correlation between objects and their linguistic pictures must not be confused with the pseudo-relations standing for and denoting. Thus, that ‘triangular’s stand for triangularity essentially involves the intralinguistic consequence uniformities governed by the consequence rules (axiomatics) of geometrical predicates.56

So the redi’s are all Red simply because they are called red; that is, the geometrical sign ‘red’ which is correlated to the redi’s is what makes them all Reds, and it makes them Red’s not through those world-word correlations but rather through the intralinguistic connections that sign ‘red’ has to other terms in the language. The objects red1 and red2 are not both red because in and of themselves they are the same in the sense of indistinguishable, that is, objectively the same, where this objective sameness is reflected in language by the occurrence and recurrence of the same sign, but rather they are both red because through the contingencies and accidents of language they happen to evoke occurrences and recurrences of a certain sign which has certain intralinguistic connections to other signs in the language, where this sign is the one which has the geometrical form ‘red’. If

382 a is red1 and b is red2 then, on this view, to say that a and b are both Red is simply to say no more than is said by predicating ‘x is Red’ = Df ‘(x = a v x = b)’ of a and b which in fact is a view that is hard to take seriously. It proposes that the qualitative sameness of things is not objective, that is, not objectively there in the things as we experience them. The properties of things are, in other words, are nothing but blobs, made into things indistinguishable qualitatively in various respects because that is the way we have come to use language: it is the blob theory of properties. If Moore finds that nothing can be more silly than the nominalism of Stout, then surely he is wrong, this blob theory of properties of Wilfrid Sellars is even more silly. If it is true, as Sellars once said, that “a naturalist ontology must be a nominalist ontology,”57 and the nominalism is that of his blob theory, then one must conclude that no one can be a naturalist. Johanna Seibt has made the point, approvingly, that, “since the basic relation of a Platonist theory of predication cannot be defined in naturalist terms, a nominalist theory of predication proves to be ‘...the very foundation of a naturalist ontology’.”58 If the rejection of Platonism means accepting the blob theory, then surely something has gone wrong. One thing that has gone wrong is the acceptance of the Principle of Localization. This is the real root of nominalism and since there is no reason to accept this Principle, we need not accept it, nor the nominalism that it implies. In fact, one can now argue that nominalism, the acceptance that ‘different3’ as applied to properties is a meaningful concept of ‘difference,’ that is, the acceptance of the claim that one can in experience identify properties in things as different3, is itself an example of a transcendental ontology ― an ontology that introduces entities that violate the Principle of Acquaintance.59 The other thing that has gone wrong is the idea that realism in the sense of accepting that properties are universals means the acceptance of Platonism, that is, the acceptance of the idea that one has to hold that these universals must, like Plato’s Forms, be entities that transcend the natural world. But if by the “natural world” one means the world as we experience it, the world as it is given to us in our ordinary sensible experience of things, then it is perfectly clear that one can have universals which are part of, and do not transcend, the

383 world as we ordinarily experience it in our inner awareness and in our sensible awareness of things. For, given the commonsense point that properties insofar as, on the one hand, they are in different things but, on the other hand, those occurrences are in and of themselves indistinguishable, that is, insofar as the properties in diverse things are the same, then properties are at once universals and also given in sensible experience.60 One can in this way be a realist and a naturalist at the same time: naturalism requires neither nominalism nor Platonism. H. Particulars Presented What of the difference between the UO and UP ontologies? What does PA have to say about these? Well, the dispute here is not about properties as universals ― the two agree that that is what properties are. The dispute is about the particular, the so-called “bare particular.” Assuming one adopts PA, then the dispute ought to be about whether the particulars UP says are there and UO says are not there are in fact given to us in our sensible experience of the world. This, however, is often not the locus of dispute. Thus, it has been said of the UP view that the particulars about which it talks are “posits”61 or “postulated entities”62, much as one would say of the entities like atoms and quarks about which physics speaks but which one cannot see that they are “theoretical entities” the existence of which is something to be postulated. Suppose one takes things to be bundles of properties, takes these properties to be universals, and takes the bundles to be diverse. All this is compatible with PA. And that means that there is nothing that is philosophically problematic about the situation, puzzling in that peculiar way in philosophical problems are puzzling. Specifically, there is no “problem of individuation” which can be solved only by the introduction of another set of entities, the particulars, which are then said to be bare in order to contrast them with universals. Thus, Bergmann tells us that particulars are there in his ontology, that they are not properties and that they “solve” the “problem of individuation”: Bare particulars neither are nor have natures. Any two of them are not intrinsically but only numerically different. That is their bareness. It is impossible for a bare particular to be “in” more than one ordinary thing .... A bare particular is a mere individuator .... It does nothing else.63

384 But why is there a problem, that is, a philosophical problem, with UO that only UP can solve. As Hochberg has put it, ... one can reasonably hold that different objects are simply different objects B there is nothing that is needed to, or indeed can, non-trivially, account for such difference. Nor need we seek to “analyze” such objects as being bundles of compresent qualities, substrata exemplifying qualities, or anything else.64

Of course, things are bundles of qualities or properties, but there is a good sense in which that judgment is not arrived at through a process of “analysis”, not in some sense of “philosophical analysis” − though it is true that one has the thing and then can attend to the various different (distinguishable) properties which are in that thing, and which are therefore predicated of it: how else could one come to predicate those properties of the thing? It is also true, however, − and Hochberg does not note this − that the UO ontology, while compatible with PA and therefore not a matter of philosophical puzzlement, does rely upon two different senses of ‘different,’ namely, ‘different’ in the sense of ‘distinguishable’ as applied to properties and ‘different’ in the sense of ‘diverse’ as applied to things. It is this that some ontologists have felt uncomfortable with. And the reason for this discomfort is not hard to find. The reason is clearly that they accept as an axiom the Principle of the Identity of Indiscernibles: Entities having exactly the same constituents are identical.

Given this Principle, then two different things cannot have all their constituents in common.65 But things which are different often do have all their properties in common. There must be something about the things, given the Principle of the Identity of Indiscernibles, that individuates them, some entities that are constituents in them and make them different. This is the socalled Problem of Individuation. Notice that it is a problem only if one accepts the Principle of the Identity of Indiscernibles. Be that as it may for the moment, it is evident that many philosophers introduce tropes precisely in order to solve this Problem of Individuation. This is true, for example, of Stout. But tropes won’t do: they require as sense of ‘different’, what we called ‘different3’, that is incompatible with PA ― it is this fact that ‘different3’ cannot be given any meaning in terms that fit properties as we experience them that led Moore to reject the view as “silly”. Hence, if one accepts the Principle of the Identity of Indiscernibles, there must be

385 present in the two different things two entities which are different. These entities are not properties; they are the particulars. Since they are there to make the two different things different, they are not properties or qualities, and are therefore said to be in and of themselves, quite bare. These bare particulars do nothing but individuate. It is clear how the argument for “bare particulars” goes. It is equally clear why many are led to say that those who accept the argument are said by others to “posit” bare particulars. One is not proceeding from an acquaintance with them: they are not, like properties or universals, legitimated by PA, but rather legitimated by an argument that proceeds on grounds other than PA, namely, on the Principle of the Identity of Indiscernibles as a premise. Thus, Hochberg contrasts the “problem of universals” which he argues can be solved as we have solved it, in effect by an appeal to PA, and the “problem of individuation.” There is, he suggests, a “clear difference” between the two problems. In the case of universals, one is not introducing a postulated entity. Rather, given that objects have properties, the question concerns the analysis of categorization of what properties are B universals, tropes, classes, etc. In the case of bare particulars or substrata, one does not ask what the ordinary object is as one asks what a property is ..., but what individuates such objects.66

The problem of universals ― which asks in virtue of what are different things the same? ― is thought to be parallel to a problem of difference B in virtue of what are objects that are the same different? It is the Principle of the Identity of Indiscernibles that makes these two problems seem parallel and seem to require analogous answers, one in terms of universals and one in terms of bare particulars. The point is that the problem of universals is solved in a way compatible with PA, where the problem of difference is solved by entities introduced by an argument which, if successful, does not suppose or presuppose that the entities satisfy PA. Moreover, this inferred entity is strange indeed. As has been said, “the required individuator must be an inferred residue devoid of any descriptive content. That is, the underlying unrepeatable must be opposed to and other than any possible characterization − Lcoke’s we-know-not-what.”67 These particulars are not there as properties are there, characterizing the object; they are “devoid” of such content. But that doesn’t matter: the argument establishes that they really are there. And it establishes this even though there is no need to suppose that we must be aware of them or that they are given to us in experience. They are “substrata” and “we-know-not-what’s.” It is this that does them in for many,

386

It is their radical emptiness that proves fatal to bare particulars.

68

If this were all there was to it, then it would indeed be the end of the story: bare particulars would have no place in one’s ontology, or at least, not if one accepted PA. For one who accepts PA, then the appropriate response is the rejection of bare particulars, because they are not legitimated by PA, and, moreover, the rejection of the Principle of Indiscernibles, which leads to their being posited but which has in fact no reason to recommend it, or at least, no reason sanctioned by PA.69 We are back, then, to UO and UP, with it apparently the case that, since the positing of bare particulars is illegitimate, UP is without support: it seems that UO alone is compatible with PA. The problem concerning bare particulars that we have been discussing turns on the apparent conflict with the Principle of Acquaintance. The arguments we have been considering are all at it were a priori, based on the Principle of the Identity of Indiscernibles taken as an axiom, and affirmed independently of PA. The existence of bare particulars has been made into a matter of dialectics alone, not a matter of being justified by PA. Some have seen this conflict between bare particulars and PA. Here is Allaire. The defense is in the first instance dialectical, based on the Principle of the Identity of Indiscernibles. He considers two diverse discs with all their properties in common. When presented together, [the two discs] are presented a numerically different. That difference is presented as is their sameness with respect to space, (shade of) color, and so on. What counts for that difference are numerically different individuals. No character or group of characters can do that. Thus, to say that they are individuals is to say that things may be merely numerically different. No matter what description one proposes, the numerical difference of two things which are alike in all (non-relational) respects must be accounted for.70

Notice the starting point: the two things are presented as different, that is, different in the sense of being diverse. Here the difference in the sense of diversity is given to one in acquaintance. But then we are asked, what “accounts for” this difference? We are now taken to be assuming that such diversity is somehow philosophically problematic. This problem that has now appeared requires solution: an entity must be found, or rather must be there, whether found or not, which does the accounting. This entity which does that is an “individual” which is taken to be a constituent of the things, there in addition to the (non-relational) properties which are present in the things. Allaire is clearly assuming something like the Principle of the Identity of

387 clearly assuming something like the Principle of the Identity of Indiscernibles to justify the inference to the individuals which account for the difference. And, now, given the presence of the individuals as constituents, the relation of diversity holding between the things can be defined in terms of the difference among the constituents. But notice that one is defining ‘diverse’ in terms of ‘different’ where the latter is not given or not wholly given in a way compatible with PA. To be sure, the various properties in the things are different in the sense of distinguishable, and that is, as we have insisted, a sense of ‘different’ justified by PA. But we are to take it that the individuals are also different from the other constituents of the things. We cannot do that, however, for those individuals are introduced not because they conform to PA but for dialectical reasons that have nothing to do with PA. How then can one say that those individuals are different in the sense of ‘distinguishable’? Allaire is taking for granted that ‘different’ is univocal, and that it acquires what meaning it has independently of PA. For one who works within an ontology shaped by PA, that is unacceptable. But Allaire also recognizes this. He goes on: To claim that both [the two tings] are collections of literally the same universals does not account for the thisness and thatness which are implicitly referred to in speaking of them as two collections. That is, the two collections of characters ― if one persists in speaking this way ― are, as presented, numerically different. Clearly, therefore, something other than a character must also be presented. That something is what proponents of the realistic analysis call a bare particular.71

Allaire is clearly arguing that we really are acquainted with bare particulars: we must be acquainted with them because they must be there. But none of this follows.72 As Hochberg has made clear, neither of these ‘musts’ need be accepted. First, that they must be presented because they must be there, does not follow.73 Second, because the two things are diverse, it does not follow that that diversity must be “accounted for” by a constituent: it could be that the diversity is primitive and unanalyzable, that the two things are simply diverse.74 For one who accepts PA, to include these entities called “bare particulars” into one’s ontology, it is necessary first to establish that one is in fact acquainted with them in our sensible experience of the world. This can be done, it has been suggested, interestingly enough by Bergmann75 but before him by William James. James was, as usual, perceptive on these things. He carefully distinguished the properties of things and the relations among them. He suggested that we

388 think of the world of which we are conscious on the patterns of the world of a bird’s life. “Like a bird’s life, it [the world as experienced] seems to be made of an alternation of flights and perching,” where the resting-places are “usually occupied by sensorial imaginations of some sort.”76 The flights are relations, the perchings are properties ― relations among and properties of sensible things. As for the resting-places, James observes that “In the sensations of hearing, touch, sight, and pain, we are accustomed to distinguish from among the other elements the element of voluminousness.”77 James notes that “this element [which he refers to as extensity] [is] discernible in each and every sensation”; and comments that “extensity, being an entirely peculiar kind of feeling indescribable except in terms of itself, and inseparable in actual experience from some sensational quality which it must accompany, can itself receive no other name than that of sensational element.”78 Extensity is, it is clear, a distinguishable part of the things we experience, and in that sense of ‘different’ is different from the properties of a thing. Each ordinary thing has, as an element within it, its extensity. It is there, upon the extensities, that perchings perch; and it is among these elements that flights take off and come to rest. Let us refer to the extensity of a thing like a red image as its “area.” The quality of redness as a property of the thing is a perching upon the area in the thing. And if one red image is to the left of another, then the relation of being to the left of is a flight that takes off from the area of the one thing and comes to rest on the area of the other. Now, one does not argue for the existence of areas: they are simply given to one. If there is disagreement, if someone says that he or she does not see them, then all one can do is what James once recommended one do is such circumstances. If you are confronted with one who says he or she does not know these entities, one who claims he or she is not acquainted with them, then, one does not offer an argument but rather, James suggests, what one does, and all that one can do is “...say to my friends, Go to certain places and act in certain ways and these objects will probably come.”79 We can say with James, it is clear, that every sensible thing comes as a piece as it were of extension; there is an area which defines each thing. Thus, we have one image - one area, or one concrete thing - one area. Now, ordinary things, images for example, are individual things; as Allaire put it, we have this image and that. An ordinary thing, a this or a that, is something complex. This complex has properties and these properties are with each other. An ordinary thing is thus a group of properties that are with one another. But it is not just a group of properties that are with one another: there is also the area that is in the thing. An ordinary concrete thing is thus a group of

389 properties together with an area; and these entities are with one another forming the thing. Furthermore, an ordinary thing is not just a thing: as a group of entities that are with each other, the ordinary thing is a fact. An ordinary thing is a particular, but it is a particular fact. The qualities in the fact do not make it a particular fact, it is not by virtue of the qualities in that fact that it is distinguished from other facts. For, after all, the qualities in the fact, the properties of the thing, are universals. That which distinguishes the fact from other such facts is the area in the fact. It is the area which is the entity which makes an ordinary thing a particular. In that sense, the area itself may be called the particular which is in the particular fact which is the ordinary thing. Since things are wholes of which areas and properties are parts, and the properties are universal, the only entity that is unique in each concrete thing is the area. As Bergmann has noted, every sensible thing comes as a piece as it were of extension; there is an area which defines each thing. Thus, we have one image - one area, or one concrete thing - one area. Now, ordinary concrete things, images for example, are individual things, we have this image and that. A concrete thing, a this or a that, is something complex. It has properties and these properties are with each other. An ordinary thing is thus a group of properties that are with one another. But it is not just a group of properties that are with one another: there is also the area that is in the thing. An ordinary concrete thing is thus a group of properties together with an area; and these entities are with one another forming the thing.. Furthermore, an ordinary thing is not just a thing: as a group of entities that are with each other, the ordinary thing is a fact. An ordinary thing is a particular, but it is a particular fact. The qualities in the fact do not make it a particular fact, it is not by virtue of the qualities in that fact that it is distinguished from other facts. For, after all, the qualities in the fact, the properties of the thing, are universals. That which distinguishes the fact from other such facts is the area in the fact. It is the area which is the entity which makes an ordinary thing a particular. In that sense, the area itself may be called the particular which is in the particular fact which is the ordinary thing. Since things are wholes of which areas and properties are parts, and the properties are universal, the only entity that is unique in each concrete thing is the area. Areas are particulars, and as such they individuate concrete things.80 When Bergmann or any philosopher like him who accepts PA says clearly that he or she is acquainted with particulars, what he or she is acquainted with are areas, where, to repeat, these areas with which he or she is acquainted are particulars, and as such they individuate concrete things.81

390 Now note that areas are distinguishable from the properties of things; they are in this sense different from the properties. So every thing has an area as a distinguishable part in it, and for each thing its area is unique to it: it is individuated by its area. There is no “problem of individuation” which must be solved by positing the existence of particulars. Nonetheless, there are particulars, that is, areas, and these particulars do individuate. And since they do individuate, they can be used to define ‘diversity’: the difference of things, their diversity, can thus be analyzed in terms of the difference among its constituents and the constituents of other things, that is, difference in the sense of ‘distinguishable.’ In other words, on this ontology ― this UP ontology B there is only one basic sense of ‘different’, and all the entities in this ontology, all the entities that it takes to be the constituents of things, are there in conformity to the Principle of Acquaintance. A UO ontology is acceptable, it passes the test of PA. But the UO ontology omits certain entities which are there, namely, the areas in things, the entities that turn out to be individuators, and which permit UP to have but one basic sense of ‘different’ where UO requires two. But this point about the number of senses of ‘different’ is not what is crucial. What is crucial is that UO does not admit, where UP does admit, certain entities, namely, areas, the extensity which is there in each thing, where the introduction of these entities is justified by the Principle of Acquaintance. The ontologist who accepts PA has no need to fear particulars, they are not I-know-not-what’s.

391

Endnotes to Study Nine

1 But for a discussion of these issues and a defense of the Principle, see F. Wilson, “Acquaintance, Ontology and Knowledge,” in the present volume, above. 2 And, in the case of a definite description, when the description is successful. 3 This introduction of the continuant issue into the discussion mars John Trentman’s paper, “Recognition, Naming and Bare Particulars,” Dialogue, 5 (1966), pp.19-30. See also the discussion of this paper in K. Barber, “Bare Particulars and Acquaintance,” ibid., 58083. 4 Some might raise the objection in a slightly different way, preferring to say that the subject term is used to refer (make a reference) and that the predicate term is used to ascribe (make an ascription). But the point I am about to make is sufficient (with appropriate translations) to say such reformulation of the objection make no difference.. 5 Here I follow H. H. Price. See his Thinking and Experience (London: Hutchinson, 1953), p. 12. 6 Since I have specified the meaning of ‘in’ and ‘has’ by giving examples, it is absolutely no objection to ask, “Are properties >in= things as books are in libraries?” or, “Do things ‘have’ properties as people have raincoats?” If such questions are asked, then the person who asks has failed to understand. J. W. Roxbee Cox fails to understand. In his essay (“Are Perceptible Qualities ‘in’ Things?”, Analysis, 23 [1963], pp. 97-103) he makes the utterly irrelevant point that the use to which I am now putting ‘in’ is not an everyday use − it is not an everyday use, but it is for all that a u se that can be characterized as commonsensical − the interpretation into what we are acquainted with ensures that. 7 This way of phrasing the distinction is Edwin B. Allaire’s. See his, “Ontology and Acquaintance: A Reply to Clatterbaugh,” Philosophy o/ Science, 32 (1965), pp. 277-80. Allaire is replying to Kenneth Clatterbaugh’s “General Ontology and the Principle of Acquaintance,” Philosophy of Science, 32 (1965), pp. 272-76, where this important distinction is not made. 8. The roots of the use of PA go back beyond Berkeley, of course. The one who made the crucial moves was Descartes. Nor do I want to suggest that Berkeley was consistent in his use of PA. He was not. See E. B. Allaire, “The Attack on Substance: Descartes to Hume,” Dialogue, 3 (1965), pp. 284-87. 9. My point is illustrative, so I am not going to pretend historical accuracy. Compare Phillip Cummins, “Perceptual Relativity and Ideas in the Mind,” Philosophy and Phenorne-

392 nological Research, 2 (1963), pp. 202-14; and “Berkeley’s Likeness Principle,” Journal of the History of Philosophy, 4 (1966), pp. 63-69. 10. Cf. John Locke, Essay concerning Human Understanding, ed. P. H. Nidditch (Oxford: Oxford University Press, 1979), Bk. II, Chap. viii, sec. 19-20. 11. Cf. Berkeley, Principles of Human Knowledge, in vol. 2 of his Works, ed. A. A. Luce and T. E. Jessop, 9 vols. (London: Nelson, 1948-57), §§ 3, 4; also § 8. Berkeley moves from something’s existing if and only if it is perceivable (i. e., from PA) which alone suffices to rule out Locke’s transcendent perceptual objects, to something’s existing if and only if it is perceived. This latter (wrong) move is not essential to Berkeley’s use of PA. For a discussion of why Berkeley moves from perceivable to perceived, see Edwin B. Allaire, “Berkeley’s Idealism,” Theoria, 29 (1963), pp. 229-44, and F. Wilson, “On the Hausmans’ ‘New Approach to Berkeley’s Ideal Reality’,” in the present volume, above. 12. Berkeley, Principles, §§ 29, 30, 33; also § 37 and § 38. 13. See footnote 10. 14. Berkeley, Principles, § 31, §65. 15. Ibid., § 49. 16. Ibid., §§ 33, 34, 59. 17. Ibid., § 3. 18. Addis appeals to PA to dismiss the notion that we are acquainted with Ideas. Someone might argue that though we are acquainted with perceptual objects, the dialectics require us to introduce into our analysis certain entities called Ideas. PA dismisses such arguments. See L. Addis, “Ryle’s Ontology of Mind,” in L. Addis and D. Lewis, Moore and Ryle: Two Ontologists (The Hague: Nijhoff, 1965), p. 63. 19. Cf. John Austin, “Are there A Priori Concepts?” in his Philosophical Papers (Oxford: Oxford University Press, 1961), pp. 1-22. 20. Ignore the psychological connotations of ‘distinguishable.’ One can distinguish the entities only because there is something about the entities themselves which makes them distinguishable. 21. Platonic Forms are often called “universals” also, but that is a quite different sense of the term. 22. I take it to be the position of G. F. Stout that there are these two basic meanings of ‘same’ and ‘different’ as applied to properties. See the passage cited by note 24. 23. That is, this philosopher has only given, as it is sometimes put, an “implicit definition”

393 for the term, a set of axioms (syntax) but no interpretation into entities with which we are acquainted (semantics): remaining uninterpreted, it remains meaningless. For discussion of the notion of “implicit definition,” see F. Wilson, “Implicit Definition Once Again,” in the present volume, below. 24. G. F. Stout, “The Nature of Universals and Propositions,” Proceedings of the British Academy, 10 (1921-23), pp. 158-59. 25. G. E. Moore, Lectures on Philosophy, ed. C. Lewy (London: Allen and Unwin, 1966), p. 56. 26. Moore, ibid., p. 56 27. Ibid., p. 56. 28. J. R. Jones, “Simple Particulars,” Philosophical Studies, 1 (1950), p. 67. 29. Ibid., p. 65. 30. Ibid., p. 66. 31. Ibid., p. 67. 32. H. Hochberg discusses these in his “Universals, Particulars, and Predication,” The Review of Metaphysics, 19 (1965-66), pp. 87-102. 33. Further moves will have to be made with respect to spatial relations. Are these exemplified by things, or by the bare particulars in things? 34. The nominalist can also define ‘diverse.’ But unlike the defender of UP, the nominalist requires two meanings of ‘different.’ 35. The discussion of these issues in the earlier version of the present study, F. Wilson, “The Role of a Principle of Acquaintance in Ontology,” The Modern Schoolman, 47 (l969), pp. 37-56, seems to have been overlooked by H. Hochberg, in his “Individuation and Individual Properties: A Study of Metaphysical Futility,” Modern Schoolman, 79 (2002), pp. 107-134. 36. E. B. Allaire, “Bare Particulars.” Philosophical Studies, 14 (1963), pp. 1-8, and also in E. B. Allaire et al. Essays in Ontology (The Hague: Nijhoff, 1963), pp. 14-21; page references are to the former. But, as I shall indicate, Allaire has come to recognize some of the infelicitous formulations of this paper. 37. Another excellent example can be found (by the author=s own admission!) in G. E. Moore’s Some Main Problems of Philosophy, (London: Allen and Unwin, 1953). In the 1952 Appendix to these 1910-11 lectures Moore points out that he had earlier failed to distinguish between a thing, a colour patch, and its colour, a property in the things: “I

394 failed, for some reason or other, to see that the colour of a patch is not identical with the patch in question; and hence I drew the absurd conclusion that, if two patches are of exactly the same colour, then they are not two patches but one and the same patch!” (p. 374) But a complete discussion of even only those points of this immensely rewarding book which are relevant to what we are presently about would take us far afield. That makes Allaire’s succinct discussion more appropriate to our purposes than Moore’s. 38. Allaire, “Bare Particulars,” p. 2; his italics. 39. Ibid., p. 4. 40. Note also the blur in the two quotations between sameness as a relation between properties, and sameness, i.e., qualitative sameness, as a relation between things. 41. Ibid., pp. 6-7. 42. Ibid., p. 7. Allaire uses ‘individual’ synonymously with ‘bare particular,’ and ‘character’ synonymously with our ‘property.’ 43. Compare Allaire’s later statement that “Bare particulars are ... the entities in things accounting for the numerical difference of things.” (“Another Look at Bare Particulars,” Philosophical Studies, 16 [1965], p. 18; his italics). 44. Compare the quotation cited by footnote 34. 45. Allaire, “Bare Particulars,” p. 1. 46. Only those assertions made with subject-predicate sentences are relevant. Compare H. W. B. Joseph: “When I judged, I distinguished in the subject I judged about a character which I predicated of it.” (Introduction to Logic [Oxford: Oxford University Press, 1916], p. 14). 47. Allaire subsequently clarifies this: “The object denoted by ‘this’ is, at the minimum, its (non-relational) properties. That is, ‘this’ denotes a thing. The issue is whether or not a thing also contains a particular, an entity which is different in kind and wholly distinct from properties.” (“Ontology and Acquaintance: A Reply to Clatterbaugh,” p. 279; his italics). 48. H. Robinson, Matter and Sense (Cambridge: Cambridge University Press, 1982), p. 30. 49. We have been over this in the Appendices to “Acquaintance, Ontology, and Knowledge,” in this volume, above. 50. ‘Parts of’ allows for entities that are too small to be seen or too far away or too large: such entities are parts of the world of sense experience so long as they can be hooked conceptually as it were by scientific inference to entities that are given to us in the world

395 which we know in our sensible experience. For an empiricist account of our knowledge of entities too small to perceive, see F. Wilson, “Empiricism and the Epistemology of Instruments,” The Monist, 78 (1995), pp. 207-229; the same reasoning will do for objects too large or objects too far away to see. 51. Cf. R. Grossmann, “Sensory Intuition and the Dogma of Localization,” in E. B. Allaire et al., Essays in Ontology, pp. 50-63.

53. See, for example, W. Sellars, “Naming and Saying,” in his Science, Perception and Reality (London: Routledge and Kegan Paul, 1963), pp. 225-246, and his Science and Metaphysics (London: Routledge and Kegan Paul, 1968). For criticism, see H. Hochberg, “Mapping, Meaning, and Metaphysics,” in his Logic, Ontology and Language (Munich: Philosophia Verlag, 1984), pp. 157-184; to which Sellars replied in his “Hochberg on Mapping, Meaning and Metaphysics,” Midwest Studies in Philosophy, 2 (1977), pp. 219220; to which Hochberg replied in his “Sellars and Goodman on Predicates, Properties and Truth,” also in his Logic, Ontology and Truth, pp.185-195. For some of the general issues, see R. Grossmann, “Common Names,” in E. B. Allaire et al., Essays in Ontology, pp. 64-77. 54. Sellars, Science and Metaphysics, p. 127. 55. Sellars, Science and Metaphysics, p. 124. 56. Sellars, Science and Metaphysics, p. 128. We have seen this previously as Sellars’ account of (cognitive) meaning, first adumbrated in his “Inference and Meaning,” Mind, n.s. 62 (1953), pp. 313-338, in our discussions in “Acquaintance, Ontology and Knowledge,” in the present volume, above. See also F. Wilson, “Placing Bergmann,” forthcoming. Sellars’ account of meaning as given by inference or consequence connections, that is, as he also puts it, by the axiomatics of the terms involved, shows that he accepts the doctrine that meaning can be conferred by “implicit definition.” But this notion of “implicit definition” has little or less to be said for it; see F. Wilson, “Implicit Definition Once Again,” in the present volume, below. Sellars’ concept of (cognitive) meaning is simply unacceptable. Others have offered the same general account of meaning, for example, M. Dummett (see F. Wilson, “Dummett on the Origins of Analytic Philosophy,” in the present volume, below). The criticisms we are making of Sellars apply to these others also. 57. W. Sellars, Naturalism and Ontology (Atascadero, CA: Ridgeview Press, 1979), p. 109. 58. Johanna Seibt, Processes and Properties: A Synoptic Study of Wilfrid Sellars’ Nominalism (Astacadero, CA: Ridgeview Press, 1990), p.184. 59. And, as was argued in F. Wilson, “Acquaintance, Ontology and Knowledge,” in the

396 present volume, above, such a transcendental lapse leads to a radical and irreparable scepticism.. 60. In this sense it is wrong to say that it is part of the realist view that “...the exemplification relation or nexus ... connects entities from different ‘worlds’,” even if one places the term ‘worlds’ in special quotes. See J. P. Moreland, Universals (Chesham, Bucks: Acumen Publishing, 2001), p.122. It is also wrong to suggest that universals must be construed as “transcendental” and as “abstract objects” which cannot enter into relations of “causal influence” in the way particulars, or only particulars, can do. (See L. Bonjour, In Defense of Pure Reason [Cambridge: Cambridge University Press, 1998], pp. 153-186; and also Moreland, Universals, p. 123.) The fact that this water is being heated causes it to be the case that this water is boiling. It is the one fact or event that causes the other fact or event, and ingredient in those two facts or events are two properties: as Hume saw so clearly, causal relations relate events only insofar as they are of certain kinds B how else could causal connections involve regularities? In the absence properties there would be no causal relations. (For a discussion of the notion that its is events which enter into causal relations, and that events are not facts but simple particulars, see F. Wilson, Explanation, Causation and Deduction [Dordrecht, The Netherlands,: D, Reidel, 1985], p. 327ff. The basic point is the simple one that an event is a fact of comparatively short temporal duration.) 61. D. W. Metz, “Individuation and Instance Ontology,” Australasian Journal of Philosophy, 79 (2001), pp. 45-61, at p. 45. 62. Thus, H. Hochberg, “Individuation and Individual Properties: A Study of Metaphysical Futility,” p. 132. 63. G. Bergmann, Realism (Madison, WI: University of Wisconsin Press, 1967), pp. 24-25. 64. Hochberg, “Individuation and Individual Properties,” p. 132. 65. One assumes here the argument of Russell, that relations do not individuate; see his “On the Relations of Universals and Particulars,” in his Logic and Knowledge, ed. R. C. Marsh (London: Allen and Unwin, 1956), pp. 103-124. See also Hochberg, “Individuation and Individual Properties,” pp. 133-134. 66. Hochberg, “Individuation and Individual Properties,” p. 132. 67. Metz, “Individuation and Instance Ontology,” p. 48. 68. Ibid. 69. Hochberg, “Individuation and Individual Properties,” pp. 132-134, makes this point clearly. See also his “Things and Qualities.” 70. Allaire, “Bare Particulars,” p. 4.

397 71. Ibid. 72. Some argue that none of this matters: who nowadays worries about acquaintance. Thus, J. P. Moreland, in his “Theories of Individuation: A Reconsideration of Bare Particulars” (Pacific Philosophical Quarterly, 79 [1998], pp. 251-263), argues that it does not matter that Allaire fails to establish that we are acquainted with the bare particulars that he establishes on dialectical grounds must be there: “Today, most philosophers would not place the type of constraints on analytic ontology that was present in Bergmann’s day. The real issue for bare particulars is whether or not the arguments for and against them are sufficient to justify their adoption as a solution to individuation, not whether they are sense perceptible.” (p. 256) But for some of us, those who are still empiricists, and do ontology in a framework established by PA, acquaintance does still matter: dialectics is not, and cannot, be everything. Acquaintance is still important even if Allaire’s argument is, as Moreland suggests, unsuccessful. However, for some of the dialectics, see Moreland, “Theories of Individuation,” and Universals; J. Hoffmann and G. Rosenkranz, Substance and Other Categories (Cambridge: Cambridge University Press, 1994); D. W. Metz, “Individuation and Instance Ontology”; M. Loux, Metaphysics: A Contemporary Introduction (New York: Routledge, 1998); Nathan Oaklander and A. Rothstein, “Loux on Particulars: Bare and Concrete,” The Modern Schoolman, 78 ( 2000), pp. 97-102; R. B. Davis, “‘Partially Clad’ Bare Particulars Exposed,” Australasian Journal of Philosophy, 81 (2003), pp. 534-348, and “The Brave New Bare Particularism,” The Modern Schoolman, 81 (2004), pp. 266-273. For criticism of the idea, fairly clear in Moreland, that dialectics are prior to acquaintance, see the Appendix to F. Wilson, “Implicit Definition Once Again,” below. 73. L. Addis, however, has suggested otherwise, or at least has suggested that it does not matter: bare particulars must exist, and therefore whether we are acquainted with them is an irrelevant issue. But it is not, and could not be, an irrelevant issue for one doing ontology within the framework of a Principle of Acquaintance. See L. Addis, “Particulars and Acquaintance,” Philosophy of Science, 34 (1967). 74. See Hochberg, “Individuation and Individual Properties,” pp. 132-134, and “Things and Qualities.” 75. Thus, Bergmann, has claimed that he accepts the Principle of Acquaintance, and claimed to be acquainted with his bare particulars. But he put his claim in an odd way, writing that “I, of course, have convinced myself that I am actually presented with two things [two particulars in two images]. Yet I am loath to rest the case on this conviction, for I am convinced that a very major part of it is dialectical.” (“Strawson’s Ontology,” in his Logic and Reality [Madison, Wisc.: University of Wisconsin Press, 1964], pp. 171-192, at p. 185). 76. William James, Principles of Psychology, 2 volumes (New York: Henry Holt, 1890), vol. 1, p. 243. 77. Ibid., vol. 1, 134.

398 78. James, Principles of Psychology, vol. 2, pp. 135-136. 79. James, Principles, vol. 1, p. 221. 80. Cf. G. Bergmann, “Synthetic A Priori,” in his Logic and Reality, pp.272-301, at p. 288. 81. But there is an important sense in which such particulars are not quite bare; see F. Wilson, “Effability, Ontology and Method,” in the present volume, below. But there is also a sense in which bareness does not matter: if particulars are bare, then so are properties; see F. Wilson, “Bareness, as in ‘Bare Particular’: Its Ubiquity,” in the present volume, above.

Ten The World and Reality in the Tractatus * Tractatus 2.04 reads: “The totality of existent atomic states of affairs (Sachverhalten) is the world (Welt).”1 In Proposition 2.06, one finds: “The existence and non-existence of atomic states of affairs is the reality (Wirklichkeit). This Proposition continues parenthetically: “The existence of atomic states of affairs we also call a positive, their non-existence a negative fact (Tatsache).” The world is, therefore, the totality of positive facts, while the total reality would have to be the totality of positive and the totality of negative facts. However, reading on, one finds in Proposition 2.063 that “The total reality is the world.” Now Wittgenstein is saying that the world is the totality of positive facts and the totality of negative facts. On the face of it, at least, there is an inconsistency. This has bothered some commentators.2 However, the bother is needless, the inconsistency only apparent. The key is Proposition 2.05: “The totality of existent states of affairs also determines which atomic states of affairs do not exist.” In the light of what has just been said, the Proposition can be restated as: the world determines the total reality. An explication of ‘determine’ can be given which both removes the appearance of inconsistency and makes reasonable sense of the Proposition. Or, more accurately, makes reasonable sense of it if the Tractatus has a realistic ontology. It will help – as it almost always helps with the Tractatus – to restate matters in terms of language.3 Consider the sentences, or, more precisely, the atomic sentences, of the Tractatus’ ideal (ontologically perspicuous) language. Consider only those sentences with monadic predicates (the world we are considering is ontologically very simple). Let (L) be the list of all the true atomic sentences. That means that (L) represents the totality of positive facts (existent atomic states of affairs). Let (L*) be the list of all false atomic sentences. Then the totality of negative facts (non-existent atomic states of affairs) is represented by (L*).4 From 2.04 we see that the * Originally appeared as “The World and Reality in the Tractatus,” The Southern Journal of Philosophy, 5 (1967), pp. 253-260.

400 world is represented by (L). 2.06 shows us that the total reality is represented by the combined lists (L) and (L*). It follows that 2.063 must be read as asserting that there is no significant difference between the list (L) alone and the combined lists (L) and (L*). If Proposition 2.05 is restated linguistically, then it asserts that the (L) “determines” the list (L*). What is this notion of “determining”? A reasonable reading would give to 2.05 at least this much content: that given the list (L) one can write down the list (L*). This minimal reading of 2.05 by itself already removes the apparent inconsistency: the list (L) is not significantly different from the combined lists (L) and (L*) because if we are given (L) then we also and thereby have (L*). This is a minimal reading of 2.05. It alone suffices to remove the apparent inconsistency. Yet more must be said. One should be prepared to show why anyone would want to affirm 2.05. Further, the reading is minimal. The clearly present logical force of the notion, that one list determines the other, must be accounted for. Consider a realistic ontology, that is, one containing two different kinds of objects, viz. particulars and characters, where the characters are universals in the sense that the same character can be exemplified by more than one particular. Suppose that in this ontology, the objects combine in one and only one way: particulars exemplify characters to form facts. Let the list (P) ‘x1’, ‘x2’, ... contain names of all and only the particulars. Let the list (C) ‘f1’, ‘f2’,... contain names of all and only the characters. Let a string of the form ‘fx’ represent the fact that the property or character f is exemplified by the particular x. One can now construct a list of strings of the form ‘fx’, each member of which represents a fact. Call this list (L) [deliberately]. One can also construct a second list (call it (L*)) of strings of the form ‘fx’ each member of which represents a character-particular combination which does not obtain. One who held such a realistic ontology might well wish to maintain the truth of the Principle of Exemplification (PE) which states that, every character is exemplified by at least one particular and every particular exemplifies at least one character. One good reason such an ontologist might have for affirming PE is to deny Platonism, i.e. to deny the view that there are unexemplified characters, or, in other words, to deny the view that characters are not dependent on particulars. Another good reason is to deny the existence of unqualified particulars, particulars existing independently of any character.5 Now, if PE be true, then, given (L) alone, one could construct (L*).

401 Conversely, a realist who maintained that, given (L) alone, one could construct (L*), would thereby be affirming PE. This much is clear. For, one could obtain (L*) from (L) only on the condition that each mark of (P) and each mark of (C) occur in at least one string in the list (L). If this condition were not fulfilled, then in order to obtain (L*) it would not suffice to know (L) alone: one would need in addition a list of those marks of (P) and (C) which were not mentioned in (L). In short, one can obtain (L*) from (L) alone only on condition that PE is true. Suppose that the Tractatus is realistic, that its objects are particulars and characters which can be exemplified by more than one particular and which, therefore, are universals.6 These objects combine into facts.7 These facts are represented by the true atomic sentences. Proposition 2.05 asserts that the list (L*) can be obtained from the list (L) alone. This is the minimal reading of 2.05. This means that if the Tractatus is realistic, then Proposition 2.05 simply states the Principle of Exemplification. If the Tractatus is realistic, then Proposition 2.05 affirms a principle which can be backed up by good philosophical reasons. In this light of this, let us now examine the whole sequence from Proposition 2.04 to 2.063. Proposition 2.04, that the world is the totality of existent atomic states of affairs, comments on Proposition 2 (“What is the case, is the existence of atomic states of affairs”). It asserts what follows from 2 and from Proposition 1 (“The world is everything that is the case”). Given this, Wittgenstein asserts in 2.05 that the totality of existent atomic states of affairs determines which atomic states of affairs do not exist. In asserting this, he asserts PE. Or, what amounts to the same, he asserts that one can infer the non-existent atomic states of affairs from the totality of existent states of affairs.8 (Compare Propositions 2.223-2.225: “In order to tell whether a picture is true or false we must compare it to reality.” “It is impossible to tell from the picture alone whether it is true or false”; and “There is no picture which is a priori true”.) In 2.06 Wittgenstein explains what he means by ‘reality’: reality is the existence or non-existence of atomic states of affairs. Propositions 2.061-2.063 comment on reality as so defined. 2.061 asserts that “Atomic states of affairs are independent of one another.” This use of ‘independent’ he explicates in the next comment, Proposition 2.062. ‘Independent’ is explicated in terms of logical inference; that ties this comment in with what was said in 2.05. Proposition 2.062 says “From the existence or non-existence of an atomic state of affairs we cannot infer the existence or non-existence of another.” In short, one part of the total reality as defined by 2.06 is (logically) independent of

402 every other part – except, Wittgenstein goes on to point out in 2.063, the part of the total reality that is represented by (L*) is not (logically) independent of that part of the total reality that is represented by (L): in 2.05 it was seen that the former can be inferred from the latter.9 The same point can also be expressed as follows. Each atomic state of affairs is logically independent of every other. That means that it is logically possible that each one exist or not exist independently of the others. But, though it is logically possible that each not exist, it is not logically possible for all not to exist. For, if all failed to exist, then no particular would exemplify any character nor would any character be exemplified by any particular. That means that it is logically necessary that some atomic states of affairs exist, a sufficient number that each particular exemplifies at least one character and that every character is exemplified at least once. It is not hard to find support for this interpretation. Firstly. (A) Proposition 2.06 defines ‘reality’. Proposition 2.05 asserts PE. 2.05 and the comments relating to 2.06 have all been previously stated in the Tractatus. The (logical) independence of the various pieces as it were of reality is asserted in 1.2, 1.21. The former tells us that “The world (Welt) divides into facts (Tatsachen).” The latter states that “Any one [i.e., fact] can either be the case or not be the case, and everything else remain the same.”10 That is, I take it, (atomic) facts are (logically) independent of one another. The comments 2.061, 2.062 merely restate these matters in terms of “atomic states of affairs,” a notion introduced in 2 and 2.01.11 (B) PE, or more accurately, what is asserted by Proposition 2.05, has already been stated in Proposition 1.12: “... the totality of facts determines both what is the case, and also all that is not the case.” Secondly. PE is held by the author of the Tractatus, not only to be true, but to be in some sense necessarily true. In Proposition 2.0122 one finds: “The thing is independent, in so far as it can occurs in all possible circumstances, but this form of independence is a form of connexion with atomic states of affairs, a form of dependence.”12 Wittgenstein is here explicating two of the philosophical uses of ‘independent’. The explication is in terms of ‘possible’. A particular (character) is independent in the sense that it is logically possible that it be connected with any character (particular) to form an atomic fact: each particular and each character is what it is and is logically and ontologically self-contained and there is, therefore, nothing about any one of the particulars (characters) entering into combination with any of the characters (particulars).13 But particulars and characters are as kinds of objects mutually dependent in the sense that it is logically impossible to have either a particular or a character which is not

403 cally impossible to have either a particular or a character which is not connected with an object of the opposite kind in such a way that it occurs in at least one atomic fact; that is, they are mutually dependent in the sense that an object of one of these kinds cannot be unless it be connected with at least one object of the other kind.14 In other words, the assertion of 2.0122, that “this form of independence ... is a form of dependence,” is simply the assertion of PE.15 In a remark parenthetical to 2.0122, Wittgenstein restates PE linguistically: “It is impossible for words to occur in two different ways, alone and in the proposition.”16 If it was not clear from the main part of 2.0122, this remark now makes it so, that Wittgenstein maintains, not merely that PE is true, but that it is necessarily true.17 Why does Wittgenstein treat PE as a necessary truth? that is, as a logical truth? To begin with, like any logical truth it does, according to Wittgenstein, say something about the facts, about the entities represented by language.18 It is, in this broad sense of ‘factual’, a factual truth. In this sense of ‘factual’, then, there is no distinction between factual truths and those which are necessary: everything is factual. And indeed there is a point to this: as Bergmann used to say, there is nothing logical about logic. However, there are still distinctions one can make. Certainly, philosophers have been inclined – strongly inclined, one should probably say – to distinguish the necessary and the contingent. It is the philosopher’s task to explicate these often unclear and generally problematic notions. The philosopher does this by drawing distinctions among the factual truths, in the broad sense of ‘fact’ in which even logical truths are among the factual truths about the world. Drawing these distinctions, one can explicate what other philosophers might reasonably have meant in calling some of these truths “necessary” (“logically necessary”) and others “contingent” (“empirical,” “factual”). How, then, would PE be reflected in language? That is, how would it be reflected in an ideal language, one which perspicuously represents one’s ontology? Consider the syntactical formation rule for atomic sentences of this language: all well-formed strings (atomic sentences) of the ideal language are of the form ‘fx’, and no other string is well-formed. Given the list of names of objects and this formation rule, all well-formed strings can be written down. This list represents the total reality. Among the sentences of this list, there are those which are true. These represent the world. But exactly which well-formed strings are true, cannot be determined by the formation rule alone; rather, it can be determined only by looking as it

404 were at the world. At the same time, given the list of all true sentences, the formation rule show that we can, and how we can, write down all false sentences. For, it says no string other than those of the form ‘fx’ are wellformed; in particular, it says no string of the form ‘f’ and no string of the form ‘x’ is a well-formed string in the ideal (ontologically perspicuous) language.19 Thus, given the total reality and the formation rule, we cannot get the world, but given the world and the formation rule, we can get the total reality. This shows two things. In the first place, it shows that the reflection in the ideal (ontologically perspicuous) language of PE is the formation rule. In the second place, it shows us how to explicate some of the philosophical uses of ‘necessary,’ and related terms such as ‘possible.’ Assumes the names of all objects are given. Together with these names, the formation rule determines all possible well-formed strings. This use of ‘possible’ is purely combinatorial, and hence commonsensical and unproblematic, needing no explication. Not needing explication, it can be used to explicate.20 And so, the statement that so and so is “possible” (philosophical use) may reasonably be explicated as saying in a confused way what can reasonably be said as, so and so can be represented by a possible (combinatorial and commonsensical use) string. Similarly, so and so is “impossible” may reasonably be explicated as, so and so is not represented by a possible string, i.e., it is represented (so to speak) by an ill-formed string.21 Finally, if one explicates “possible world” as that which is represented by a list of well-formed strings in which the name of each object occurs at least once, then the formation rule determines all possible worlds – again explicating a problematic notion, that of a “possible world,” using the non-problematic notion of combinatorial possibility. This then explicates the sense of ‘necessary’ in which the formation rule is “necessary”; for, the traditional formula goes, the necessary is what holds in all possible worlds: the same formation holds in each list that defines a possible world.22 To repeat what was said before, the formation rule, while it is in the sense just indicated “necessary,” is for all that a factual truth. If the world were different (e.g. if it contained characters which existed independently of any particular) another formation rule would be needed. The claim that it is not a factual truth is philosophical; it involves a philosophical use of ‘factual’. This use of ‘factual’ can be explicated thus: given the total reality and the formation rule we cannot get the world, but given the world and

405 the formation rule, we can get the total reality; in this sense, what’s in the world depends on the facts. Thus, on the now-explicated narrower use of ‘fact’, all facts are possible, none are necessary. A reasonable sense has been provided in which PE is necessary, i.e., in which the formation rule represents a necessary truth. But this does not show why Wittgenstein should assimilate the formation rule to logical truths. Having proceeded thus far in its explication of the modalities, the Tractatus now blurs issues. Complications have to be introduced. Wittgenstein introduces them, but, unfortunately, not too carefully. So they lead to a blurring of the distinction that must be drawn between a formation rule and a sentence which is necessary in the sense of being a tautology – a second sense of ‘necessary’ which Wittgenstein has not yet introduced. What is involved can easily be located. Note that in the just stated explication that “so and so is ‘impossible’” is explicated with “so and so is represented by an ill-formed string”. Here “impossible” and “ill-formed” coincide. Similarly, “necessary” statements are all “metalinguistic” in the sense that they are statements in the syntax language (since all statements in the object language represent possibilities, not necessities). However, this is true only so long as the ideal language contains only atomic sentences. Once connectives are introduced, and molecular sentences as well as atomic sentences are well-formed, then the patterns break down. For, a new distinction has to be introduced. This is the analytic-synthetic distinction. Some molecular well-formed strings (though no atomic string) can now reasonably be called necessary, either necessary truths (tautologies) or necessary falsehoods (contradictions). The truth table apparatus provides the explicatory sense of ‘all possible,’ viz. “all possible interpretations into a two-element Boolean algebra,” where ‘possible’ is now being used, once again, in the philosophically unproblematic sense of “combinatorially possible.” It is at this point that the blurs come in. “Necessary” truths are metalinguistic. Tautologies are “necessary” truths. Hence, tautologies are metalinguistic. The fallacy is obvious. Yet, that Wittgenstein was misled by just this equivocation, I have no doubt. One has merely to read 6.1264: “The significant proposition (sinnvollle Satz) asserts something and its proof shows that this is so; in logic every proposition is a form of proof. / Every proposition of logic is a modus ponens presented in signs. (And modus ponens cannot be expressed by a proposition.)” The assimilation of object-language tautologies to metalinguistic rules of deduction is obvious. Note, too, the explicit denial that tautologies are propositions. Note also

406 that both tautologies and contradictions, on the one hand, and ill-formed statements, on the other, are meaningless – though, to be sure, the former are sinnlos while the latter are unsinnig (see 4.461 for the former, 4.0611 for the latter − of this, more in a moment). The denial that tautologies are propositions arises from the same blur, in effect the blur between object language and metalanguage, though it arises in a slightly different way. Upon the first explication of “necessary”, “possible” and “impossible”, all facts are possible, none necessary and none impossible. But tautologies and contradictions are necessary and impossible respectively. Hence, they don’t represent facts. Hence, they are radically different from such factual truths as (say) “p1 v p2". The equivocation on ‘fact’ is clear: once it means what is represented by a true sentence when one is operating in a context which involves only atomic sentences, and once it means what is represented by any sentence that is non-tautologous and non-contradictory, be it atomic or molecular.23 That Wittgenstein was misled by this equivocation, I again have no doubt. Proposition 3.3 tells us that “Only the proposition has sense (Sinn) ...”, and in 4.461 we find that “Tautology and contradiction are without sense (sinnlos).” It follows that tautologies and contradictions are not propositions. And indeed, we have just seen this in 4.1264, which says, as we noted, that “The significant proposition (sinnvolle Satz) asserts something ...”. So a tautology or a contradiction, which are senseless – sinnlos – , are not significant propositions – sinnvolle Satz – , and are, therefore, one gets the idea, much of a piece with other strings, the ill-formed strings, which are also not significant propositions.24 But of course, even if they are not propositions, they nonetheless are well-formed. Wittgenstein immediately realizes this, for 4.4611 is immediately qualified with: “Tautologies and contradictions are not, however, meaningless (unsinnig); they are a part of the symbolism ...”. But even if they are not quite meaningless, they are still not really genuine propositions. In this respect they fall in the same category as formation rules: they too are not propositions.25

One concluding remark. As is well known, according to the Tractatus every atomic sentence contains at least two names.26 Proposition 2.05 affirms that from a list of all true sentences we can obtain a list of all the false sentences. That is the minimal interpretation which can be given to the term ‘determine’ in that Proposition.

407 A nominalist is one whose ontology contains only particulars,27 no characters understood as entities capable of being exemplifiable by more than one particular (though they might be characters taken as tropes, entities as particular as the individuals they characterize). Such an ontologist could very well hold that there are “monadic facts.”28 Indeed, I can see no reason why this ontologist would deny the existence of “monadic facts.” Or, to say the same thing differently, I can see no reason why a nominalist would want to insist that all facts are at least “dyadic”, containing at least two objects. The realist can, as we saw, give good philosophical reasons for denying the existence of “monadic facts.” He or she wants, for example, to avoid Platonism, that is, the view that there are unexemplified characters, or wants to avoid unqualified particulars. But I can think of no such reason a nominalist might advance for denying the existence of “monadic facts.”29 Now, in the Tractatus, every sentence involves at least two names. If there were “monadic facts” in the Tractatus’ ontology, then these would be represented by strings consisting os a single name. One could not, therefroe, from nothing but a list of all true atomic sentences (concatenation of names) obtain a list of all false atomic sentences (or, what is the same, the list of sentences representing all negative facts).30 For there would be the “monadic facts” which would not be mentioned by the sentences of the first list, but which would have to be mentioned by sentences of the second list. Thus, if the Tractatus were nominalistic, there would be no reason to deny the existence of “monadic facts,” and therefore no reason to affirm the minimal interpretation to be given to 2.06, viz. that the list of all false sentences can be obtained from nothing more than a list of all true sentences. But at the same time, if the Tractatus is realistic, that is, has an ontology containing both particulars and characters understood as universals, then 2.05 is simply the assertion of PE. And there are good reasons for one with such an ontology to affirm PE. In short, if the Tractatus is nominalistic, it has no reason to affirm what 2.05 affirms, but if it were realistic there are good reasons to affirm what 2.05 affirms. This provides further evidence in support of the realistic interpretation of the Tractatus.

408 Endnotes to Study Ten

1.L. Wittgenstein, Tractatus Logico-Philosophicus, Intro. by Bertrabd Russell, trans. C. K. Ogden, with original German on facing pages ( London : Kegan Paul, Trench, Trubner & Co., 1922). 2.See, for example, George Pitcher, The Philosophy of Wittgenstein (Englewood Cliffs, NJ: Prentice-Hall, 1964), p. 48; and Erik Stenius, Wittgenstein's Tractatus : A Critical Exposition of Its Main Lines of Thought (Oxford: Blackwell, 1960), p. 51. Stenius holds that the Tractatus is realistic in its ontology, while Pitcher holds it to be nominalistic. That, as we shall see in the concluding remark of the present study, makes the former’s worry less understandable than the latter’s. 3.This shift between atomic states of affairs (what language is about) and sentences (language), in this paragraph and throughout, is justified by Proposition 4.0311: “One name stands for one thing, and another for another thing, and they are connected together. And so the whole, like a living picture, presents the atomic state of affairs.” 4.(L*) is the list of all false sentences. With the usual truth-table stipulation for negation, that is, for “~”, this list can be changed into a list of true sentences simply by placing a ‘~’ in front of each of the false sentences. 5.For the sense of ‘(in)dependent’ in this and the preceding sentence, see note 14, below. 6. I ignore relations. See note 15, below. 7.This realistic reading of the Tractatus is traditional. Ramsey, Bergmann and Stenius have all asserted it, indeed more or less taking for granted, it seems that clear. But some have argued that this reading is mistaken, and that the work is nominalistic. E. B. Allaire has ably disposed of their arguments in his essay, “The ‘Tractatus’: Nominalistic or Realistic?” in E. B. Allaire et al, Essays in Ontology (The Hague: Nijhoff, 1963), pp. 148-165. The present essay supports this traditional reading by explaining the otherwise obscure Proposition 2.05. See my concluding remark. 8.The ‘infer’ here is meant to convey a logical force. This is justified if the Tractatus construes PE as a necessary truth. The Tractatus does so construe PE. See note 17, below. 9.Max Black holds that “There seems, on the face of it, to be a conflict between 2.062 and 2.05” (Max Black, A Companion to Wittgenstein’s Tractatus [Cambridge: Cambridge University Press, 1964], p. 72). What has just been said shows that one need attribute no such conflict to the Tractatus.

409

10.This shows that ‘Tatsache’ must here mean “atomic fact.” For, if the sentence ‘p’ were true and was “changed” to false, it is not true that everything would remain the same if one were talking about molecular facts: ‘~p’, for example, which was false would “become” true. But it is also true that Wittgenstein also speaks of “negative facts” (negative Tatsache), as in Proposition 2.06. This shows that ‘fact’ sometimes means (indeed, it usually means this) what is represented by a true atomic sentence, but that (less commonly) it means what is represented by any true sentence, be it atomic or molecular. This blur, between a narrower and a broader sense of ‘fact’, plays, as we shall see, an important role in the Tractatus. See note 23, below. 11.Proposition 2 reads: “What is the case, the fact, is the existence of atomic states of affairs.” 2.02 reads: “An atomic state of affairs is a combination of objects ...”. 12.Wittgenstein is here explicating the sense of ‘(in)dependent’ mentioned above, in note 5, in which characters and particulars are mutually dependent, each kind upon the other kind. At least, that is what he is doing if the ontology is realistic. For a discussion of the issues centring around “independence,” see E. B. Allaire, “Existence, Independence and Universals,” Philosophical Review, 69 (1960), pp. 485-96; reprinted in E. B. Allaire et al., Essays in Ontology, pp. 3-13. 13.This logical and ontological independence of things is grounded in the very being of objects – each object is what it is and not another thing. This is the core of logical atomism, be it in Russell, Wittgenstein, or Bergmann. Such a logical or ontological atomism is justified philosophically by appeal to a Principle of Acquaintance. See the essay on “Acquaintance, Ontology and Knowledge,” in this volume, above. Wittgenstein has logical atomism as central to the ontology of the Tractatus, but then waffles when it comes to colour incompatibilities– red is what it is, he insists, but he also insists that it is part of the being of red that something red is not green. On this issue, with regard to the Tractatus, see E. B. Allaire, “Tractatus 6.3751,” Analysis, 19 (1959), pp. 100-105; and on the issue more generally, see G. Bergmann, “The Revolt against Logical Atomism,” Philosophical Quarterly, 7 (1957), pp. 323-339, and 8 (1960), pp. 1-14; reprinted in his Meaning and Existence (Madison, WI.: University of Wisconsin Press, 1959), pp. 39-72. 14.See also the essay on “Effability, Ontology and Method,” in the present volume, below.. 15.I have deliberately ignored relations. If these are introduced into the ontology, the dialectics of dependence-independence are slightly more complicated. See H. Hochberg, “Elementarism Independence, and Universals,” Philosophical Studies, 12 (1961), pp. 36-43; reprinted in E. B. Allaire et al., Essays in Ontology, pp. 22-29. Wittgenstein largely ignores the complication that relations bring to the dialectics of independence. 16.See also 3.3: “Only the proposition has sense (Sinn); only in the context of a propo-

410 sition has a name meaning (Bedeutung, i.e., reference).” No sort of object can exist as the referent of a name standing alone; no sort of object can exist as the referent of a name unless that name occurs in at least one atomic sentence – where that sentence has as its meaning (Sinn) the fact which consists of that object in combination with another object. This proposition is, in other words, a restatement of PE. But Wittgenstein is also arguing here: he is restating his point that particulars and characters are mutually dependent as an argument against Frege. 17.That it is a necessary truth justifies the use of ‘infer’ above. See note 8, above.. It also accounts for the logical force of ‘determine’ in Proposition 2.05. 18.Cf. Proposition 6.124: “The logical propositions describe the scaffolding of the world, or rather they present it ...”. 19.Once again, Proposition 3.3: “Only the proposition has sense; only in the context of a proposition has a name meaning.” 20.For an extended treatment of the virtues of explicating philosophically problematic uses of ‘necessary,’ ‘possible,’ etc., by means of combinatorial possibility, see F. Wilson, Laws and Other Worlds (Dordrecht, Holland: D. Reidel, 1986). Naturally enough, the current fad to take possible worlds as ontologically real, in an unexplicated sense of ‘possible,’ is not to be taken seriously. But Laws and Other Worlds does at least undertakes to explicate, as accounts such as that of D. Lewis do not., the concept, which is clearly philosophically problematic, of possibility as it appears in the notion of a “possible world.” The idea for the proposed explication goes back to the Tractatus, of course. On this issue of the relevance to ontology of the concept of “possible world,” as that notion is currently used, compare H. Hochberg, “The Radical Hylomorphism of Bergmann’s Aristotelian Metaphysics and the Ontology of Relations,” The Modern Schoolman, 78 (2001), pp. 275-288, at p. 282. Hochberg (p. 282 note) draws attention to G. Bergmann, “The Philosophical Relevance of Modal Logic,” Mind, n.s. 69 (1960), which argues quite clearly, and as conclusively as these things can get in philosophy, that modal logics of the sorts described and axiomatized by C. I. Lewis, have little philosophical relevance, and none for ontology. Bergmann’s paper seems to have been systematically overlooked by those many who now take such Lewis-type modal logics seriously. One would have thought a serious search of the relevant literature would have preceded the efforts of those many who now fill the journals with the dots and dashes, and the diamonds and squares and funny arrows, and all those other chicken scratches, that are characteristic of essays on modal logics and their purported applications to metaphysics. 21.As Wittgenstein speaks, a well-formed string has a sense in the sense of Sinn; an

411 ill-formed string is unsinnig. 22.Suppose ‘red’ names an object. That makes “‘Red’ is a name” a true statement of the metalanguage. It is not the formation rule alone that determines all possible worlds, in the just explicated sense of ‘possible world’; the list of names also enters into this determination. That means that, like the formation rule, “‘Red’ is a name” is a necessary truth. And as some now speak, this implies the ‘Red’ is a rigid designator. In any case, this shows the reasonable sense of what some have meant when they have asserted that “Red exists” is a necessary truth. Wittgenstein sees this quite clearly. Or at least, I see no other way to explain the following sequence of Propositions. 2.02: “The object is simple.” 2.021: “Objects constitute (bilden) the substance of the world (Welt).” 2.022: “It is clear that however different from the real one an imagined [possible; cf. Proposition 3 ff] world may be, it must have something – a form – in common with the real world.” 2.023: This fixed form consists of objects.” It is not surprising, I think, that the later Wittgenstein was worried both about the status of such statements as “Red exists”, and about Platonism (which amounts to worries about PE, since it is that which secures that there are no unexemplified characters, and since it is a worry about PE it amounts to a worry about the formation rule for atomic sentences of a perspicuous language). See The Blue Book (Oxford: Blackwell, 1960), p. 31 for the former, p. 17 for the latter. 23.We saw earlier that Wittgenstein makes just this blur on the use of ‘fact’. See note 10, above. 24.Compare also, once again, 2.221-2.225: “What the picture represents is its sense (Sinn).” “In the agreement or disagreement of its sense with reality (that is, Wirklichkeit), its truth or falsity consists.” “In order to discover whether the picture is true or false we must compare it with reality.” “It cannot be discovered from the picture alone whether it is true or false.” “There is no picture which is a priori true.” All this makes sense only if one is restricting oneself to atomic sentences (propositions), excluding molecular sentences. 25.The assimilation of tautology to formation rule and the impossible to the meaningless came to dominate the later Wittgenstein; compare E. B. Allaire, “Tractatus 6.3751”. 26.Cf. 4.22: “The elementary proposition consists of names. It is a connection, a concatenation, of names.” 27.Miss Anscombe has these as bare particulars (see G. E. M. Anscombe, An Introduction to Wittgenstein’s Tractatus [London: Hutchinson University Library, 1959], ch. 7), but this reading is quite implausible. Rather more plausible is a reading which takes objects to be tropes, as in Wilfrid Sellars, “Naming and Saying,” in his Science, Per-

412 ception, and Reality (London: Routledge and Kegan Paul, 1963), pp.225-246. 28.Note that this is not really possible for one who speaks of facts in the way Wittgenstein does. For the latter, a fact, that is, a basic or atomic fact, is what is represented or pictured by a true atomic sentence. Since an atomic sentence contains at least two names, each Tractarian fact contains at least two objects. 29.This is pointed out by Wilfrid Sellars in his “Naming and Saying,” p. 237ff. 30.For that matter, what would a negative “monadic fact” look lie? A “monadic fact” would be represented by a single name. But you can’t negate names! Maybe, it might be suggested, one could think of it as represented by a name which does not name. But what sense does that make? After all, a name that does not name is merely a meaningless mark or noise.

Eleven Grossmann on the Categorial Structure of the World* Reinhardt Grossmann’s study on The Categorial Structure of the World 1 is one of the most significant of recent works in ontology. Although Quine convinced a number for some years that the substance of ontology is to be found in the formula that “to be is to be the value of a variable,” this formula in fact for the most part described the appearances, not the reality. For, as Grossmann contends, rightly, I believe, all appearances to the contrary notwithstanding, it is not the existence of certain entities which is the subject of ontological dispute, but rather their nature: to what ontological categories are they correctly to be assigned? In this book, Grossmann’s aim is “to give a complete and accurate list of the categories of the world” (xv). But given his conviction about the nature of ontology, the result is that when he deals with a given category, he not only argues for that categorization but also examines alternatives. Thus, Grossmann argues both that individuals are correctly categorized as particulars, and incorrectly categorized as bundles of properties; properties are argued to be universals and not particulars (that is, tropes located in space and/or time), numbers are argued to be quantifiers and neither multitudes of units nor properties of properties nor classes of either properties or classes. At the same time Grossmann attempts a systematic examination of the ontological innovations since Descartes. Until the latter, ontology had as its basis the Aristotelian ontology of substance and accident.2 Descartes himself was a member of this tradition. In fact, it is still a major force. But for all that the latter does remain so, the substance tradition was sharply criticized – decisively criticized – in the seventeenth and eighteenth centuries.3 The Aristotelian doctrine of substance was attacked from the left by the empiricists (Berkeley and Hume) and from the right by the rationalists (Spinoza and Leibniz). As a consequence, ordinary things, individuals, 1

Originally appeared as a Critical Notice of Reinhardt Grossmann, The Categorial Structure of the World (Bloomington: Indiana University Press, 1983), Canadian Journal of Philosophy, 16 (1986), pp. 163-180.

414 were re-categorized as bundle of properties, and the Aristotelian distinction between essential and accidental properties disappeared. An issue never adequately faced by the Aristotelian tradition – what are relations? – acquired new force. New questions arose – what is the difference between a bundle of properties and a mere class of properties? what precisely is the role of relations in the constitution of bundles? – and these force upon one as never before the question, what is the ontological status of relations? Pierce and Russell were to clarify the notions of relations and quantifiers, Frege and Russell were to clarify the notions of number and class. Bolzano, Meinong and Russell were to clarify the notion of fact. The nineteenth century saw in effect the final rejection – alas, not the elimination — of the Aristotelian ontology, and, as part of that rejection, the introduction of a number of new ontological categories: relations, classes, facts. Grossmann attempts to take stock of these innovations, to sort out what is reasonable and unreasonable, and to defend a system of categories that does justice to these developments. The Aristotelian tradition did not begin by looking about the world and asking what sorts of things do exist and then proceed to categorize them. Rather, it argued on an a priori basis that there were substances and accidents, and each substance had a nature or essence which determined the kind of substance it was and which it shared with other substances of the same kind, and if there were things we knew that did not fit these patterns – e.g., relations – then they were declared not to exist. Far from being a cautious sort of empiricism, as its defenders so often claim, the Aristotelian tradition proceeded on the basis of dogmatically asserted a priori principles. Nor could it do otherwise. It had its own a priori thesis about knowledge and this was imposed on the world in a way that simply denied the phenomenological detail that did not fit the patterns. This dogma about knowledge was that “like knows like” and this was interpreted to mean that the knowing mind became identical with the object known. Since the mind could not literally become the material object presented in sense, sense could not be a form of knowledge; what came to be known, rather, what came to be in the knowing mind, was the nature or essence of the sensed object. This doctrine of the two eyes – the eye of the senses, that is, awareness of sensible particulars which the object known presented to the knower, and the mind’s eye, which knew the universal, the nature or essence in the material object known – this doctrine of the two eyes is the most pernicious feature of the Aristotelian legacy. It prevented philosophers from recognizing that in sense experience (and in inner awareness)

415 we are presented not only with particulars but also with their properties, and not just particulars and properties but also the facts of the particulars having certain properties; and not only properties but also relations and the facts of various particulars standing in certain relations. When the empiricists attacked Aristotelian natures and essences, they also attacked the mind’s “eye” that saw these things. That left them with the eye of the senses that grasped only sensible particulars. The resulting nominalism was, of course, untenable, and one can again and again feel Locke, Berkeley and Hume struggling to affirm phenomenological fact while still gripped by this uneliminated Aristotelian dogma about the eye of the senses. However, as the rejection of substances forced an examination of the problem of relations, so the examination of the latter permitted a reexamination of the relation of a mind to what it knows. This elimination of the dross of the Aristotelian tradition permitted philosophers to accept, on the one hand, without dogmatic distortion phenomenological accounts of what is given in experience,4 including acts of experiencing themselves; and also permitted them, on the other hand, to develop an account of experiencing adequate both to phenomenological detail and the philosophical demands of categorial analysis. Grossmann’s method consists in just this genuine empiricism that takes seriously as its starting point that which is given in experience − that is, that which is given both in sensible experience and in inner awareness. This empiricism recognizes that what we are aware of are complexes – states of affairs, or, equivalently, facts – and their constituents; and also that among the states of affairs with which we are presented are acts of experiencing – including perceiving, remembering, believing, thinking, supposing, loving, hating, willing and so on. What this empiricism then attempts to do is categorize these entities in a philosophically satisfactory way, and to lay out the laws that describe how the entities in the various categories relate to each other. The point is to establish categories appeal to which allows the solution of various philosophical problems.5 I Grossmann divides the world into two basic categories: things and facts. Among things one finds individuals, properties and relations. Individuals are characterized by properties from the various sensory dimensions, and among these the spatial and temporal ones are especially important. In particular, an individual is, according to Grossmann, a temporally enduring

416 entity with other individuals as temporal parts. As for properties, there are not only properties but also properties of properties. Relations are, contrary to the Aristotelian tradition, irreducible, and have a sense − a relation goes from one thing to another − and, moreover, contrary to Bradley, can obtain between entities without being related to them in turn. Grossmann also admits classes as a category distinct from properties. Contrary to Russell, classes are not eliminable in favour of properties.6 Russell of course so argued in order to avoid the well-known paradoxes. Grossmann’s discussion of this topic is especially challenging. He argues that the traditional responses are wrong – type theory, the denial that properties can have properties, the class-set distinction, the vicious circle principle are all canvassed and found wanting. Rather, Grossmann suggests, the paradoxes are nothing other than straight-forward non-existence proofs which happen to shock our intuitions. Another class of things, according to Grossmann, is that of the quantifiers. Numbers are among the quantifiers, and it is argued that attempts to treat numbers as “multitudes,” as properties, or as classes are mistaken. In particular, the Frege-Russell “definition” of a number as a class of classes is not a definition, in the strict sense, but rather is an informative identity statement.7 Besides numbers there are also the general quantifiers, all and some, but also, Grossmann argues, no and the. In addition to things, there are, as we have said, facts: Grossmann argues that factuality is a fundamental category. What one is presented with in experience are states off affairs. States of affairs consist of things standing in various relations. Thus, we are presented with Jones kicking Smith, Ed exemplifying (the colour) gray, grey exemplifying (the property of being a ) colour, Salzburg being the birthplace of Mozart, and of course cases like Macbeth seeing a dagger, and so on, But as the last example makes clear, not every state of affairs with which we are presented exists. Facts are those states of affairs which exist. The intentional relation (a thing) connects thoughts to the states of affairs those thoughts are about; being presented with a state of affairs consists of one’s present thought intending, standing in the intentional relation to, that state of affairs. Since thought can be about a state of affairs that does not exist, the intentional relation is unlike ordinary relations, such as kicking or even the relation of exemplification, all of which connect only existing entities. Intentionality can relate the existent to the non-existent. A final category of thing that Grossmann proposes is that of struc8 tures. Structures include the wholes the gestalters described, definite de-

417 scriptions (which are abstract structures of a certain kind and are to be distinguished from description expressions. Structures presuppose relations; they consist of entities in relations to each other. Structures may be, and often are, isomorphic to each other. However, since isomorphic structures need not be identical, there are no implicit definitions, and, for the same reason, entities with a shared isomorphic structure cannot, for that reason anyway, be reduced to each other. Structures are, in effect, complex features common to several fact, as, for example, the axioms of a Boolean algebra define a complex structure that appears in every fact that consists in a set of entities satisfying those axioms. Two further categories stand out as unique. Grosmann argues that there are not two modes of being, one for things and one for facts. There is, to the contrary, only one mode of being, namely existence.9 Existence is not of course a property of non-existent entities; for, existence can never belong to an entity that does not exist. But also it is not a property of existent entities. A property divides things into two classes, those that have it and those that don’t. But existence does not share this feature; it is not something that an existent may or may not have. While, on the one hand, an existent entity and existence are not one and the same thing, nonetheless, on the other hand, existence is not something “added on” to an existent. That is, while any property – including such categorial properties as that of being a thing or that of being a particular (i.e. the property of particularity) – stands in the relation of exemplification to an entity, existence is not exemplified at all; existence belongs to an entity much more intimately. It is, Grossmann argues, different also from exemplification, from concreteness, and from object of thought; it is rather most appropriately to be considered as the “ultimate substratum for all attributes” (403ff). Like existence and unlike categories such as thing and property, negation is not a property. Grossmann argues that negation attaches itself to states of affairs; there are no negative properties, nor is there a relation of negative exemplification. Negation, clearly, attaches only to non-existent states of affairs. When negation attaches to a non-existent state of affairs the result is a (complex) fact, an existent state of affairs. In this respect, negation is like the intentional relation in being able to attach to what is not there (359). It is evident that Grossmann’s position has the odd consequence, which he accepts (416), that non-existent entities, namely, nonexistent states of affairs, can be constituents of facts, that is, existent states of affairs, and specifically they can be constituents of the facts that are the negations of the non-existent states of affairs.

418 This has certain similarities to Meinong’s position that non-existent things can be constituents of facts, e.g. the non-existent golden mountain which Meinong maintains is a constituent of the fact that the golden mountain is golden. Grossmann, however, rejects Meinong’s position.10 He accepts the principle that “No simple object of whatever kind can be part of a state of affairs before the mind as the object of any sort of thought unless it has first been perceived or experienced.”11 This is in effect the traditional empiricist Principle of Acquaintance (PA). We do perceive negation, that is, perceive it in perceiving a state of affairs of which it is a constituent, e.g. when we perceive that Peter is not in the room; or again, we do perceive numbers, that is, perceive them in perceiving a state of affairs of which they are constituents, e.g. when we perceive that there are three apples in the basket. But we do not perceive the golden mountain; we have never, in our sense perception, observed it. This and PA has the consequence that the golden mountain in person as it were is never part of an object of thought, nor, therefore, is it a constituent of the state of affairs that is before my mind when I think that the golden mountain is golden. Rather, Grossmann argues, the state of affairs that the golden mountain is golden is a complex state of affairs involving the quantifier the, and, more specifically, the special structure, called a description, which is represented by the linguistic expression ‘the golden mountain’. Note that on this view, the description is not a linguistic entity (269ff); it is, rather, what the linguistic expression represents. As for the connection between the (non-linguistic) description and what it describes, that is established by the relation of identity (278ff). In particular, the description is not in any sense part of what it describes. The point about the description the golden mountain is that there is no existent state of affairs of which it is a constituent and in which it is related by identity to some individual. Thus, when one asserts that the golden mountain does not exist, what one is asserting is that “there is no thing that is (identical with) the golden mountain.” As for the assertion that the golden mountain is golden, this true statement represents the fact that the property golden is part of the description the golden mountain. If, with Meinong, one holds that the expression ‘the golden mountain’ names a non-existent thing, then not only does one violate PA, but one ends up doing considerable violence to the laws of logic, the straightening out of which requires nothing short of vast tomes.12 Grossmann’s position on descriptions, in contrast, requires neither the abandonment of PA nor any modification of the laws of logic. On either Meinong’s or Grossmann’s view, the laws of logic are very general laws concerning rela-

419 tions among entities. For Grossmann,, propositional logic formulates under what general conditions states of affairs obtain or exist, while predicate logic formulates under what general conditions properties are exemplified and relations hold (360ff). Necessity is a matter of lawfulness, and there are different kinds of necessity, logical, physical, etc. (367ff). Nagel’s “Logic without Ontology”13 convinced many that this view of logic is false. But Morris Cohen had earlier held the opposing position: the laws of logic have genuine grounding in the ontological structure of the world. The notion that formal propositions are empty of meaning is a persistent radical confusion. It is true in a sense that every form is independent of its matter. But a formal act is one that is the same for all regardless of the individual differences in the class to which is applies; and so the rules of logic or pure mathematics universally apply to all propositions irrespective of differences of their material content. But this does not mean that logical or mathematical forms can exist apart from all reference to any possible content. On the contrary the most formal propositions are those which apply to all kinds of entities, and reference to such possible application is essential to their meaning.14

Grossmann’s is a return to this earlier tradition,15 one that is, one must say, more sensitive to the ontological issues than is Nagel’s. Nagel’s thesis is that the principles of logic are useful tools for keeping discourse consistent.16 What Nagel’s argument neglects is that two valued logic maintains consistency only if there is in the form of the world a ground for the applicability of two-valued logic: the world must be two-valued. The position of Grossmann, like that of Cohen, attempts to do justice to the ontological ground of logic.17 The point is that for Grossmann, in contrast to Meinong, the ontology does not require an extraordinary logic; to the contrary, the laws of logic that it requires are none other than those on which there is uniform agreement that these are the laws of logic. All this is equally true for Grossmann’s claim that non-existent states of affairs can be constituents of facts. In contras to Meinong’s non-existing things, non-existent states of affairs do not violate PA, since they are complex and even if they do not exist but are presented in (non-veridical) perceptions their simple constituents can be existing things that do not violate PA provided only that these constituents have been presented in other acts of perception. Moreover, Grossmann’s claim that non-existent states of affairs are constituents of facts, that is, of existent states of affairs, is fully compatible with the ordinary laws of logic, again in contrast to what Meinong has to hold with respect to non-existent things.

420 And in addition, Grossmann’s claim that non-existent states of affairs can be constituents of facts fits in with his other thesis that the intentional relation is one that can (in the case of false thoughts) connect (existing) things to non-existent states of affairs; in this case, too, non-existent states of affairs are constituents of facts, to wit, facts to the effect that this (existent) mental act or thought intends this (non-existent) state of affairs. Thus, in contrast to Meinong’s jungle, Grossmann’s ontology is sober and down to earth. At bottom, unlike Meinong, Grossmann is not prepared to let the dialectic move him wither it will; for Grossmann, the dialectic of philosopy must always operate within the empiricist constraints. Nonetheless, there will remain many who prefer landscapes that are not only less luxuriant than Meinongian jungles, but less luxuriant, too, than the ordered parklands of Grossmann’s ontology. But a preference for deserts, like one for Ockham’s razor, is not a sound method of argument in ontology. However, Russell’s insistence that philosophers have a robust sense of reality does have its point. Arguments phrased in such terms may well reflect a disquietude that is fully legitimate. For an empiricist at least, what they signal is that the ontology is not just proliferating (kinds of) entities, but is doing so in violation of the fundamental empiricist commitment to restricting the existence claims of one’s ontology to entities given in perception and experience. Thus, consider negation. We do have acts of perceiving which have negative states of affairs as their intentions, as when we observe that Pierre is not present in the room. Thus, Sartre gives a well-known account of negation in Being and Nothingness.18 Sartre reports how, upon entering a café and looking about for an acquaintance of his, Pierre, he notices that “He is not here.” Sartre initially questions the idea that he has actually been given this fact in perception; it seems strange to him to say that one can perceive an absence or that one is presented with – nothing. Sartre reflects, however, on the situation, both on his looking over all the faces and things which are not Pierre, and on the growing recognition that Pierre really is not there, and comes to realize that he is indeed presented with a something that is negative, that there is there in world, and given to him, a negativity. In his awareness of things in the world, he is acquainted with a “flickering of nothingness” that serves as the foundation (truth maker) for his judgment that “Pierre is not here.” Grossmann, as I have said, argues convincingly that this feature of negativity cannot be categorized in terms of, or, perhaps more accurately, analyzed away, first, in terms of incompatibility19 – incompatibility presupposes negation rather than accounts for it20 (355); nor,

421 second, in terms of negative exemplification21 – this as it were fragments negation into many cases of different negations, one for each kind of atomic fact22 (356); nor, third, in terms of negative properties23 – the relations between a property and its negation, e.g., mutual exclusion, will have to be more than contingent and in fact quite unlike other relations among properties.24 (357) He concludes (358) that therefore negation is an uneliminable ontological category of its own. One may well want to grant that something like this is so, without going on to agree with Grossmann that this implies that besides all the non-negative facts in the world, every true negated statement also represents a fact. Suppose that ‘Fa’ is true and that ‘Ga’ is false Then it is certainly plausible to include in one’s ontological inventory the fact Fa But is it also plausible to include as Grossmann does (359), ~Ga and not only these but in addition, as Grossmann also does (352), such further facts as ~~Fa and ~~~Ga etc.? The empiricist feels uncomfortable. Yet, given that the empiricist begins with perception, as Grossmann insists, and the undeniable fact that negative states of affairs are intentions of acts of perceiving, then how does one avoid admitting this abundance and superabundance of facts into one’s ontology? In particular, can one avoid admitting any negative facts? Russell once remarked that “There is implanted in the human breast an almost unquenchable desire of finding some way to avoid the admission that negative facts are as ultimate as positive.”25 Russell argues that we need negative facts as truth-makers for certain sentences. ...it is better to take negative facts as ultimate. Otherwise you will find it so difficult to say what it is that corresponds to a proposition. When, e.g., you have a false positive proposition, say ‘Socrates is alive’, it is false because of a fact in the real world. A thing cannot be false except … because of a fact, so you find it extremely difficult to say exactly what happens when you make a positive assertion that is false, unless you are going to admit negative facts.26

422 Grossmann argues that we cannot avoid negative facts. “I think it is just plain common sense that there are such negative facts as that the earth is not flat and that three times three is not ten” (355) Since negation cannot be eliminated, he concludes, “therefore, that negation is an entity which forms its own category” (358).27 This not quite how Russell puts it, and puts the argument, but they amount to the same. Bergmann, too, argues that negative facts are unavoidable.28 But the fact that the desire to avoid negative facts in unquenchable is telling. Like the desire for desert landscapes. Is it really necessary to admit negative facts? One can, I think, give a negative answer to this question: there are no negative facts. Though I should hasten to add that it does not follow that there is no objective ground for negativity. This position is contrary to that of Grossmann. One can defend this position, I think, by developing two, not unrelated, lines of criticism against Grossmann (and Russell and Bergmann). II The first line of criticism involves the recognition that granting ontological status does not always consist in admitting an entity. Sartre gives an example. It is rather complicated for purposes of ontological discourse, but the idea is clear enough. After seeing what is in the room, Sartre concludes Pierre is not there. To simplify (but not, I think, to falsify), suppose that instead of the room in the café, we are given in perception the very simple world consisting of the facts Fa Gb where the individuals a and b (Pierre and Maggie, if you wish) exemplify, respectively, the properties F and G. We have represented the structure of entities in this world perspicuously in language using two sentences. These sentences each have the form (*) fx where ‘f’ is replaced by a term referring to a property and ‘x’ is replaced by a term referring to an individual.29 Given what we perceive, (*) represents a pattern in what is perceived. But of course, other combinations of entities we perceive are not among the combinations or facts given in perception but are compatible with (*).30 We have in fact the following list of combinations that share the perceived pattern (*): Fa, Gb, Fb, Ga

423 The first two are actual: they are given in perception. But the second two are not actual: they are merely possible. So, what does the latter mean? It does not mean that besides the two actual facts there are in addition two other entities, namely, the two merely possible facts Fb, Ga. Grossmann, of course, would argue otherwise. For him, there are indeed two further entities, namely the two actual negative facts ~Fa, ~Ga But however one talks, it is clear that these amount to much the same: there is little difference ontologically between talk of “merely possible facts” and “actual negative facts.” The point to be made is that in our simple world, it is not the case that in addition to the two positive facts, there are two further fact-like entities. Or at least, although it is clear that there are combinations that are possible but not actual, we are not required to construe this to mean that besides the two actual facts there are in addition as further entities two merely possible facts. We can instead proceed as do Humeans with respect to causation,31 and treat the formation rule (*) as a law, which on the one hand simply describes a pattern discovered in the world that we perceive but on the other hand places constraints upon the combinations which the mind can think.32 In this sense, there is no possibility apart from the capacity of the mind to move beyond the actual; and what counts as possible and not merely illformed nonsense is determined by the constraints on thought which are constituted by the habits of thought – in Hume’s terms, the “association of ideas” – imposed on thought by the patterns that thought discovers in the entities which it experiences. That is, the entities that we experience share the pattern (*), and this experience as it were imprints this pattern on the mind: when thinking of this world and the things in it, we are habituated to conceive of things in this way, i.e. the pattern (*), rather than some other way. So we think of Fb, Ga as non-actual but really possible combinations, whereas the patterns aa, FG are equally non-actual, but, given our habits formed by what is actual, are simply not possible.33 It is not true, then, that in this small world that we are considering that Fa and Gb are facts while Fb and Ga as mere possibilities are separate entities in their own right (“possible facts”). Rather, the ontological ground of those possibilities consists in the observed pattern (“constant conjunc-

424 tion”) that individuals always exemplify properties and properties are always exemplified by individuals.34 No other combinations are observed, so it is just combinations of this sort and no others that appear in our representation of this world. Surveying this world, we are led to expect this sort of combination and not any other sort. Given the actual world, it is this sort of combination that one by habit comes to expect, such combinations alone constitute what is possible: what is possible is the projection by habit of a pattern that pervades the actual. Thus, while one can insist that possibility has an ontological ground, one does not have to insist that that ground is an entity, or, rather, an entity over and above the (actual) facts of the world. Possible states of affairs are described by strings, combinations of names of individuals and predicates, and specifically, those strings which are compatible with the formation rule (*). Actual states of affairs or facts are those possibilities which occur in the picture that we draw of the world that is given to us through our perceivings. Those strings are true, we say, that describes the facts in the world given that is given to us perceptually; and those strings are false that do not occur among the strings that are true. The last can be put in a way that gives the illusion that besides actual facts, that is, one should no doubt insist, besides the facts, there are also possible facts – there is in our world no entity that is a possible fact, nor, therefore, any facts that, to contrast them with the merely possible, must be called actual – there are in the world only facts: but we might create the illusion that besides the facts there are other entities which are the merely possible or even the false facts – for, we might say, albeit misleadingly, that those strings that are false do describe states of affairs, and that those states of affairs are the ones that are merely possible. But we do have the list of strings the formation rule for which states that (*) is their prototype.35 This list is a list of possible states of affairs. This list divides into two sets, those strings which are true and those which are false. These sets are mutually exclusive.36 The world permits and requires this sort of division: this is a fact about the world, that it is twovalued. It is this two-valuedness of the world that provides the ontological basis for our language having the logical structure that it does.37 Our capacity to establish a formation rule that permits us on the basis of perception to unequivocally divide possibilities into the actual and the merely possible is not only psychological but depends upon the world that is perceived having a certain pattern or structure: this pattern is the ontological ground of the two-valuedness of the world.38 Again, what is important is

425 not an entity but rather the patterns that we discover in the entities that we perceive. Negation now enters in a fairly obvious way. It is (to speak succinctly) the object-language reflection of the metalinguistic predicate ‘false’. That is, (again to speak succinctly), the two-valuedness of the world permits, via the usual truth-table definitions, the introduction of negation into our object-language description of the world. Negation thus has an ontological ground, namely, the two-valuedness of the world, but that ground is, contrary to Grossmann, not an entity. Consider, now, our object-language to be so extended as to include negation. The small world that we are supposing that we perceive is fully described by the strings (sentences) (x) Fa, Gb but the set of true sentences about it will be (p) Fa, Gb, ~Fb, ~Ga What Russell argued, and following him Grossmann, is that each of the sentences in (p) requires a truth-maker, and that therefore we have to admit negative facts as the truth-makers of the negative sentences. Which means we need negation as an entity – one, indeed, in its own category. One might argue that once one is given that the set of sentences of (x) is the set of true sentences, then the absence of ‘Fb’ and ‘Ga’ from this list shows that these two sentences are false, i.e. that the sentences ‘~Fb’ and ‘~Ga’ are true. The truth-makers for these negated sentences are the absences from the world depicted by (x) of the corresponding positive states of affairs. Russell objected. The claim is: if A does not love B then what makes that true is the absence from the world of the state of affairs that A loves B. “But,” Russell objects, “the absence of a fact is itself a negative fact; it is the fact that there is not such a fact as A loving B. Thus, we cannot escape from negative facts in this way.”39 Absence just means that there is no such fact as A loving B in the world. This is not a matter of a proposition that does not correspond to anything, but rather the fact that A does not love B. On Russell’s view there must be a fact that grounds the truth or falsity of propositions. Either absence of a fact is to be taken as the same as a negative fact, or it is just a feature of language. If it is the latter, then, for Russell, it cannot be the ground for the truth of the proposition that ‘A does not love B’. Hochberg40 has also claimed that the absence of a positive fact entails the existence of a negative fact. Say that the world contains two things: a white square and a black square. To use his example, consider the

426 four positive sentences: W(a), B(b), S(a), S(b). From these sentences there is no way to infer ‘~W(b)’. But suppose this negative proposition is true. Its truth cannot be inferred from the first list, so the facts represented by the sentences on this list cannot constitute its truth-maker. It therefore requires a truth-maker of its own. If it is said that the absence of the positive state of affairs represented by ‘W(b)’, then the argument, completely cogent, is that this absence of a positive facts amounts ontologically to the presence of a negative fact.41 There is an asymmetry between positive propositions and negative propositions if one claims that the fact that makes ‘B(b)’ true is also the same fact that makes ‘~W(b)’ true since this is different than the sense in which one fact makes ‘W(a)’ true and ‘~W(a)’ false.42 Hochberg claims, as Russell does, that the pair “positive fact and negative fact” accomplishes the same thing that “presence and absence” would do. To take up the point using our example, this argument gets its point from the apparent inability to infer the sentences of (p) from the sentences of (x). Therefore the facts (x) cannot be the truth-maker or makers for the negative sentences of (p). But these negative sentences are true – and are true, not as somehow a matter of convention, but objectively, true by virtue of something about the world and the way it is. Since it can’t be the facts (x) that are the truth-makers, there must be (so the argument goes) other facts about the world that make the negative sentences true. These further facts are the absences of certain facts from the world, or, what is said to be the same, the negative facts of the world. But the argument assumes a false premise, that the sentences of (p) cannot be inferred from the set of sentences of (x). For, given the description of our world Fa, Gb one can infer that the strings (q) Fb, Ga are absent from this list, and the absence of these from (x) is what makes the negative propositions of (p) true. The absence of a sentence from the list amounts to admitting a negative fact only if the positive facts of (x) cannot be the truth-makers of the negative facts of (p). But if one can infer that the strings which are absent from (x) are precisely those of (q), then one has the positive facts of (x) serving as truth-makers for the negative sentences of (p). The inference depends on two facts about the world. There is, first, the determination that the strings of (q) are those that are absent; this determination is made by the formation rule (*). There is, sec-

427 ond, the further determination that, as they are not true, then they are false. This determination, that they – the sentences of (q) – are false is made by the inference: since these are absent from (x) and since they are therefore not part of the description of the way the world is, then they are false – and this inference is justified by the fact that the world is two-valued. Hence: the argument for negative facts as truth-makers for negative sentences depends upon the assumption that we cannot infer the truth of the negative propositions from the truth of the positive propositions. But we can so infer. So the argument that we need negative facts as truthmakers for negative sentences fails. Thus, we can ground objectively the truth of those negative propositions (strings, sentences) that describe our world. But contrary to Grossmann – and Russell, and Hochberg and Bergmann – that ground is not an entity. The ground of negativity is really there in the world, but it is not among the things in the world. In a sense – in a sense – , then, what grounds the truth of a negative statement is the absence of a fact or state of affairs; in a sense, we have accounted for the truth of negative propositions by having such propositions refer to nothing, or an absence of any fact whatsoever. Such a view, taken literally, has been proposed by Oaklander and Miracchi.43 They argue that negative propositions do not correspond to anything in the world. Their claim is that the sentence, ‘This is red does not exist’ can be altered to ‘It is false that this is red,’ and this statement of absence does not require a negative fact to ground the absence of this fact. Rather the absence “can be treated as a feature of the ‘language’ without ontological significance, since it implies nothing more than that a certain positive sentence does not correspond to anything.”44 On this view, one accounts for absence by claiming the absence is just a feature of the language. The perspicuous language of the ontologist will have two roles, that of representing facts, and that of transcribing “the logic of sentences contained in a natural language.”45 Language so used contains both true and false sentences, but the role of language is that of authorizing inferences among such sentences and others built out of them by connectives and quantifiers.46 But there is language as used to perspicuously represent the ontology of one’s world: in this use, language “does not contain both true and false sentences, but only true sentences.”47 But this by itself won’t do: it makes the truth of negative propositions a feature simply of the conventions of language, and does not square with our commonsense view of truth, namely, that the truth of (non-analytic) propositions is determined by non-linguistic features of the world. The view that we have suggested, fol-

428 lowing Morris Cohen and Everett Hall but also Grossmann, is that the logic of language reflects deep seated facts or truths about the world. More specifically, we have suggested that, while it is the case that negativity does as it were arise only by virtue of language, it is nonetheless not purely conventional: rather, the conventions reflect the two deep seated truths that facts in the world have the form (*) which is reflected in the formation rule for the language, and in the other deep seated truth that the world is twovalued. Negativity is thus grounded in the objective structure of the world. It is true that, as Oaklander and Miracchi state, that “the most crucial way in which sentences in [one’s perspicuous language] represent sentences in a natural language is that they are capable of being true or false,”48 but this capacity is not one that depends on language and its rules alone: language has this capacity because two-valuedness is part of the ontological structure of the world. The radical separation, proposed by Oaklander and Miracchi, of language as representing logical relations which have nothing to do with ontology and language as representing ontological structure and categories is not tenable. To be sure, if one accepts Russell’s theory of truth-makers, that every true sentence, positive or negative, requires a fact as its truth-maker, then one should reject this account of absence. But the admission of this account of truth-makers leads to the admission of negative facts as separate entities, and this is unacceptable. But why accept Russell’s (and Hochberg’s and Grossmann’s) account of truth-makers? Oaklander and Miracchi reject this account of truth-makers, but wrongly infer that the capacity of language to represent logical relations needs no ontological ground: negative sentences represent mere absences. However, these authors notwithstanding, one does need an objective ground, a ground in the facts of the world, for the truth of negative propositions. The point is that this ground need not be an entity, a thing in a category of its own, as Grossmann and these others claim. One can provide the required ontological ground without requiring negative facts: there are other ways, or so we have argued, to objectively ground negativity in the structure of the world.49

III Now recall the question we are addressing: Is it really necessary to admit negative facts? We are proposing a negative answer to this question: our ontology need not be as luxuriant in entities as that which Grossmann pro-

429 poses. We have developed one line of criticism that supports this negative answer – the ontological ground of negativity need not be an entity. But there is a second line of criticism that must also be developed. This line is not unrelated to the first. Consider again our object-language to be extended as we have suggested to include negation. The small world that we are supposing we perceive is described by the strings (x) Fa, Gb but the set of true sentences about this world will be (p) Fa, Gb, ~Fb, ~Ga (p) records all the possible states of affairs in the world, (x) records those that are actual. What we have argued is that in effect we infer (p) from (x). Sartre’s example fits this pattern. He perceives the absence of Pierre. But this perception of the absence is achieved only after he (Sartre) looks carefully about the room: he perceives something that is a possibility for that room but which is not actual in the room, but it arises only after what is actually there has been given to him. But what is important is that he does perceive the absence: the negativity may be grounded in the actual, but for all that it is there in what is given to one in perception. Grossmann agrees with Sartre on this point: we perceive negative states of affairs.50 In this he is no doubt correct. To make the point in terms of the simple example we have used: In our small possible world, we infer (p) from (x). We assume that we perceive the facts (x). But we not only infer (p) but besides that, as Sartre makes clear, we also perceive the facts (p): the negativity is given to us in perception. If this is so, if negativity is given to us in perception, if we perceive what is not as well as what is, – and it is so –, then, given the empiricist commitment to PA and the primacy of perception, or, what is the same, given the commitment to a phenomenological method, then how can one avoid admitting negative facts, and therefore negation, into one’s ontology. We have suggested that, contrary to Grossmann, we need not admit negation as a separate entity in our ontology, but it would seem that, given our shared commitment to the empiricist’s PA, we must after all agree with Grossmann’s view and admit negation into our ontology. But there is the worry about luxuriant ontological landscapes, and Russell’s urge to avoid negative facts. Given an empiricist commitment to the primacy of perception, how can one avoid admitting negative facts, and therefore negation, into one’s ontology? The answer that fits the empiricist tradition would seem to this, that

430 one distinguish what one perceives (perceives in the broad sense) from what one is acquainted with (perceives in the narrow sense). For our small world, what one is acquainted with is that Fa, Gb What one perceives is that Fa, Gb, ~Fb, ~Ga The surplus that is in the latter over the former is, we have argued, a matter of inference on the basis of laws. In this case, the laws in question are the formation rule and the two-valuedness of the world, categorial laws describing how (the entities of) the basic categories systematically related to each other.51 Now, if “I” is the intentional relation in which thoughts stand to their objects, that is, if I is the relation that relates thoughts to what they are about, and if t is the perceiving that places the small world before consciousness, then we have (a1) t I (Fa) but also (a2) t I (~Fb) (a1), it was just suggested, is acquaintance and (a2) is perception. Our empiricist Principle of Acquaintance has us admit into our ontology the entities in the state of affairs that is the content of the thought in the case of acquaintance (a1). But (a2) does not differ from (a1) in the way it intends its content. So if we admit into our ontology the entities in the state of affairs that is the content of (a1), then by parallel reasoning we ought to admit the entities which are in the state of affairs that is intended in (a2). But negation is among those entities. So we seem committed after all to admitting negation into our ontology. At least, we must do so if we hold, as Grossmann does (200), that intentionality is a simple unanalyzable relation. So we cannot avoid admitting negative facts so long as we take intentionality to be an unanalyable relation. This leads to the promised second line of criticism against Grossmann, which is that the “relation” of intentionality need not be taken to be unanalyzable. Now, there is no doubt that there are mental acts and that these are distinct from images.52 Moreover, having a certain propositional content is a property of such acts; as Grossmann says, there is a certain feature of mental acts such that they intend states of affairs. Call such a feature a thought. Then the thought that p intends the state of affairs p. But in addition, the thought that p is the feature in consciousness that is the causal ba-

431 sis of the disposition to utter the sentence ‘p’ that, by virtue of the worldword connections of language, refers to or represents the state of affairs p. The thought that p is thus linked to the state of affairs p through the causal regularity or regularities and dispositions that link the thought to the disposition to utter the sentence ‘p’ and regularities and dispositions that yield the referential meaning of the sentence; call this latter the causal reference of the thought. There is no doubt that thoughts in this sense causally refer to the states of affairs that they intend. The suggestion to be made is no doubt obvious: analyze the intentionality of the thought that p as the causal reference that p.53 A sentence refers to the state of affairs such that, if the latter exists, the sentence is true. One now distinguishes the semantic (including the syntactic) regularities that govern the descriptive terms of a sentence from the regularities that govern the logical terms (connectives, quantifiers). The latter determine the reference of a sentence in terms of the references of atomic sentences, while the semantic regularities determine the reference of atomic sentences. Finally, there are linguistic patterns that determine not only reference but assertibility. Ideally, one asserts a proposition if and only if it is true, and more specifically only if one has evidence that it is true. For the empiricist this evidence consists primarily, if one may speak loosely, of coming into contact with facts, and inferences drawn from such basic data. In the basic case of the atomic sentences in terms of which the truth conditions of all other sentences are specified, the disposition to assert such a proposition should be evoked by the fact that it refers to or is about. This fact will, of course, be an existing state of affairs since only existing states of affairs can be causally efficacious. One must, of course, allow for perceptual error, but the ideal, to which various situations more or less closely approximate in often well known ways, consists in errorfree observation. It is clear that one now has the required distinction that we mentioned above, between perception and acquaintance. One perceives whatever is or is in the intention of a perceptual act, but one is acquainted with only those (atomic) facts that directly evoke the disposition to assert the sentences that are about them. One can now argue that, or complete the argument that negation is not there in the world. In our little world we have the facts (x) Fa, Gb These facts exist. But there is nothing which exists which corresponds to the sentences

432 (n) ~Fb, ~Ga To be sure, the truth of these sentences is grounded in the facts – the existing facts – of the world, but there are no entities corresponding to these negative sentences; their truth is grounded in the facts of the world but unlike the non-negated or positive sentences (x), the negated sentences do not have entities which are their truth-makers. Nonetheless, though the sentences (n) lack truth-makers, one can still perceive these facts. For example, one can perceive that ~Ga – one can perceive the absence of Pierre. This perception involves the thought that ~Ga and therefore the disposition to assert the sentence ‘~Ga’. This disposition is present by virtue, first, of the fact that one is acquainted with both and only the facts Fa and Gb, that is, by virtue of the fact that these facts evoke the dispositions to assert the atomic sentences ‘Fa’ and ‘Gb’ and only these atomic sentences; and by virtue, second, of the semantic regularities that permit one to further assert on this basis the sentences ‘~Fb’ and ‘~Ga’. Thus, the perception of the negative facts proceeds on the basis of an inference from what one is acquainted with. This is brief indeed – one should go on to discuss error, the role of sensory contents, and much else – but it does indicate however sketchily the direction in which an analysis of intentionality other than that of Grossmann might proceed.54 In particular, it indicates a line of argument for holding that we do indeed perceive negation, and various other connectives, the quantifiers, numbers, and so on, one need not admit that they exist as entities but rather appear in the world through out thinking about it.55 And, to repeat and to emphasize, to argue that they are in this way not entities is not also to argue that the facts that they appear in the world as we perceive it has no ontological ground. IV If this alternative to Grossmann’s position is defensible, then the result is an ontology that is a desert indeed compared to that of Grossmann. Connectives, quantifiers, numbers, structures, all contrary to Grossmann, do not exist. At the same time, the basic thrust of his case remains: contrary to the standard view of, e.g. Nagel or (to mention another) Carnap, these things do have an ontological ground: the alternative just sketched does not dispute this, but only the claim that this ground consists of entities. From the viewpoint of this alternative, Grossmann’s error consists, first, in his insistence that every feature of an object of thought must be on-

433 tologically grounded in an entity, and, second, in his insistence that intentionality is not analyzable. And, as we earlier suggested, the two are connected. For, Grossmann’s failure to recognize that intentionality cannot be analyzed in terms of causal reference consists in his insistence that it be construed as an entity something like an ordinary relation. There is one point at which Grossmann raises an issue which challenges this verdict, and, more specifically, challenges the notion that intentionality can be analyzed in terms of causal relations. He suggests that if one proceeds, as our alternative suggests, to analyze intentionality causally in terms of regularities and dispositions, then one cannot account for how the belief that not-p is “attuned” to a non-existent state of affairs (193). But this is no problem: on every view, even Grossmann’s language and, more specifically, sentences are “attuned” to non-existent states of affairs. The problem is not whether such attunement can occur but whether intentionality can be analyzed in terms of such an attunement of sentences. Given Grossmann’s empiricist tendencies, it is precisely at a point such as this that Grossmann should appeal to experience to settle the question of the analyzability of intentionality. But nowhere in his discussion of intentionality (189-203) do we find Grossmann making such an appeal. Rather, Grossmann turns to dialectics, precisely the move that the empiricist should not make.56 Be that as it may, he argues that the reference of sentences is conventional and therefore capable of change by decision, whereas “we can no more change the fact that a certain belief in the fact that the earth is round than we can change the fact that two plus two is four” (202). This is by no means decisive, however. For, on the alternative sketched, while the fact that the marks ‘p’ referentially mean the state of affairs that p is a matter of convention, it is not a matter of convention that the thought that p is the causal basis in consciousness of a disposition to utter a sentence that refers to or represents that state of affairs that p. Thus, while the connection between a sentence and that state of affairs that it represents is conventional, it is not accidental that the thought that p should be precisely that entity that causes one to be disposed to utter a sentence the referential meaning of which is p. Grossmann also argues (203) that in order to establish the convention that, say, ‘red’ means or refers to (the property) red, one must “connect the one with the other in a certain indefinable fashion” (203);57 conventional meaning thus presupposes rather than accounts for the causal reference of sentences. But the argument is spurious. There are semantic regularities governing the introduction of new names; these are such that hitherto un-

434 observed objects can evoke not only the disposition to apply hitherto unused names to themselves, but (often) also the settled disposition to continue to use these names. This certainly means that when an object evokes the first use of a new name, the thought of that object is novel. But one need not conclude that such novel “acts of meaning or referring” (203) somehow create the conventions.58 It may be so, but to suppose it without further argument, as Grossmann does, is to stray dangerously close to the old idea of the romantics that all meaning in the world is the active creation of the self that transcends, but can causally shape, the world of ordinary experience.59 Thus, although there is more to the dialectic than this,60 I think it safe to say that Grossmann’s case for the unanalyzability is less than clear-cut. And moreover, the appeal to experience is missing precisely where one would expect to find it. In short, if Grossmann’s empiricism finds negation, the quantifiers, etc., to be simple entities in the world, it does so only because it takes intentionality to be a simple relation, and it does this more for reasons of ontological dogma, that the ontological ground of every feature of the world that we experience must be an entity, than for reasons of an empiricist sort.

I have raised a number of questions about the ontology that Grossman gives us and defends in this book on The Categorial Structure of the World. In raising these questions, however, I wish it to be understood that this book about which they are raised is significant and important. Today, there are those, like David Lewis, who do ontology through talk about “possible worlds” and other such chuzzerai; for the most part such ontologies, though superficially elaborated in scholastic detail, are so remarkably naive in both their articulation and their support as to be unworthy of detailed discussion. Grossmann’s work stands head and shoulders above this in its sophistication of detail and argument. We also have today those, like Sylvan (né Routley), who take apparently perverse delight in retailing the worst excesses of Meinong. Grossmann’s work is sobriety indeed compared to this, as it is compared to that of Meinong himself. In short, Grossmann’s achievement is solid and substantial; and unlike so much of what we now find in the literature, his is work that any empiricist and anyone interested in analytic metaphysics must take seriously. Highly recommended.

435 Endnotes to Study Eleven

1.Reinhardt Grossmann, The Categorial Structure of the World (Bloomington, IN: Indiana University Press, 1983). Unless otherwise noted, page references in parentheses in the text refer to this volume. 2.Cf. F. Wilson, The Logic and Methodology of Science in Early Modern Thought (Toronto: University of Toronto Press, 2000). 3.Cf. ibid. Also R. H. Popkin, The History of Scepticism from Savonarola to Bayle (Oxford: Oxford University Press, 2003), rev. and expanded edition of The History of Scepticism from Erasmus to Descartes (Assen: Van Gorcum 1963); and R. A. Watson, The Downfall of Cartesianism (The Hague: Nijhoff, 1966). 4.See Study One, above, “Acquaintance, Ontology and Knowledge,” and Study Fourteen, below, “Effability, Ontology and Method.” 5.On this method, compare the discussion below, in the study on “Effability, Ontology and Method.” 6.Conversely, neither are properties reducible to classes. For greater detail on this, see Grossmann’s Ontological Reduction (Bloomington IN: University of Indiana Press, 1973), esp. Part II. 7.See also Grossmann, Ontological Reduction, Part I. 8.See also Grossmann, Ontological Reduction, Part III; and his “Structures, Functions and Forms,” in M. Schirn, ed., Studien zu Frege, vol. 2 (Stuttgart-Consnstatt: Friedrich Fromman, 1976). 9.On the issue of modes of being, are there one or several?, see Grossmann’s Phenomenology and Existentialism: An Introduction (London: Routledge and Kegan Paul, 1984), Part III, “Martin Heidegger: ‘The Meaning of Being’.” 10.See also Grossmann, “Nonexistnet Objects versus Definite Descriptions,” Australasian Journal of Philosophy, 62 (1984), pp. 363-77; and also his Meinong (London: Routledge and Kegan Paul, 1974). 11.Experiencing includes sensing and perceiving in which we are aware of sensible physical objects, and also inner awareness, that is, the kind of mental act through which we are conscious of the contents of our own minds, or, more accurately, conscious of our own mental states. 12.See, e.g. R. Sylvan (né Routley), Exploring Meinong’s Jungle and Beyond (Can-

436 berra: Departmental Monograph # 3, Philosophy Department, Research School of Social Sciences, Australian National University, 1980). The exploration goes on for 1035 pages, plus plates and introductory material. 13.E. Nagel, “Logic without Ontology,” in Y. Kirkorian, ed., Naturalism and the Human Spirit (New York: Columbia University Press, 1944). Nagel’s view was subjected to some telling criticisms in a “Symposium on the Relation of Logic and Metaphysics,” (Philosophical Review 48 [1949]). There were contributions by E. J. Nelson (“The Relation of Logic to Metaphysics,” pp. 1-12), A. Ambrose (“Everett J. Nelson on ‘The Relation of Logic to Metaphysics’,” pp. 13-16), and Everett W. Hall (“The Metaphysics of Logic,” pp. 16-25), with a reply by Nagel (“In Defense of Logic without Metaphysics,” pp. 26-34). 14.M. R. Cohen, Reason and Nature, second edition (Glencoe, IL: The Free Press, 1953), p. 196. The first edition is 1931. 15.Bertrand Russell of course had earlier held this position; see his Problems of Philosophy (London: Williams & Norgate, 1912). The view was shared by G. Frege. But Canrap, though a student of Frege, was more influenced by Wittgenstein’s Tractatus, and came to propose that one could do logic without ontology. See his Logical Syntax of Language. Carnap’s view was adopted by most of the other positivists, and persuaded Nagel, too. Grossmann’s teacher, G. Bergmann, alone among the positivists, was not persuaded by Carnap’s conventionalism, and came to defend an ontologically realistic interpretation of logic, akin to the positions of Russell, Frege, and Cohen. See his “Ineffability, Ontology and Method,” in his Logic and Reality (Madison WI: University of Wisconsin Press, 1964). Also his Realism; a Critique of Brentano and Meinong (Madison WI: University of Wisconsin Press, 1967) and his New Foundations of Ontology (Madison WI: University of Wisconsin Press, 1992). Grossmann takes up and develops themes from Bergmann’s logical realism. 16.The argument is that the rules of logic are regulative, not descriptive, and that their function – their only function – is to regulate inquiry. Notice the weasel word ‘function.’ Hall makes the essential point. He grants that logical principles function to regulate inquiry. But, “It still would not be the case that a logical principle just is a way of regulating inquiry. A logical principle might well be a tool of inquiry much as an axe is a tool of woodcutting – by having a positive nature of its own that fits it for its function” (“The Metaphysics of Logic,” p. 18). Logic presupposes that the world has a certain categorial structure, where “Categories are features (I purposefully use a loose word) of the world revealed in the major syntactical forms necessary to a clarified language adequate to talk about the world” (p. 24). Grossmann has a similar view of categories, and the way they are reflected in a language that is ontologically perspicuous (an “ideal language”). 17.Notice Wittgenstein’s remark that “It is clear that something about the world must be indicated by the fact that certain combinations of symbols – whose essence involves

437 the possession of a determinate character – are tautologies” (Tractatus LogicoPhilosophicus [trans. D. F. Pears and B. F. McGuiness (London: Routledge and Kegan Paul, 1961)], 6.124). 18. See Jean-Paul Sartre, Being and Nothingness, trans. by Hazel Barnes (New York: Philosophical Library, 1956) “The Origin of Negation”, pp. 33-85, and especially pp. 40-42. 19.This has been suggested by such diverse philosophers as P. Strawson, Introduction to Logical Theory (London: Methuen, 1952), and B. Bosanquet, Logic, or, The Morphology of Knowledge, two vols., (Oxford: Oxford University Press, 1888). (There are more similarities than this between Oxford “ordinary laguage” philosophy and the philosophy of the British idealists.) See also Jay F. Rosenberg, “Russell on Negative Facts” [Noûs, 6 (1972), p. 27-40]. Rosenberg introduces the notion of relational families (such as “North” and “South”, etc.), the members of which mutually exclude one another; negativity enters due to the necessary exclusion by one member of the relational family by another member of the family. The idea that negation can be analyzed in terms of incompatibility is most clearly stated in R. Demos, “A Discussion of Certain Types of Negative Proposition, Mind, n.s. 26 (1917), pp. 188-96. Russell takes up Demos’ arguments in his “Lectures on the Philosophy of Logical Atomism,” (delivered in 1918) [in his Logic and Knowledge, ed. R. C. Marsh (London: Unwin, 1956)], p. 212ff; and in his “On Propositions: What They Are and How They Mean’ (also in Logic and Knowledge), p. 288ff. Grossmann follows Russell. 20.To say that two properties F and G are incompatible is to say that “for all x, x is F only if x is not G”, and note the ‘not’. The alternative is to take “incompatibility” as primitive and unanalyzable (as Demos argues might be the case, following Bosanquet – Strawson, apparently unaware of these earlier discussions, does not see that there is an issue here, in any case he does not discuss it), but that raises real problems of its own, as both Russell and Grossmann point out. 21.This has been suggested by Donald Brownstein [“Negative Exemplification”, American Philosophical Quarterly, 10 (1973), pp. 43-50]. He considers a spot b, and the quality white. There are two possible states of affairs into which these parts can combine. The spot b may have the quality white or may not. Brownstein introduces an analogy which he uses to defend the idea of negative exemplification. He supposes that one has two circles C1 and C2. These have the same diameter and can be arranged geometrically in two different fundamental ways. Either the circles are placed in perfect coincidence or they are placed so that at least one point on C1 does not coincide with C2. In either case there is some determinate spatial relation between these circles (Brownstein p. 47). Similarly, he suggests, just as is the case with circles in spatial relations, so for colours. The object b must stand in some determinate relation: either b is white or it is not white. Brownstein notes: Just as the difference between the circumstance in which

438 C1 coincides perfectly with C2 and the circumstances in which C1 and C2 do not coincide perfectly is a difference in the relations by which C1 and C2 are related, so it seems reasonable to suggest, the difference between b’s being white and b’s not being white is a difference in the “relations” or ties by which b and white are tied (ibid.). So b is related somehow to white. In fact, there are two ways b may be related to white – negatively or positively. The parallel to the case of circles suggests that b and white differ in the two cases by the relations in which they stand to each other. These relations would seem to be, on the one hand, the tie of exemplification, and, on the other hand, the tie of negative exemplification. If b can enter into one state of affairs by one pf these relations, exemplification, then it can also stand in the other relation to enter into the other state of affairs. But this is hardly an argument. What it states is that if there are two ties, exemplification and negative exemplification, then there is a parallel to the case of the circles. But why accept the antecedent of the conditional? Why suppose that there is such a tie as negative exemplification? We are given no reason, and in particular, no counter to Grossmann’s arguments. For further discussion of the possibility of negative exemplification, see also Herbert Hochberg, “Negation and Generality”, Noûs, 3 (1969), pp. 325-343. 22. “~Fa” will make negative exemplification a two-place relation, but “~Rab” will require a negation that is three-place, and so on. 23.Plato’s treatment of the categories of Being and Non-being suggests something like this. 24.The presence of the quality “~F”, e.g. non-red, and the absence of “F” (red) seem to be the exact same thing. We introduce a pair of qualities, namely two complements, all the time. One multiplies properties – there are “F” and “~F” – and has relations amongst these properties – “F” and “~F” mutually exclude one another. But then there will be the negation of the latter: “~~F”. Here we have necessary co-exemplification. But does one have co-exemplification of two properties or are the properties identical? – “~~F” seems should be the same as “F” – in which case one has the exemplification of a single property, but represented in language in two ways. There is moreover the problem of fragmenting negation – “~Rxy” seems a different sort of negativity than “~Fx”. 25.Russell, “On Propositions,” p. 287. In his lectures on the “Philosophy of Logical Atomism,” he reports that “When I was lecturing on this subject at Harvard, I argued that there were negative facts, and it nearly produced a riot: the class would not hear of there being negative facts” (Logic and Knowledge, p. 211). 26.Russell, “Philosophy of Logical Atomism,” p. 214. 27.Grossmann makes the same case for the ineliminability of negation in his Phe-

439 nomenology and Existentialism: An Introduction (London: Routledge and Kegan Paul, 1984), p. 233ff. Here he finds common ground with Sartre against the denial of negative facts to be found in Bergson. But the case Grossmann makes is essentially the same, identical in many of its details, to the case that is made in his Categorial Structure of the World. 28.Bergmann, “Ineffability, Ontology and Method.” 29.Thus, Wittgenstein, Tractatus: “… the notation for generality contains a prototype” ( 3.24). 30.Thus, Wittgenstein, Tractatus: “It is obvious that an imagined world, however different from the real one, must have something – a form – in common with it” (2.022). 31 For more on the Humean position, see the essay below on “Bergmann’s Hidden Aristotelianism.” 32.Cf. F. Wilson, “Hume’s Theory of Mental Activity,” in D. F. Norton et al. eds., McGill Hume Studies (San Diego CA: Austin Hill Press, 1979); and Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1997).. 33.Thus, Wittgenstein, Tractatus: “...a propositional sign cannot be contained in itself...” (3.332) 34.Thus, Wittgenstein, Tractatus: “One might say ... that only connexions that are subject to law are thinkable” (6.361). 35 The term ‘prototype’ is used by Wittgenstein in this context; cf. Tractatus: “… the generality-sign contains a prototype” (3.24). 36.Thus, Wittgenstein, Tractatus: “A proposition must restrict reality to two alternatives: yes or no” (4.023). And: “Propositions represent the existence and non-existence of states of affairs” (4.1). And: “If an elementary proposition is true, the state of affairs exists: if an elementary proposition is false, the state of affairs does not exist” (4.25). 37.As Grossmann puts it, “the truth-making relation [for logical complex sentences] cannot be arbitrary.” So, for example, the laws for conjunction cannot be arbitrary, and “the laws of implication cannot be arbitrary either” (354) 38.Compare the interesting discussion of E. W. Hall, “The Metaphysics of Logic,” Philosophical Review, 58 (1949), pp. 113-118. 39.Russell, “On Propositions,” p. 288. 40.Hochberg, “Negation and Generality.”

440

41.Hochberg, “Negation and Generality,” p. 326. 42.Herbert Hochberg, Thought, Fact and Reference, (Minneapolis: University of Minnesota Press, 1979), p. 329. 43.Nathan Oaklander and Silvano Miracchi, “Russell, Negative Facts, and Ontology,” Philosophy of Science, 47 (1980), pp. 434-455. 44.Oaklander and Miracchi, p. 453. 45.Ibid., p. 436. 46.One presumes that they have in mind something like Carnap’s notion of the logical. Carnap has said (in his Logical Syntax of Language [London: Routledge and Kegan Paul, 1937]) that logical truth is analytic. A proposition is analytic if it asserts an Lconsequence, that is, a consequence inferred from the null class of premises by means of L-transformation rules. Since L-transformation rules are simply a certain sort of syntactical rules, Carnap means, one presumes, that entailment is a purely syntactical relation. Carnap was later to allow “reduction sentences” and, more generally, “meaning postulates” as analytic truths, though they are not L-derivable from a null sent of premises. See his “Testability and Meaning” [Philosophy of Science, 3 (1936), pp. 419-471] and “Meaning Postulates” [“Meaning Postulates,” Philosophical Studies, 3 (1952), pp. 65-73]. Thus a sentence is a logical truth just in case it is an L-consequence of the null set of premises or a set of premises containing reduction sentences or meaning postulates (or both) but containing no other premises. 47.Oaklander and Miracchi, p. 438. 48.Ibid., p. 437. 49.What holds for negation holds for the three other one-place connectives that the truth tables define, all 16 two-place connectives that the truth tables define, all 64 three-place connectives, all 4

2 2 four place connectives, ... , all 22

n

n-place connectives, etc. Each of this infinity of further connectives exists on Grossmann’s view; but he chooses to ignore them. In the same way he ignores, besides some unary quantifiers, all the binary, ternary, ... , n-ary quantifiers. (For each n, there are n

2

( 2 2 −1)

n-ary quantificational modes.)

50.See R. Grossmann, Phenomenology and Existentialism: An Introduction (London: Routledge and Kegan Paul, 1984), Part IV (“Jean-Paul Sartre: The Search for Freedom”), p. 235.

441

51.These are, to repeat, laws – matters of fact, in a suitably broad sense of ‘matter of fact’. As Bergmann once insisted, there is nothing logical about logic. This is the point that Hall made against Nagel. Grossmann would agree. 52.Cf. G. Humphrey, Thinking: An Introduction to Its Experimental Psychology (New York: Wiley, 1963), Ch. I-IV. Also R. Grossmann, The Structure of Mind (Madison WI: University of Wisconsin Press, 1965); and the studies below, “Marras on Sellars on Thought and Language” and “Human Action and a Natural Science of Human Being.” 53.See also the studies below on “The Aboutness of Thought” and “Effability, Ontology and Method.” 54.See also the studies below, “The Aboutness of Thought” and “Effability, Ontology and Method.” W. Sellars has also explored this idea; see his “Notes on Intentionality” in his Philosophical Perspectives (Springfield, IL: Charles C. Thomas, 1967), pp. 308-21. But this exploration is from Sellars’ own perspective. Specifically, while full of insights, his position ends up in an intolerable position through his construing cognitive meaning in terms of word-word connections (L-[transformation-]rules and P-rules) rather than reference (word-word connections): see his “Meaning and Inference,” Mind, n.s. 62 (1963). Selllars’ proposal concerning cognitive meaning amounts to asserting the claim that all concepts are theory laden. I have discussed the latter claim and its disastrous consequences in another context; see F. Wilson, “Logical Necessity in Carnap’s Later Philosophy” (in A. Hausman and F. Wilson, Carnap and Goodman: Two Formalists [The Hague: Nijhoff, 1967]), Chapter III. On the same issue, see K. Kordig, The Justification of Scientific Change (Dordrecht, The Netherlands: D. Reidel, 177). In addition, while Sellars analyzes the meaning of sentences in terms of semantic regularities, he fails to recognize that his account of meaning cannot straight off be applied to mental acts (see below, “Marras on Sellars on Thought and Language”). To do the latter is in effect to ignore mental acts, through wrongly suggesting that thoughts (not: intentionality) can be analyzed into imagery and dispositions to have imagery; but this simply won’t do (cf. the work summarized in Humphrey, Thinking). An extended discussion of a variety of aspects of Sellars’ position can be found in H. Hochberg, Thought, Fact and Reference (Minneapolis MN: University of Minnesota Press, 1978). See also in this volume the studies “Acquaintance, Ontology and Knowledge,” above, and “Universals, Particulars, and Tropes: the Role of a Principle of Acquaintance in Ontology,” below. 55.Something like this is perhaps to be found in some thoughts of the later Russell, in his Inquiry into Meaning and Truth (New York: W. W. Norton, 1940). In this Inquiry, Russell argues that believing p is a motor impulse to accept the sentence ‘p’ and that

442 believing ‘~p’ is the inhibition of this motor impulse: “Negation expresses a state of mind in which certain impulses exist but are inhibited” (p. 266). This view merits greater attention than can be given here, and certainly more than it is generally accorded. But it clearly indicates that he abandoned his earlier view of the uneliminability of negation. We earlier noted Russell’s comment that “There is implanted in the human breast an almost unquenchable desire of finding some way to avoid the admission that negative facts are as ultimate as positive” (“On Propositions,” p. 287); he clearly found, to his own satisfaction, a way of quenching that desire. 56.I have tried to delineate the roles of dialectics and experience (acquaintance) in the study below, “Effability, Ontology and Method.” 57.For a similar point, see H. Hochberg, Thought, Fact and Reference, p’ 399ff. 58.Compare the discussion in the study below, “The Aboutness of Thought.” 59.One who holds, as I believe Grossmann does (see his The Structure of Mind), that the thesis of psychophysiological parallelism correctly describes the mind-body relationship, ought to agree with this point; see the study below, “Marras on Sellars on Thought and Language.” 60.Compare the valuable discussion in L. Addis, “Natural Signs,” Review of Metaphysics, 36 (1983), pp. 543-568.

Twelve Bergmann’s Hidden Aristotelianism*

The ontology of Gustav Bergmann contains properties and particulars and, composed of them, facts. It has facts in which a particular exemplifies a property, and facts in which several particulars exemplify a relation. These are the atomic facts. But it contains some other facts besides these. In particular, it contains general facts: in addition to the facts that this thing, call it a, is F, that a is G, and so on, and facts like this other thing b which is not G – besides individual facts like these, there is the general fact that for everything, if it is F then it is G. That is, besides the facts Fa, Fb, Fc, ... there is also the general fact Every individual is F or, in symbols, the general fact (x)(Fx)1 Bergmann is here disagreeing with Wittgenstein, who proposed in the Tractatus that the general statement, Every individual is F, is to be understood as an indefinitely long conjunction of statements of individual fact. In holding this position, Wittgenstein was disagreeing with Russell, who argued for general facts in his “Lectures on Logical Atomism.”2 Bergmann is on Russell’s side. Indeed, he repeats Russell’s argument in justifying his own view. And he offers a second argument of his own. Bergmann in an early essay on “Logical Positivism” made it that one of the defining characteristics of that philosophical position is that all who adopt that position “hold Humean views on causality and induction...”.3 Bergmann is among these positivists.4 What I propose to argue is that when Bergman introduces generality he introduces an entity that is contrary to that Humean position on causality, and moreover in fact violates the Principle of Acquaintance which, * Originally presented at the Conference on the Philosophy of Gustav Bergmann at the University of Aix-Marseilles in Aix-en-Provence; forthcoming in the Proceedings of that conference edited by Prof. G. Monnoyer, who organized the conference.

444 as he argues, provides the basic argument for the Humean position on causality. The inconsistency can be removed, I shall argue, by recognizing that we are in fact not acquainted with generality (nor with existence) and therefore abandoning generality as an existent.

The case that Bergmann makes for generality consists of two arguments. I propose to examine this case, both his arguments. But before getting on to that I want to trace a tradition in philosophy, one in which Bergmann’s views can in fact be situated, though I doubt that Bergmann would see things in quite this way. In any case, locating him in this tradition will lay the groundwork for my reasons for thinking his view on general facts is mistaken.

We have the following syllogism: Water when heated boils This water is being heated so, this is (will be) boiling In symbols this is: (x)(Fx e Gx) Fa so, Ga The pattern is familiar. We have all seen explanations and predictions of this form. We have the event Fa and it falls under the pattern recorded in the major premise. If the generality which is the major premise is true then, given that the event Fa occurs then we know that the further event Ga will occur. That is, given the generalization, and supposing it to be true, the one event is succeeded by the other event: if the generalization is true, then the two events are, we might say, connected – though we should not at this point anyway give this a specific ontological significance. It is the nature of this connection that we have to discuss. Is this a generalization that is true just in case all its instances are true, saying no more than that all those instances are true? Or is this generalization not only true but also a fact that is there in addition to its instances and irreducible to them? The early Bergmann answered the first of these questions affirmatively and the second negatively, but, accepting various arguments and feeling certain pressures, he eventually changed his mind and answered the first question negatively and the second affirmatively: generality, he came to insist, has ontological status – it is a subsistent which exists alongside other existents. It is this later view that we wish to examine.

445 Bergmann was, of course, not the first to hold that some at least among the general facts have a special ontological status. In fact, such a view appeared early in the history of philosophy. Here is some of this background.

We can in fact begin at no better place than at the beginning, with Socrates. The scene is set in the neatest of Plato’s dialogues, the Phaedo.5 Socrates is sitting in prison. According to the earlier dialogue, the Crito, his friends offered to arrange things so he could escape to Thebes. But Socrates refused: following the demands of justice he determined that it would be wrong to escape his native city and run off to Thebes. He is, instead, determined to remain in his cell and drink the hemlock. So, we have one event, Socrates sitting in his cell. There are two events that could follow that event, namely, running off to Thebes and remaining in Athens and drinking the hemlock. There is nothing in the first event, Socrates sitting there, that indicates or determines that it will be followed by one rather than the other. But in fact it is the drinking of the hemlock that follows. Why? Socrates introduces into the discussion the philosophy of Anaxagoras.6 Anaxagoras had a sort of naturalism, in terms of which he hoped to explain what happens in the world.7 The world begins as a mixture of everything with everything, every part contains within itself an infinity of parts. These parts are of many kinds, bright and dark, for example; whatever else one can say, it is clear that the parts have properties of sensible things: explanation proceeds in terms of these parts, and it is naturalistic in the sense that things in this world are all of sensible sorts.8 There are no kinds in Anaxagoras that transcend and lie outside the natural world but in terms of which we are expected to explain causally what happens in the world of ordinary experience.9 Of course, some of the parts, no doubt most of them, are too small for the eye to notice, but they are like the things we sense and are therefore parts of the world we sensibly experience. In answer to Parmenides, the world consists of a real diversity of parts each of which is in itself unchanging. What changes are their relations. Then, once set in motion these parts begin to accumulate to form the bodies as we know them, including our own bodies. In fact, everything contains infinitely many parts, and among this infinity are parts gathered together as bits bone, and an infinitely of such bits. There are, too, infinitely many groups which are bits of flesh. Moreover, every bit of flesh contains not just bits of flesh but also bits of bones and bits of everything else. After an initial starting rotation, the parts started to come together. The initial movement was caused by mind or noØs, but afterwards the processes of change seem to have been conceived as more or less

446 mechanical.10 But mind still has a role to play, though much more ordinary than the role it played in producing the initial push. Mind was itself a sort of substance. To be sure, it is a special sort of substance: it is fiery. But it is of a sort of which there is no question that we can become aware of it. It too is part of the natural world of things we can know in our ordinary experience of things. Now, ordinary things have parts of the initial mixture concentrated or gathered together in various special ways: ordinary things arise out of the original mixture through the motions of its parts. Originally at rest, the mixture is put into motion through a push from mind or noØs which gives rise to an initial rotary motion. Mind or noØs is concentrated before its initial push,11 but then divides itself among the individual things that began to emerge.12 After the initial push, the world emerges through changes that seem to have been more or less mechanical, their patterns determined by the natures of the bits and of the initial jolt, with mind playing only a little role, now organizing and directing individual things.13 But mind itself is, as we said, a natural thing.14 The person seems, on this account of Anaxagoras, to have been considered a body, consisting of skin and bones, structured and moved by mind, which organizes things so as they survive as wholes for as long as they can, but in all this persons, it seems, are not distinctively different from dogs, and perhaps not even from oysters. Socrates objects.15 If Anaxagoras were correct, then, like a dog seeking only bodily well-being and survival, he would have run off to Thebes. But he is not doing that, so Anaxagoras must be wrong. The difficulty is the role of mind or nous. Socrates has already established that mind or nous, the soul, moves the body. Thus far he agrees with Anaxagoras. But what is needed to explain Socrates’ sitting in jail being followed by his drinking the hemlock is the fact that he is moved by a vision of the just. This leads him into his alternative explanation. Earlier he has established two things. One is that things in the world of sensible experience, one’s body, in particular, is moved by the soul and that this active entity is simple, an indivisible pure activity, and therefore unlike the things of the world of sense which are divisible and separable into parts, as, e.g. Socrates’ sitting there in his cell is separable from the events that will succeed it. The second thing that he has established is the existence of the forms as perfect exemplars of the things we observe in sense. Having secured agreement that all equals in our ordinary experience are imperfectly equal, he then argues on the basis of the principle that in order to judge something to be not F one needs to have the concept of F – in order to know that something is not red one must know what it is to be red. So, to judge something to be imperfectly equal we must have the concept of perfect equality. The form of perfect equality must therefore exist, not in the world of ordinary experience, but in another world.

447 What holds for equality holds for other things, perfectly straight lines, perfect triangles, and, of course, perfect moral virtue or justice. He uses these metaphysical principles to give an explanation of the sequence of events that Anaxagoras is (it is claimed) unable to give. Socrates sitting in his cell is followed by his drinking the hemlock because the soul, his soul, is striving to imitate in his outward sensible appearances the form of perfect human justice, the ideal good that moves his soul in one direction and not another, the form the vision of which moves him to do what is right or just and not what the doggie would do.16 There is a problem:17 why does the soul of Socrates strive after the form of human justice rather than the form of doggie justice or oyster justice? Separation is to be explained in terms of a connection being effected by a soul imitating a form. This is the model. So the soul striving after human justice has to be explained by the soul striving after the super-form of justice – from which it is separate so its striving after that needs another form to strive after, that of the super-duper-form of justice – and so on without end. And since there is no end, nothing ever gets explained. This is the Third Man Argument of Aristotle. To solve the problem in a way that saves the basic form of explanation Aristotle ends the separation of the forms from the souls. Each soul, or, as Aristotle comes to speak, each substance has intrinsic and inseparable from itself a form or essence. Then patterns in the world of ordinary experience which are separate from one another – apparently separable, so far as our senses know – because they are not really separate: rather they are connected by the activity of the substance moving in a direction determined by its form or essence.18 The sequences of the separable events of sense, of ordinary experience, are explained through their being a tie between them such that one event necessitates the succeeding event by virtue of the activity of the soul in which those events occur where the observable pattern reflects the necessary structure or form or essence of the active substance. The sequences of ordinary events in the world of sense are explained in terms of entities outside this world, the sequence is really one in which its parts are not after all separable but connected by the striving of the active substance qua having a certain form that determines the direction of the striving. Moreover, the explanation is in terms of best: the form of human is the form of human justice, and the way things are is explained by the soul aiming at the best. What is is explained in terms of what ought to be.19 Contrast this to the world of Anaxagoras. In that world, there is no unity, only separable parts. There are sequences of events, and even regular patterns: after all, stuffs are stuffs and each kind behaves in its own ways and combines and separates in its own ways. So there is regularity and patterns. But these pat-

448 terns are contingent, separable and without any real or necessary connection to one another. Even mind or noØs, while of a unique sort of substance, and not mixed with the other stuffs, is not itself a unity; it too consists of parts. What is not a unity cannot unify and therefore, on the Socratic-Aristotelian model, cannot explain. Mind, moreover, is of this world, the world that exists in space and time, the world we know by sense. Mind is not something that is outside this world and certainly not something that grasps the eternal forms and therefore partakes of that eternal being. There is no explaining of what is in terms of the best, no teleology that determines the way things are in terms of striving for what ought to be. So we have ordinary events explained by being instances of patterns: All human is animal Socrates is a human Hence, Socrates is animal On this, everyone agrees, Anaxagoras, Socrates, Plato, and Aristotle, and everyone since then. If we ignore the initial shove, this is all there is to explanation on Anaxagoras’ scheme.20 But for the Platonist and the Aristotelian, this is not enough: there is not yet any real unity. For the Platonist and the Aristotelian, the major premise, recording the pattern All human is animal if there is to be an explanation, must do more than merely record a pattern observed in ordinary experience: it must of course represent that, but further must represent a real connection, something necessary and not merely contingent: it must represent a necessary structure that so binds the characters that H (= human) in its very being is inseparable from A (= animal) in its very being. H and A must be so intimately connected that neither the contrary, stating their incompatibility, No human is animal nor the contradictory Some human is not animal is conceivable. The connection of H and A must be a timeless structure that rules that these possibilities are impossibilities. As Aristotle worked these things out,21 the required structure is that of a species as part of a genus – the species human within the genus animal, with the connection between effected by the specific difference, in this case the specific difference of being rational. This necessary structure is displayed in what Aristotle referred to as a scientific syllogism:

449 All rational is animal All human is animal All human is animal This syllogism displays the logical and ontological structure of the species human. This form is given equivalently in the real definition of the species human: human is rational animal This real definition is a necessary truth, a timeless truth about the logical and ontological structures of the forms of substances. More generally, and putting it in terms of a more recent logic, we have characteristics F and G, and the observable pattern All F are G or (x)(Fx e Gx) We have this connecting the two events Fa and Ga: (x)(Fx e Gx) Fa so, Ga According to the experience of sense, being F and being G are separable events. The pattern represented by “All F are G” is merely a regularity, a pattern amongst events that is merely contingent. There is no deeper or stronger connection, and therefore, on the Socratic-Aristotelian model, no genuine connection. In order that there be a genuine explanation, there must be a necessary connection. This in effect Socrates’ argument against Anaxagoras, or part of that argument anyway. The very same argument has been put more recently by Fred Dretske. In his essay on “Laws of Nature,”22 he argues that mere regularities are simply unable to explain anything. To say that a law is a universal truth having explanatory power is like saying that a chair is a breath of air used to seat people. You cannot make a silk purse out of a sow’s ear, not even a very good sow’s ear; and you cannot make a generalization, not even a purely universal generalization, explain its instances. The fact that every F is G fails to explain why any F is G, and it fails to explain it, not because its explanatory efforts are too feeble to have attracted our attention, but because the explanatory attempt is never even made. The fact that all men are mortal does not explain why you and I are mortal; it says (in the sense of implies) that we are mortal, but it does not suggest why this might be so .... Subsuming an instance under a universal generalization has exactly the same explanatory power as deriving Q from P & Q. None.23

The important point is that laws, that is, generalizations that genuinely explain,

450 “tell us what (in some sense) must happen, not merely what has and will happen (given certain initial conditions).”24 Mere matter of fact regularities do not explain: this claim is the substance Dretske’s argument (as it was the substance of Socrates’ argument against Anaxagoras). Statements of such generalizations amount logically to nothing more than conjunctions. And conjunctions do not explain their conjuncts. If a generalization is to explain, then it must show how the events in question were necessitated, not merely record their occurrence. In some sense, causal laws must be necessary; they must be such that an event that violates them is impossible, they must be such that their contraries must be inconceivable. This argument is of a piece with that of Socrates against Anaxagoras. Thus, the regularity represented by “All F are G” will not by itself be explanatory unless it is the reflection in sense of a necessary connection among forms which are the ontological reality behind the world of sensible appearances. That is, we must have a relation among forms (^) R(F, G) where R is a second order relation amongst first order characteristics F and G. Since this is a relation amongst forms, it is a relation which holds amongst them as outside the world of sense and of space and time. It holds not merely everywhere and everywhen, not merely omnipresently and omnitemporally but eternally, that is, outside the world of space and time, existing not in this world we know by sense, but existing as beings in the calm of a timeless and unchanging order. Since the order is timeless and therefore changeless, it is necessary. This necessity secures as necessary the connection, only apparent to sense, amongst the characteristics F and G , so that the pattern represented by “All F are G”is not merely a contingent regularity but a connection which is necessary and therefore genuinely explanatory. And, since the fact that R(F, G) holds timelessly, it can be argued that it is known a priori, as Plato and Plato’s Socrates argued, or at least in a nonperceptual intuition, as Aristotle argued. However, there are problems. It is worth noting, first, that, while there was a certain neatness to the necessary connection in the real definitions of the Platonic-Aristotelian metaphysics, once one moves away from this traditional ontology, things cease to be so neat. Causal regularities come to take on more complicated forms. The secondorder relation that constitutes the tie that secures causal relatedness should be seen for what it is, a somewhat odd relation. We not only have the above twoplace instance of it, but also three- and many-place instances of it. Thus, we could also have the regularity

451 (x)[(Fx & Gx) e Hx] If this is to have the necessity of a genuine causal collection, if it is to be a law and not a merely contingent regularity, then we will need the second-order relational statement R(F, G; H) where this will have to be distinguished from R’(F; G, H) which guarantees that (x)[Fx e (Gx & Hx)] is a law, if it be a law. We also have laws like For every shark there is a pilot fish close by which would go into the symbolism as (x)[Sx e (›y)(Py & Cxy)] which, one supposes, is guaranteed by some second order fact like R”(S; P; C-,-) or perhaps something like R”’[S, (›y)(Py & C-, y)] or whatever – it is in fact quite difficult to say exactly what it would look like. But one gets the idea. The point here is that the necessary tie among properties or characteristics of things is intended to secure as causally necessary what otherwise would be contingent regularities among separable facts. Dretske’s causal connection is not just one connection but a whole family of relations. We are owed a story about how the members of this family are to be understood ontologically. Dretske does not give us that story. Be this last as it may, the Dretske suggestion for securing the metaphysical or ontological necessity those regularities that are to count as causal has of late been taken up by several other metaphysicians, among them D. M. Armstrong,25 M. Tooley,26 and J. Brown.27 Unfortunately, there is a fatal flaw in the position that these philosophers have advocated. The crucial point is that in order for the fact that R(F, G) to guarantee the truth of the regularity, it must also be true, and not merely true but necessarily true, that (@) R(F, G) e (x)(Fx e Gx) The difficulty is that in terms of the standard notion of entailment given by the usual rules of formal logic, the sentence (@) is not a necessary truth: the logical form of (@) is not that of a logical truth, since the antecedent is an atomic statement (of the second order) and the formal relations of standard logic imply that no atomic statement ever entails a general truth. But (@) is precisely the sort of statement that must be necessary if structural relations among properties

452 are to provide a metaphysical ground for regularities among the individuals that exemplify those properties. Moreover, the relational statement (^) R(F, G) so far as formal logic is concerned is an atomic statement and, again, therefore not a necessary truth: the logical form of (^) is that of an atomic statement and our standard logic implies that all such statements are contingent, not necessary. The problem is that (^) and (@) are not what they are required to be: necessary. Their problem is to show the necessity of the relational fact (^) and the principle of inference (@). As John Earman has put it, What remains to be worked out [on the view that causal relations are relations among universals] is the formal semantics of the entailment relation [that holds between the statement about universals and the matter of fact regularity]; whether this can be done consistently ... remains to be seen.28

Earman puts the matter rather too cautiously: there is no reason to suppose that it can be done, no reason to think that one can find, in the formal logic of the empiricist, grounds to make plausible the claim that a second order atomic statement about properties, “R(F, G)”, should entail a first order generalization, “(x)(Fx e Gx)”. Earman is making the point that was in effect made earlier by Russell in his Inquiry into Meaning and Truth.29 Considering the generality All humans are mortal30 he takes up what he calls, not inappropriately, the “Aristotelian” suggestion, that there is an a priori connection among the predicates: “We might be tempted to say that one of these predicates implies the other, and use this as an analysis of what is “expressed” by “all humans are mortal.” But, he goes on to argue, This Aristotelian interpretation ... overlooks the fact that the connection is not between the predicates as such, but only between the predicates as predicated of one subject. “A is human”31 involves “A is mortal,” but not “B is mortal.”32

The objection was clearly and explicitly developed by Bergmann, in his essay “On Non-Perceptual Intuition,”33 and it has been developed by others also.34 It turns out that Dretske’s proposal is hardly original. Bergmann gave the criticism originally in response to C. D. Broad, who had argued Dretske’s thesis in his commentary on McTaggart,35 who had also defended a version of Dretske’s thesis. Unfortunately, Dretske seems not to know either the history of his proposal nor the criticisms which have been raised against it.36

453 The tradition of Plato and Aristotle had in fact a way out of this problem. That was the active soul. Recall from the logic text tradition that Aristotle argued with regard to syllogisms that the first figure M is P S is M S is P was the perfect figure, which could justify syllogisms in the other figures by deriving or “reducing” those of the other figures from those of the first figure, but the syllogisms of which could not itself be so justified by reduction. However, Aristotle also argued, the syllogisms which occur in this first figure could be justified by what came to be called the dictum de omni et nullo. This is the maxim that “whatever can be affirmed (or denied) of a class can be affirmed (or denied) of everything included in the class.” In the traditional metaphysics this did indeed have a place. According to that tradition, the proposition (a) S is P is about substantial forms. But it is true about individuals that (b) All S are P in the sense that (b’) All individuals which are S are also P If no substantial forms exist, as Anaxagoras would have it, this latter is of course the only sense that could be given to the asserted generality: it merely records a pattern amongst things. The problem facing McTaggart, Broad, Dretske, et al., is to account for why (a) implies (b) = (b’). The traditional ontology deriving from Aristotle and from Socrates’ critique of Anaxagoras has a solution to this problem. On this ontology, the forms as the forms of active entities, souls or substances, cause the individuals of which they are the forms to be in a way that guarantees the truth of (b) = (b’): what is predicable of the universal or form is predicable of the individual substances subordinate to it. S is P This is S so, This is P The metaphysics of active substances thus guarantees that (b) = (b’) is true if (a) is true. But furthermore, this ontology of active substances ensures that (b) = (b’) is not just true but necessarily true. For, the active substance, the This, has as its form the species S, and has that form necessarily. So, of its own ontological necessity, since it is S, it of necessity acts to make itself P: it brings it about that it conforms in its outward appearances to the necessary structure of its inner form or essence. And this will be true of every thing, that is, every sub-

454 stance which has that same form or essence. The form thus guarantees that every object with that form will have the same pattern in its outwards appearances. (b) = (b’) will therefore not only be true but be necessarily true. And, given the structure of the metaphysics, since (a) is about forms and timelessly true, and therefore knowable a priori, so also the truth of (b) = (b’) is knowable a priori. On the substance metaphysics deriving form Socrates’ encounter with Anaxagoras, Earman’s problem is solved, and Bergmann’s objection is met. Thus, on the traditional view, the dictum de omni et nullo expresses a fundamental principle or law about the ontological structure of the universe; as John Stuart Mill was later to put it, the principle stated that “the entire nature and properties of the substantia secunda formed part of the nature and properties of each of the individual substances called by the same name...”37 Now, as Broad presents the suggestion that Dretske later takes up, he quite rightly notes, appropriately as a commentator, that the proposal is present already in McTaggart, who had argued that observed instances of things being both F and G, no matter how numerous, have no tendency to support the conclusion that ‘All instances whatever of F are instances of G’.38 Hence, Unless some further premise, which is known a priori, be added to the empirical premise, the latter can never give us any ground to suspect the presence of the relation of conveyance [McTaggart’s term for the relation of causal necessitation].39

But at the same time, McTaggart also holds, as Broad puts it, that In the case of laws of nature, human beings cannot see by direct inspection that the relation of conveyance holds between the terms.40

This problem is not even acknowledged by Dretske, but it is significant. For, this problem, which Broad calls epistemological,41 implies that this account of causation that is being proposed, as a second-order relation between properties, even if we waive Bergmann’s objection, is of no use in solving the problem of induction. Hume was the first to make the point clearly. He argued that there is no necessity to causation: Anaxagoras was right, all that is involved in explaining the processes of this world, this world in which we carry out our ordinary lives, and which we know by means of our senses (and by means of our inner awareness), is the regularities that obtain among the ontologically separable events that we find in this world and have imposed upon us. If we examine carefully our knowledge of this world we find neither within it nor within ourselves as part of this world, any entities that could transform these regularities into objec-

455 tive necessary connections. Of course, these patterns are regularities, generalities which are timeless patterns, but nothing about them or the entities in the world that makes it a matter of ontological necessity that the world unfolds according to these patterns rather than those. Hume proposes,42 or, rather, argues that objectively causation simply is regularity: to ascribe a causal relation is to propose a general pattern describing the way things actually are. The justification for this claim is based, first, on the failure of those who defend objective necessities to direct our attention to these objective necessities. After all, we make causal judgments all the time – my flicking the switch turns on the light, the bark of the tree is rough to the touch, always rough, my willing my arm to rise is inevitably (though not necessarily) followed by my arm going up (most of the time, at least)43 – and since making causal judgments occurs all the time, if necessary connections were required for us to make such a judgment, then we should constantly be aware of them, they should be a constant part of or experience of the world and of our being in it. But they aren’t, and so the reasonable person would exclude them from his or her ontology. As Hume puts it, The small success, which has been met with in all the attempts to fix this power, has at last obliged philosophers to conclude, that the ultimate force and efficacy of nature is unknown to us, and that ’tis in vain we search for it in all the known qualities of matter.44

And Hume argues, second, that as we are acquainted with them in ordinary experience the properties of things are presented as ontologically and therefore logically self-contained: any characteristic like F, say human or say red, is given to us such that they do not stand in objective necessary connections to any other characteristic, say mortal or say not green. And so we have Hume’s definition of “cause” as nothing more than objective regularity among entities that ontologically are separable: “an object precedent and contiguous to another, and where all the objects resembling the former are placed in a like relation of priority and contiguity to those objects, that resemble the latter...”.45 Wittgenstein made the same point in his usual oracular way.46 “The world is everything that is the case.” (Tractatus, ¶ 1) These facts are represented by elementary propositions and “One elementary proposition cannot be deduced from another” (Tract., ¶ 5.134); he then comments to emphasize the point, “There is no possible way of making an inference from the existence of one situation to the existence of another, entirely different situation” (Trac., ¶ 5.135). The facts of this world as it is given to us in experience, and the objects, particulars and characteristics, are logically and ontologically self-contained:

456 one does not imply any other, each is what it is and neither is another entity nor is tied in its being to another entity. Hume made the relevant point: as Wittgenstein puts the point, with respect to inferring one basic fact from another, “There is no causal nexus to justify such an inference” (Trac., ¶ 5.136). And again, to emphasize and make the point more dramatically, he goes on to state that We cannot infer the events of the future from those of the present. Belief in the causal nexus is superstition. (Trac., ¶ 5.1361)47

Bergmann was later to propose that “as long as one sticks to cautious generalities” then it is safe to say that “all Logical Positivists,” including of course himself, among other things “agree that they ... hold Humean views on causality and induction....”48 This is based, like Wittgenstein’s similar view, and like Hume’s view, on the logical and ontological independence of basic facts, that is, as he (Bergmann) also puts it, upon an “insistence on the selfcontainedness of the given...”.49 This latter means that factual generalities are all contingent, including such statements as those of colour incompatibilities: Whatever is red is not green or in symbols (x)(Rx e ~Gx)50 To know an entity, a particular, a characteristic or a fact, we do not need to know any other entity. To know what redness is it is not necessary to know greenness, and to know that something is red it is not necessary to know that it is not green. As Bergmann also put the point “There are several things with which we become acquainted if they are once presented to us. If one such thing is presented to us again, we recognize it.”51 This being true, the world as we know it in ordinary experience is “atomistic,”52 the parts as we know them in their givenness are logically and ontologically separable: there are, once again, no objective necessary connections. McTaggart attempts to get around this point. He argues that he can establish a priori that there are such connections among properties: “...it is possible to know à priori that some such laws are valid if anything does exist.”53 But, as Broad indicates, we cannot know specifically which relations hold in the actual universe.54 For, we are not acquainted with any such relation: it is not given to us in our experience of the sensible world, nor is the activity or soul that makes the connection between the world of forms and the world ordinary experience given to us in our ordinary experience. In that experience all we do and all we can record is mere regularity. But if this is so, that is, if we assume as given the epistemological problem raised by Hume, then there is an insurmountable diffi-

457 culty with McTaggart’s claim upon which one must insist. The point is that we must know the sort of thing that is claimed to be there if that claim is to be intelligible. If we claim that there are unicorns, then we must know what sort of thing a unicorn is if that claim is to be intelligible. If we claim that there are boojums, then we must know what sort of thing a boojum is if that claim is to be intelligible. We know the sort dog because we are acquainted in ordinary experience with Fido, Spot, etc., that is, instances of the sort dog. We know the sort unicorn not because we are acquainted with instances of that sort – after all, there are no unicorns – , but because we can define the concept “unicorn” in terms of other sorts – horse and horn – which are such that we have been acquainted with instances of those sorts in ordinary experience. But we do not know the sort boojum because neither can we define it nor are we acquainted with instances of that sort. Boojums are snarks, and we have never seen a snark either, so we don’t even know the sort of this sort. These words, ‘boojum’ and ‘snark’ are just sounds.55 So the claim that there are boojums is unintelligible where the claims that there are dogs and that there are unicorns are both intelligible. Similarly, we can claim that there are objective necessary connections of the sort defended by McTaggart, etc., only if either we are acquainted with instances of the kind or we can define them in terms of kinds instances of which we are acquainted with. But such relations are held to be unanalysable. And then there is the epistemological problem: we have not encountered any of these entities in ordinary experience. But if such relations are unanalyzable and we are not acquainted with them in experience, then the concept of such a relation is like the concept of a boojum: unintelligible. This is a problem for the recent advocates of this position such as Dretske, Tooley, Brown or Armstrong, and also a problem for its earlier advocates, Plato and Aristotle. To be sure, the recent advocates do make a reply of sorts: they suggest that we can treat the required relations among universals as “theoretical entities.” Thus, Brown introduces the objective causal or nomological relation as a relation among forms as a “theoretical entity posited for theoretical reasons.”56 Armstrong makes much the same point: “The postulation of a connection between universals can provide an explanation of an observed regularity in a way that postulating a Humean uniformity cannot.”57 However, first off, one does not “postulate” a Humean uniformity: such a uniformity is simply an observed regularity. The use of such language is nothing other than a cheap attempt to make it seem that the two positions are somehow parallel, in that they both “postulate” something – but affirming an observed regularity is very different from “postulating” an unobserved tie of ontological necessity. Moreover, “postulating” a tie as a “theoretical entity” is also unlike the “postulation” of the

458 “theoretical entities” in physics, the affirmation, usually on good inductive grounds, of entities, atoms or quarks or whatnot, that are unobservable by sense.58 The “postulation” of a causal nexus or a tie of ontological necessity is totally different from the postulation of small parts that we cannot see – for example, the postulation of a speck of dust to explain why my watch won’t work.59 In this latter sort of postulation, the entities postulated share certain features with the entities of which we are aware in ordinary experience. Although we may not experience the entities themselves, and, indeed, even though they may be too small or too far away for us ever to experience them, nonetheless, they are of kinds with which we are acquainted.60 But this does not hold for the nomological connection: by hypothesis this is something totally different in kind from anything with which we are acquainted. This raises problems, surely, for anyone – problems that should not be brushed under the carpet by a blithe and unexamined use of the term ‘postulation’.61 Broad in fact attempts to avoid this problem. He argues that we are in fact aware by a sort of “non-perceptual intuition” of instances of the crucial relation, “conveyance” as he calls it, following McTaggart. Broad indicates62 that we have in the example of everything that is shaped has extension: “we recognise by mere inspection and reflexion that nothing could possibly have shape and lack extension”.63 But this will not do. It may well be that the proposition that everything that has shape is extended is in some sense of the phrase a synthetic a priori truth.64 But Broad has not solved the problem of giving us an instance of the relation of conveyance. For, if conveyance is a relation among properties that guarantees the truth of the generalization in question, then it must entail that regularity. That is, we must have (@) as a necessary truth, as Bergmann argued. And Broad does not provide us with an account of entailment that will do the job. So Broad has after all not provided us with an instance of the sort of relation that he and McTaggart need in order give conceptual content to the claim that there are second order relations among properties constituting ties of causal necessity. Actually, Broad elsewhere gives another example that he thinks is a case where we experience a tie of causal necessity. This is the case of the will. “It is perfectly plain,” he tell us, “that, in the case of volition and voluntary movement, there is a connection between the cause and the effect which is not present in other cases of causation and which does make it plausible to hold that in this one case the nature of the effect can be foreseen by merely reflecting on the nature of the cause.”65 Hume argued that this case is in fact no different from other cases of causation: cause and effect are logically and ontologically sepa-

459 rable.66 But, Broad goes on, “The peculiarity of a volition as a cause-factor is that it involves as an essential part of it the idea of the effect.”67 But surely Broad is confusing the causal connection with the intentional connection. In the case of volition, the volition which is the cause contains an idea which has as its intention the event which is the effect. It does not follow, as Broad suggests it does, that cause and effect are not logically and ontologically separable: it is easy to conceive the cause occurring but not the effect. So I think Hume is correct here also, and Broad is wrong, we no more experience a causal tie in the case of volition than we do in the case of billiard balls. It is no doubt true that in willing the effect I do not predict the effect. But there is nothing in the idea that causation is regularity that requires that in willing my arm to go up and its so going up, I must think of the regularity that must – and does – obtain with respect to volitions of that sort and movements of that sort if the former are to cause the latter. So again Broad has not provided us with an instance of the sort of relation that he needs in order give conceptual content to the claim that there are second order relations among properties constituting ties of causal necessity. This is true in general. The defenders of the thesis that laws are relations among universals provide no clear account of entailment such that a statement of such a relationship entails a generalization about the particulars that exemplify those universals. Thus, though Dretske, Armstrong, Brown and Tooley, following Plato and Aristotle, all assert that the connection (@) is an entailment relation they make no effort to indicate the logical form that provides the ground of the necessitation. The antecedent of (@) is a relation among properties, and therefore in some sense a necessary truth, at least in the sense of being timeless. However, on any standard account of entailment, there is no reason to expect a non-general relational fact about properties to entail a generalization about the particulars which have those properties. Whatever else a law is, it is a regularity. The concern is whether there is anything more to laws. Dretske et al., insouciantly following Broad, argue that there is something more, to wit, the relation among universals. But if that is to make sense, then the relation among universals does indeed have to entail the matter-of-fact regularity, as they quite rightly see. But, alas! they provide no account of that entailment relationship. To assert that one exists is hardly to explain it. What precisely is the logical form that is the ground of the necessity alleged to hold between the antecedent and the consequent, between the relational fact about universals and the generalization about the particulars falling under those universals? In the absence of any reasonable account of logical form that would show the necessity of (@), the notion that we can construe laws as relations among universals is simply a non-starter. Plato and Aristotle of course do provide an answer

460 where Dretske et al. do not: they do ensure the necessary truth of (@). They do this by appeal to the activity of a soul or substance to guarantee that the connection among the forms is exemplified in their outward appearance as observable things. This is not an entailment relation, but it is an ontological guarantee. However, just as the Humean and Wittgenstein and Bergmann reject objective necessary connections, so they also reject the active soul or substance: we are acquainted with neither, and therefore neither can be admitted into our ontology. Indeed, Dretske et al. also reject this Platonic or Aristotelian solution of an active substance – at the least, they invoke no such entity – but they thereby deprive themselves of a solution to the problem of guaranteeing the necessity of the observed regularities among ordinary things.

Here is where we are. What the defender of objective necessary connections needs is something like this. We have F caused G (or: F causes G) which can be construed as This event, which is F-like, causes this other event, which is G-like or more briefly as (This being F-like) causes (this being G-like) or still more briefly as (c) Fa causes Ga We can see here the logical features that a causal nexus or necessary tie must have.68 First, the tie causes connects two facts. It is a connective of sorts, connecting atomic sentences into complex sentences. It must effect a connection between the individual facts that are the cause and the effect. It must therefore effect a connection between the individuals in the pair of facts it connects, it must connect them through a connection among those causally relevant properties which the individuals exemplify. Second, as (c) stands, the connective causes is not a truth-functional connective: the truth-values of the two atomic sentences in (c) do not by themselves determine the truth-value of the complex sentence. The truth-value of the complex sentence will depend on other things, too, e.g. F and G cannot be any old kinds, they must be kinds that are causally connected. Any connection of the atomic facts that purports to be the causal tie must depend on the kinds that are said to be the causally relevant kinds. This is in effect the third point that is to be emphasized: the statement (c)

461 will be true only if the atomic sentences mention kinds that are causally connected. The kinds must be what are often called “natural kinds,” that is, kinds that are in themselves so bound together that they determine that a fact of the one kind necessitates causally the existence of a fact of the other kind. They must as kinds themselves be bound together in the causal nexus. In this respect they must in effect be Aristotelian forms or essences. [This third point requires that there be a connection among properties of the sort described by (^): R(F, G) The second point requires that the kinds F and G are “natural kinds” in the sense that when (c) obtains then, necessarily, the regularity (r) (x)(Fx e Gx) also obtains. In effect, it requires that the connective causes requires that the implication (@) R(f, g) e (x)(Fx e Gx) hold as a necessary truth. These then ensure that the first point is fulfilled.] Fourth, the causal tie must vary from context to context. Sometimes the “natural kinds” will determine a regularity of one logical form, sometimes a regularity of another logical form, sometimes involving relations at other times not. These are the four crucial features of any connective that can do what the causal nexus or tie is supposed to do, according to those who propose it. The proposals of Dretske et al. all fail to satisfy these crucial features. There are other issues. The proposed nomological connective causes is on the face of it non-truth-functional. If we make it truth-functional by making it a second order relation among characteristics as in (^) R(F, G), thereby making it an atomic fact among atomic facts, separable from and logically unconnected to other atomic facts, then it does not do the required job; this was the point made by Bergmann against Broad (and McTaggart), the point which was later taken up by Earman. But if it is non-truth-functional, then it comes into conflict with the standard explication of the analytic-synthetic distinction which presupposes that the only primitive connectives are truthfunctional. Call this the first epistemological problem. Further, if the causal tie is taken as primitive then there is the problem of how we know when it obtains and, more deeply, how we even know what it is. This is the difficulty Hume raised. It was referred to by Broad as the “epistemological problem” facing one who proposed anything like McTaggart’s nexus of “conveyance.” Let us refer to it as the second epistemological problem.

462 This second problem is essentially an appeal to the effect that such an entity cannot be admitted into one’s ontology on the grounds that it violates the empiricist’s PA. In effect so is the first the same appeal. For, the usual explication of the analytic-synthetic distinction presupposes the logical independence of all atomic facts. But that can be secured only if the facts are ontologically independent. But properties and particulars and relations are presented as ontologically self-contained – they are not in their being connected to other entities. PA therefore secures the ontological independence of the entities in the world we are given in sense and in inner awareness; and hence in turn it secures the logical independence of atomic facts that the usual explication of the analyticsynthetic distinction depends upon. With all this in mind, we can now turn to Bergmann’s general facts, the general facts that exist over and above particular atomic facts. We have Fa & Ga Fb & Gb Fc & Gc ........ and so on as a list of the facts in the world, but besides these we have to attach to this list as an additional fact the general fact (d) (x)(Fx e Gx), as somehow a further fact not merely a fact about but one that is within the world. This fact is to be counted as an additional fact when listing the facts that obtain in the world. This unanalyzable general fact that Bergmann includes among the facts obtaining within the world, I am going to suggest, has the four crucial features needed by a “connective” if it is to count as giving us the tie or nexus of causal necessity. It is even possible to give it a form that simulates the form (^) ‘R(F, G)’ that is proposed by Dretske & Co. to account for causal necessity. Let us define ‘C(f, g)’ is short for ‘(x)(fx e gx)’ where ‘f’ and ‘g’ are free variables. Then, given (d), we have (dd) C(F, G) as a (defined) relation among the characteristics F and G understood as universals.69 It of course immediately follows that (de) C(F, G) e (x)(Fx e Gx) is analytically true, true ex vi terminorum, thus meeting Earman’s point that for a causal relation among abstract forms the principle (@) must be a necessary

463 truth. Now look at the four crucial features which anything that purports to play the role of a causal connective or tie as in (c) Fa causes Ga I am proposing that in Bergmann’s ontology the relation ‘C(F, G)’ as given in (dd) satisfies the four crucial features and therefore functions as a nomological tie or nexus. First: The tie or connective must connect any two atomic facts representing the cause and the effect. But if Fa obtains and so does the general fact (d) then the fact Ga must exist. The existence of the general fact guarantees that, if the fact which is the cause obtains then it is required that the fact which is the effect also obtains. Since the general fact exists in its own right, this guarantee is ontological and objective, there in the world, linking cause and effect. Second: The general fact (d) connects the atomic facts as (c) requires in a way that is in effect truth-functional: there is nothing about the logical form of (d) that conflicts with the standard truth-functional logic deriving from Russell’s (and Whitehead’s) Principia Mathematica by way of Wittgenstein’s Tractatus. At the same time, the general fact (d) connects the atomic facts in a way that clearly depends on the kinds occurring in those atomic facts. Third: The kinds are straight forwardly “natural kinds” simply because they do exist as kinds which occur in the general fact (d), or, if you wish, in (dd) understood as a relational fact connecting characteristics of things understood as universals.70 Fourth: The causal tie must vary from context to context. But whatever the form of the causal connection, the general fact of the required form will exist. Thus, if the general fact (d) constitutes the causal tie (c), then automatically the regularity (r) obtains. As Bergmann would have it, (r) simply is the general fact (d) that ties together the causally connected atomic facts: so, of necessity the fact (d) that ties together the atomic facts in (c) implies regularity (r). To put the same point in another way, one that makes clear that Earman’s objection is met, the fact (d), in the form of (dd), guarantees that (de) is analytic and that the regularity (r), following from (dd) and (de), holds. Thus, Bergmann’s irreducible general facts have the four crucial features that any connection must have if it is to constitute the causal nexus that is supposed to tie together individual facts with an objective necessity. I conclude that the general fact that Bergmann insists exists over and above its instances constitutes what is in effect a causal tie or nexus that structures the kinds that occur in it in such a way that it is guaranteed that if something falls under the one kind (or kinds) it will also be of the other kind (or kinds). Notice, first, that

464 it links the kinds, as defenders of causal ties from Aristotle to Broad to Dretske have required; but notice, second, that it so binds these that the relevant pattern obtains amongst the individuals exemplifying those kinds, as the tie proposed by Broad or Dretske does not do, i.e. it meets the Bergmann-Earman objection to causal ties; and, notice, third and finally, that it secures as its reflection among observed particulars the pattern required by the connection amongst the characteristics without introducing the active soul or substance by which that correspondence was secured by Plato and Aristotle. I conclude that Bergmann’s unanalyzable general facts, while ostensibly simply facts among facts, are really, despite that appearance, objective necessary causal connections: these facts are objective necessary connections disguised as contingent regularities. What I am arguing, of course, is that Bergmann’s irreducible general facts constitute a sort of Aristotelianism – that this is his hidden Aristotelianism.71 But of course we have also listed two epistemological problems. How does Bergmann’s causal tie, his irreducible general facts, fare with these? The first epistemological problem has to do with the analytic-synthetic distinction. Any proposed causal tie or connective must satisfy the condition of being truth-functional; otherwise the traditional explication of the analyticsynthetic distinction is compromised. But there is nothing in the logical form of the general fact (d) that is in conflict with the standard explication of the analytic-synthetic distinction. So this first epistemological problem is solved. The second epistemological problem involves our knowledge of the proposed tie: the objection to the tie as proposed by Plato and Aristotle or as proposed more recently by Broad and McTaggart and even more recently by Dretske, Tooley and others, is not given to us in our ordinary experience of the world. If Bergmann cannot solve this second epistemological problem, then this is a serious objection to Bergmann’s view – it is essentially the Humean objection to any sort of tie meant to secure objective causal necessities. Bergmann attempts to meet this objection head-on: he asserts that he is acquainted with some cases of the general facts which he takes to be facts over and above the individual facts that are presented. It might be argued that Bergmann claims the general fact is something that he is not really acquainted with, and that he has argued himself into thinking that he really is presented with an irreducible general fact. Certainly, he does offer an argument that there must be general facts, an argument which appears originally in Russell. As we shall see, I think there is a point to such a comment. In any case, there are two aspects to Bergmann’s case for general

465 facts, namely, the argument deriving from Russell, and the claim that he (Bergmann) is acquainted with such facts. To be sure, there are only a few general facts with which he is (he claims) acquainted. So most of our causal judgments are simply guesses as it were that the regularities of the world are this way rather than that. For most of our causal judgments we judge that a regularity obtains but we are not acquainted with the general fact we would have to know, have to be acquainted with, if we are to know that the regularity we are guessing at really is a genuine causal regularity – one where the separable atomic facts that have that pattern are not merely conjoined but are genuinely tied together by an irreducible general fact. So most of our causal judgments are simply as the Humean says they are. But at least we do know some irreducible general facts, and therefore know the entities that in these facts constitute the causal nexus. This means that Bergmann is at least one step better than the Platonist or the Aristotelian or Broad or Dretske. Unlike these latter defenders of the causal nexus or nomological tie, Bergmann can at least claim to know the sort of entity that the tie is: since he knows in his experience of the world some instances of the tie, he can claim to know the sort of entity it is. Where these others claim to know what the causal tie is but can in fact give no instances in our experience of that tie, they are confined in practice to holding that causal judgments are judgments of regularity only: in their theories they may be anti-Humean but in their practice they are after all Humeans. Bergmann, too, is in practice like them, a Humean – “merely” a Humean, if you wish – , but unlike them he can reasonably claim to know what it is that a regularity must be like if it is to be genuinely causal and not merely a “mere” regularity. That is, he can claim reasonably so to know, provided that the case he makes for the existence of generality is sound. That case we must now examine. There is, in the first place, the argument, deriving from Russell, for the existence of general facts. And there is, second, the claim he makes that he is acquainted with generality. We can deal with them in order. We first consider the argument deriving from Russell. It is this.72 The ontologist aims to give a complete description of the world. Let us suppose we have a small world of four objects, two particulars a and b, and two characters F and G, where these form the two facts (i) Fa, Gb

466 It would seem that the sentences (i) give a complete description of the facts of this world. But Russell argues that the description is complete only if we include in our knowledge claims about the world the claim that these are all the facts in this world; our knowledge of the world is cimplete only if we know, for example, that the only particulars in this world are a and b – in symbols (ii) (x)(x = a v x = b) Our description of the model world is not complete if all we have are the descriptions (i), it is complete only if we add (ii) to the facts we know. So, besides the individual facts (i) we must include in our inventory of facts the general fact (ii). The argument can be put in another way. Suppose our world consists of a and b and that these are both H. That means that the atomic facts are given by the list (iii) Ha, Hb Instead of saying that the world is described by the list (iii), we might say that it is described by the conjunction (iv) Ha & Hb But the statement that (v) (x)(Hx) is stronger than (iv): according to the usual rules of logic, (v) entails (iv) but (iv) does not entail (v). So, in order to make our description logically complete, we must include (v) among the list of facts that exist. Generality is indispensable. Here is how Russell puts it: It is perfectly clear, I think, that when you have enumerated all the atomic facts in the world, it is a further fact about the world that those are all the atomic facts there are about the world, and that is just as much an objective fact about the world as any of them are. It is clear, I think, that you must admit general facts distinct from and over and above particular facts.73

Or so the argument goes. But it seems to me not to be sound. Consider the world (i). Surely, if in this world a is F and b is G, then (i) does give a complete picture of how things stand in that world. If I am asked to draw a picture of, let us say, our cat, then I draw the face, the mouth, the nose, the ears, the whiskers, and so on, through all the features of the cat. I then give you the picture. I do not then have to add that those are all the features of the cat. It is built into the idea of giving you a fair picture of the cat that everything is included in the drawing that I hand you. It is true that I have given all the fea-

467 tures in what I have drawn, but it is not an additional feature that just those are all the features. Similarly, for the world of (i), this list provides a complete picture of what obtains in that world: there is no additional fact that just this is everything that is in the world. Similarly for the world (iii) where everything is H: it is in the idea of a complete picture of that world that the list (iii) includes everything that is there to be pictured, and that that picture shows that everything is H. It is not necessary to include as an additional part of the picture the universal statement (v) representing as part of the picture and somehow additional the general fact (v). So, the general statement is not indispensable to a description of the world. So the Bergmann-Russell argument for the existence of general facts fails. This argument against general facts is, I think, essentially Wittgenstein’s. In the Tractatus he tells us that the world is a world of (atomic) facts, “everything that is the case.” These facts are pictured by elementary propositions. It is part of the very idea of giving a complete picture of the world that every (atomic) fact is pictured, and the names of objects are such that every name occurs in at least one picture. If the objects [in the world] are given, then at the same time we are given all objects. If elementary propositions are given, then at the same time all elementary propositions are given. (Tractatus, ¶ 5.524)

A universal proposition is a conjunction of elementary propositions. If 0 has as its values all the values of a function fx for all values of x, then N(Z) = ~(›x)(fx). (Tractatus, ¶ 5.52)

In the notation of the Tractatus, ‘0’ is a variable that takes propositions as its values, and ‘N(Z)’ represents their joint denial.74 Thus, if we have ‘fa’ and ‘fb’ as values of 0, then ‘N(Z)’ is ~fa & ~fb which is logically equivalent to ~(fa v fb) which, Wittgenstein is saying, is logically equivalent to ~(›x)(fx) And this is logically equivalent to (x)(~fx) The point is that here we see Wittgenstein affirming that the existentially quantified statement is equivalent to a disjunction of atomic propositions, which

468 means in turn that a universally quantified statement is equivalent to a conjunction of atomic statements. So, if the list fa, fb gives a complete picture of the world (this is a really small world, but that does not affect the point), then the generality Everything is f , that is, the statement that (x)(fx) amounts simply to the statement that fa & fb This what Wittgenstein meant when he said that Any proposition is a truth-function of elementary propositions. (Tractatus, ¶ 5)75

Any general proposition describing the world is simply a conjunction of atomic propositions. There is no general fact over and above the individual or particular atomic facts. Russell was later to make much the same point in his Inquiry into Meaning and Truth: Theoretically, given sufficient capacity, we could express in the object-language every non-linguistic occurrence.76

that is, every non-linguistic occurrence in the world that we are acquainted with in our experience of things, and where, as he speaks, “The words ... [which are] characteristic of logic”, including ‘true,’ ‘false,’ ‘not,’‘some,’ and (what is important for us) ‘all’, are “words which occur in the secondary language and in all higher languages, but not in the object-language,”77 that is, not in the language that records the facts of the world as given to us in sensible experience. As he went on to put the point, “‘Had we but enough time,’ we could dispense with general propositions. Instead of ‘all humans are mortal’, we could say ‘Socrates is mortal,’ ‘Plato is mortal,’ [‘Hypatia is mortal,’] and so on.”78 We do need general propositions, but not in our picturing of the world as we experience it, and therefore not as part of our ontology: In fact, however, this [enumerating ‘Socrates is mortal’ etc.] would take too long, and our vocabulary of names is insufficient. We must therefore use general propositions. But the subjective character of logical words appears in this, that the state of the world which makes a general proposition true can only be indicated by means of a

469 general proposition.79

Russell is here arguing that his former commitment to the existence of general facts alongside atomic facts is mistaken. It is indeed the case, as he argued in the “Logical Atomism” lectures, and as Bergmann also argued, that the general proposition is indispensable, and says something that we cannot say with atomic propositions alone. In the logic of our language, generality is indispensable. But in the language that is used to picture the world (what Russell here calls the “object-language”), there are only statements of atomic fact. As we go about describing the world, statements of general fact are indispensable, but the world that we are describing does not contain anything but atomic facts.80 More specifically, when we describe the world in the sense of giving a complete picture of it in language, then we give a list of atomic sentences; but in order to further assert that this picture is complete, to so describe the picture as given by the list, then at that point we need to say that, indeed, this list includes everything that needs to be said. Generality does not appear in the (object) language that pictures the world; it appears only at the level, the meta-level is you wish, in which we speak about the picture and its relation to the world it pictures. And this will be true for not only the description of the world as presently given but for any description of facts in the world that are not (yet) and perhaps not ever given, being in the past, or elsewhere beyond our perceptual powers or in the future. So, once again we may conclude that there is no general fact over and above the individual or particular atomic facts. Herbert Hochberg has criticized this argument against the existence of irreducible general facts.81 He does not suggest that every regularity is an irreducible general fact – even as some cigars are just cigars, so some regularities are just regularities –, but he does argue that there are some generalities that must be taken as irreducible in one’s ontology. Consider a three-membered universe: {a, b, c} Hochberg argues that it is a fact and, indeed, an irreducible fact that this is a three-membered universe. This fact, he argues, not merely shows itself, as we have argued, following Wittgenstein, but must be recognized in the truth of the generality (x)(x = a v x = b v x = c) which gives the structure of the universe as three-membered. If we have, say, the property F in this universe, and have moreover that Fa

470 then Hochberg’s point is that the fact that Fa is irreducible in this universe and that the generality describing the universe as being three-membered is equally a fact and equally a fact that is irreducible. One can, however, make the same point without acknowledging that there exists a general fact to that effect, a fact that is there but is irreducible to any non-general fact about the universe. At least, so I would suggest. Hochberg’s point requires the special predicate x=avx=bvx=c which specifies the universe by enumerating its members.. Let us abbreviate this predicate by the expression Ex Call such a predicate an “enumeration predicate.” Clearly, such a predicate can be defined for any universe (so long as it is finite – but that’s another story, to consider the infinite universe would take us far afield). Now, the predicate ‘Ex’ that we have just defined holds of or is true of each and every member of our little model universe, there is no member for which it is not true. If we now look at a four-membered universe {a, b, c, d} then our just-defined ‘Ex’ is true of three of its members but not of the fourth. And if we now look at a two-membered universe {a, b} then if we apply ‘Ex’ then its first two disjuncts are true but the third disjunct is meaningless since the term ‘c’ does not refer to any entity in this universe (assuming the rule of unum nomen unum nominatum). Evidently, for each universe there is an enumeration predicate, and only one enumeration predicate such that (1) it applies to each individual in the universe, (2) it applies to every individual in the universe, and (3) it contains no disjunct that is meaningless because the constant which appears in that disjunct does not refer. Call such a predicate the enumeration predicate characteristic of that universe. So ‘Ex’ is the enumeration predicate characteristic of our little three-membered universe. Hochberg holds that we cannot capture the fact that our three-membered universe is of a certain size, that is, is three-membered, by holding that its threememberedness shows itself in any picture of that world (as Wittgenstein held and as we have argued). Rather, we can capture that fact of threememberedness only if we hold the generalization (x)(Ex) to be true (which it is) and to be irreducible to any statement or combination of statements about the atomic facts of that universe. But surely all that we need to recognize is that the conjunction

471 Ea & Eb & Ec is true, that there is no conjunct of ‘Ex’ which is meaningless, and that there is no statement of the form Ex for example Ed which is false. Surely recognizing that the enumeration predicate characteristic of an n-membered universe is true of each and every one of the members of the universe and that it has no meaningless disjuncts, captures as it were the truth the universe is n-membered: it is not necessary to go beyond this and add And, ah yes, besides all this, it is also the case that there is an additional fact represented by ‘(x)(Ex)’ where ‘Ex’ is the enumeration predicate characteristic of the universe. So Hochberg is wrong: there is no need to include among the list of irreducible facts in the universe besides the atomic facts also the general fact represented by that universal generalization which states that the enumeration predicate characteristic of the universe holds of every individual in the universe. I therefore reject Hochberg’s argument for the existence of at least a few irreducible general facts. The criticism that we have developed following Wittgenstein and Russell of the Russell-Bergmann82 argument for the existence of general facts based on the logical indispensability of generality seems to be a sound criticism. Return now to Bergmann. We must consider the second argument that he can solve the second epistemological problem that confronts any proposal that there is some special causal nexus. This second argument that he gives for the existence of generality is based on the claim that he is acquainted in his experience of the world with generality as part of the logical structure of that world. Bergmann has argued consistently that ontology should be done in conformity with the Principle of Acquaintance – the rule that nothing is to be admitted into one’s ontology unless one is acquainted with that entity or with entities of that sort. In his case of the ontologization of generality, his making it into an entity in his ontology, is justified by the Principle of Acquaintance. That is, he claims that generality is presented to him in his ordinary experience of the world. I merely claim that (a) generality is presented on some such occasions [that is, occasions when “one uses or could use, for the best of reasons, a general sentence”], and (b) if it were not, we would not on any occasion know what the quantifier meant.83

472 “I take (b) to be obvious,” he comments.84 But in reply one might perhaps suggest (as some have done) that he is not so much acquainted with generality as convinced by the dialectics that he is acquainted with it. In any case, he gives the following example.85 We have two individuals a and b. Individual b has the property C of being a circle, while a has the property S of being a square. Besides the characteristics C and S, there is Rxy the relational characteristic of x being inside y. The situation is this: a is inside b, and is the only square inside b. Bergmann supposes himself to say (!) This square is the only one inside this circle The transcription of (!) is not (+) Rab that is, not this alone, but rather the conjunction of this together with (++) (x)(x = a v ~Rxb) or, more fully, taking into account the characteristics square and circle, (x)[(Sx & Rxb) e x = a] Bergmann now notes that (++) contains the generality operator. He then comments that I conclude that generality is on this occasion presented to me in connection with the fact [Rab].

and adds that Instead of saying that it is (on this occasion) presented in connection with [Rab], one might as well say that it is (on this occasion) presented in addition to [Rab].86

What can one say? Take the situation that Bergmann has described. It is there, in the world, given to us in our sense experience of that world and of those facts in the world. If one is asked to give a linguistic picture of the situation, then one would give: (s) Sa , Cb, Rab That is how one would describe the situation, picture it in language. The point is that (s) is the complete picture. To be sure, not everything that we know about the situation is there in the linguistic picture we make of it. Thus, from the picture we can see that the square is inside the circle and that it is the only square that is inside the circle. That this is true, is something we discover in looking at the picture. It is not something that is said in the linguistic picture (s), as a fact alongside as it were the facts that are pictured, that is, the “other facts.” It is not there as part of the picture, but still, it is a fact that is shown to us by the picture we make of the situation when we made a complete picture of

473 the situation. In that picture, it is shown that a is the only square inside the circle b: if there were another square there then there would be another atomic fact of the form Sx and another relational fact of the form Rxb. Given that the picture is complete, given that it is a complete picture of this little world, then there is no need to add that in addition to the facts pictured, there is beside them the further fact that b is the only square inside the circle a. Bergmann’s situation is given in sense experience, and we know about that situation what Bergmann records in the quantified statement. But that knowledge is not given to us in our sense experience of the world. That general fact is before the mind, but it is not a fact with which we are acquainted, that is, the general fact which is before our mind is not one with which we are acquainted in our ordinary sense experience of the world. It derives from our thinking about that world and our picturing of it, but it is not in the world as we sensibly experience it. That is why, as Russell put it in the Inquiry, the description of the world as we experience it appears in our object-language but generality as a logical concept appears only in our secondary language which we use as we reflect in thought about the world described in the object-language. I conclude that Bergmann’s claim that he is acquainted with generality is not sound. However, it must also be said that the general fact is certainly before his mind, it is in the intention of an act of thinking about the situation – indeed, one cannot doubt that that act, that mode of thinking, can be called perceptual: the general fact is not in the world as we know it in our sense experience, but nonetheless it is something about the world that we perceive.87 Indeed, Bergmann later came to the conclusion that generality is in fact not given in our sensible experience of the world. In his New Foundations of Ontology, our sensible experience of the world consists of acts which he refers to as “primary Perceivings and Imaginings (and undoubtedly some ‘Feelings’).”88 But general facts are, he now acknowledges, entities that we cannot perceive or imagine. The awareness of a general fact must be a believing or an entertaining, not a perceiving or an imagining.89 Bergmann is therefore here granting Russell’s point and ours that general facts are not really among the facts of the world as it is given to us in sensible experience. They are therefore excluded by the Principle of Acquaintance. But Bergmann gives up that Principle, at least in its original sense, where ‘acquaintance’ means ‘sensible awareness’ (and inner awareness). He now expands ‘acquaintance’ to mean any act of awareness in the sense of thinking about, thus including believings and entertainings (and any other sort of thinking) as being ways of being acquainted with facts. He now holds that “whatever

474 is thinkable exists.”90 Thus, the intention of any act exists. And we can think of or about general facts. So they exist. But why accept that principle? Surely it gets us into insurmountable problems. Thus, I can think that Toronto is east of Brockville For, that is something I disbelieve, and in disbelieving it, I am thinking it, and so it exists. So Bergmann’s new Principle of Acquaintance commits us to the existence of false facts – surely something intolerable, even if these false facts are (as Bergmann would put it) pervaded by the mode of potentiality rather than actuality. Put in these terms, one can begin to see how Bergmann could come to hold that, since a general fact is the object or intention of thought, he is acquainted with a fact existing alongside other facts, other objects in the intentions of thoughts. We therefore have as the intention of a thought a fact, namely, the general fact, that we are not presented with in our sensible experience of the world. Does this fact therefore exist? Or, more generally, what, then, do we say about intentional objects? Bergmann says they all exist. Is that so? That seems intolerable as a conclusion, no matter what Bergmann says about such things as the mode of potentiality. We must ask, what gets one into this problem? It seems clear that not every intentional object exists. There are no false facts, for example. Yet Bergmann insists that every such object does exist. There seems to be something that has gone wrong with how he is thinking about intentionality. So, let us recall Bergmann’s formula for representing thoughts: +p,Mp which says that the thought that p means (the fact) that p. ‘p’ as a sentence is the text of the thought. This text is perhaps actually said or, as most often occurs, there is a short-term disposition to utter that text. ‘+ p ,’ is the name of a simple character, the thought. So at least Bergmann proposes. Actually, it would seem better to think of ‘+ p ,’ as a definite description: it is a definite description to this effect, “the thought of which ‘p’ is the text.”91 The important feature in Bergmann’s account of intentionality is ‘M’. When Bergmann first introduced this to represent the intentionality of thought, he thought of ‘M’ as linguistic, and that a statement to the effect that +p,Mp is true by virtue of its linguistic form alone – which makes it a sort of logical truth, since, like all the positivists, he took logical truths to be truths which hold

475 by virtue of their linguistic form alone.92 But he gradually came to the conclusion that, like all of logic, this formal fact too had to be ontologized: the linguistic ‘M’ had to be construed as representing something existing in the world.93 This last meant treating M as a sort of a relation, that is, a relation similar to an ordinary relation like kicks. But if it is true that Beverley kicks Georgie then both the relata, Beverley and Georgie, must exist. So, if M is a relation, then its relata must exist. Hence, if we have +p,Mp then both relata, and in particular the relatum p must exist. Hence, if we are thinking of or aware of or entertaining the fact that (++) (x)(x = a v ~Rxb) then we have + (x)(x = a v ~Rxb) ,M (x)(x = a v ~Rxb) and the second relatum must exist. Similarly, if Bergmann is presented with the fact about the sensible world that Ca then we have + Ca ,M Ca and again the second relatum exists. So, when Bergmann is aware of the fact that Ca which exists, he similarly is aware of the general fact (++) which also exists, exists alongside the individual fact Ca. So there we have it: there are general facts, there in the world, alongside particular or atomic facts. We can think them so they must exist. There is a mistake here. It is just wrong, we have been arguing, to hold that the general fact exists alongside the particular facts; such a generality does not exist (though the generality often enough is true). Something has clearly gone wrong. What has gone wrong, I suggest, is Bergmann’s treatment of M as a relation. More correct is his original view that this represents something linguistic, rather than being something like a real relation similar to kicks. If we think of ‘p’ means p as a linguistic statement about the meaning of the sentence ‘p’, then means, that is, M, will allow us to have the thought that p, i.e., + p ,, to have p as its object without there being any requirement that p exist. Thus, in ‘Toronto is east of Brockville’ means Toronto is east of Brockville we have a true statement of meaning even though it is false that, and therefore not an existing fact that

476 Toronto is east of Brockville This, as we have said, is intolerable: there are not in the world alongside the (positive) facts another set of entities, the false facts. Bergmann is committed to the view that these objects of thought which do not exist do after all exist.94 Surely we must, as Russell once put it, as philosophers have a robust sense of reality; false facts violate, or should violate, anyone’s robust sense of reality. Besides, and this is what is crucial, we are not acquainted with them as being in the world as it is given to us in our sense experience of it: they violate the Principle of Acquaintance. Thus, thinking of the ‘M’ in ‘p’ M p as linguistic yields an account of intentionality that does not force one into the position in which Bergmann finds himself, that of requiring that the object of thought, the intention, whatever it may be, exist. One can accept the Principle of Acquaintance without requiring oneself to hold that every object of thought is a case where we are acquainted with something that exists. Treating ‘M’ as linguistic enables one to hold, on the one hand, that generalities are among the objects of thought, while also holding, on the other hand, that such an object of thought is not a fact alongside the facts we are acquainted with in our sensible experience of the ordinary world. The sentence ‘p’ is the text of the thought. Its meaning, in the linguistic sense, is given by the various patterns of language that govern it and the terms that occur in it – these are regularities and dispositions that constitute worldword connections (e.g., rules of reference), word-word connections (e.g., rules of syntax, rules of logic, statements of generality), and word-world connections (rules for speech acts), and involve, in complicated ways, the causal role of rule expressions (resolutives and imperatives).95 This meaning that determines what ‘p’ means, and in particular if it is atomic what it refers to and pictures, is a matter of patterns and dispositions. These by their very nature are not wholly present in consciousness when we use ‘p’ or are disposed to use ‘p’. Patterns and dispositions are known by inference, yet when we use ‘p’ we are aware non-inferentially of that meaning: There is present in consciousness a characteristic or feature that is the simple feature of that conscious state that causes us to use or be disposed to use ‘p’. ‘+ p ,’ is a definite description of this feature: it is the feature that causes us to utter or be disposed to utter ‘p’, the text of the conscious state; and +p,Mp represents the fact that the feature referred to by ‘+ p ,’ causes the fact that one utters or is disposed to utter the text ‘p’. Thus, the meaning of the conscious state, what it intends, is given by the linguistic patterns governing ‘p’ and the

477 terms that occur in it. ‘M’ gives it that the conscious state has as its meaning or intention the fact that ‘p’ means, it also represents that there is present in consciousness the thought that causes us to utter or to be disposed to utter the sentence ‘p’ with that meaning.96 Understanding intentionality in this way means that we are not treating the text of every act of awareness, every intentional state, as having an object that has any ontological status. And so, perceiving that a is the only square inside the circle b does not require one to give ontological status to the general fact.

However, there remains another point that Bergmann makes, again following Russell. Logically speaking, the generalization (x)(Fx) is stronger than Fa & Fb & Fc & ... no matter the length of the conjunction. So “generality” cannot mean “very long conjunction.” Nor can it mean what an ellipsis “…” means; an ellipsis does not occur in any symbolism that purports to be logic. What, then, does it mean? Bergmann suggests that one couldn’t know the meaning of generality unless it was an entity that is presented to one. Since there are no general facts over and above the atomic facts, or so we have argued, we cannot have acquired our concept of generality from acquaintance with the entity generality as part of a general fact. We have, therefore, yet to explain the meaning of the concept and how we acquire that meaning. Wittgenstein makes the relevant point, I think. He holds, correctly we are arguing, that ontologically there is no difference between the general proposition ‘(x)(Fx)’ and the conjunction ‘Fa & Fb & Fc & ...’: recall his comment that Any proposition is a truth-function of elementary propositions. (Tractatus, ¶ 5.53)

This does not mean that generality is the same thing as ‘(long) conjunction’ or “ellipsis.”.Thus, he tells us that What is peculiar to the generality-sign [as in ‘(x)(Fx)’] is first, that it indicates a logical prototype, and second that it gives prominence to constants. (Tractatus, ¶5.522)

What ‘(x)(Fx)’ does is give us a prototype for a fact: this prototype is ‘Fx’.97 This prototype gives prominence to, in particular, the constant ‘F’: the prototype is therefore the prototype of a fact in which the characteristic F is exempli-

478 fied. Thus, what ‘(x)(Fx)’ means is more than what a conjunction means: it means that as one continues to construct pictures of the facts of the world, that is, as one continues to give the elementary propositions that picture the facts of the world, then every individual, no matter what, will fit the prototype ‘Fx’: for every individual there is or will be or was, that individual is prototypically F. The generalization is not a conjunction but a declaration that pictures of the world will have in them facts that fits this prototype, an anticipation of the structure of facts in the picture of the world, whatever picture it is, however large the world becomes. This, then, is the meaning of the general quantifier. The meaning of an elementary proposition is given in terms of the state of affairs that it pictures. An elementary proposition is correct as a description of the world just in case that the state of affairs that it pictures is there in the world as it is given to us. The meaning of the general proposition lies in its giving a picture of the pictures that will be there in any complete description of the world.98 As for a regularity of the form (x)(Fx e Gx) this too should be understood in terms of prototypes. It says that if a fact fits the prototype Fx then it will also fit the prototype Gx while if there is a fact that does not fit the first of these prototypes then it may or may not fit the other prototype. It lays down, in other words, that future pictures will fit one of these three prototypes: Fx, Gx ~Fx, Gx ~Fx, ~Gx 99 This notion of generalizations as describing the truths of known facts and as projections of the forms of facts yet-to-be observed is likely what John Stuart Mill meant when he argued that “General propositions are merely registers of such inferences already made, and short formulae for making more.”100 And now we are brought back to Socrates and Anaxagoras. The world as we experience it consists of separable parts, parts that are logically and ontologically separable, parts that are conjoined but merely conjoined. If Anaxagoras is correct, then there is nothing more, only these parts and the patterns that describe how parts of just these sorts are associated with parts of just these other sorts. If Socrates is correct, then this account of the world misses something: what it misses is the tie that provides the reason why these parts are

479 associated with those parts. Why is there this pattern rather than that? Because, Socrates proposes, there is a reason for this, an ontological ground, for this pattern being more than a mere pattern, but instead one that must hold of the world as we experience it, a ground or tie which makes of this pattern more than a “mere” pattern. Recall that Socrates put his point in terms of his own situation. Here he is, Socrates, sitting in prison. According to the earlier dialogue, the Crito, his friends offered to arrange things so he could escape to Thebes. But Socrates refused: following the demands of justice he determined that it would be wrong to escape his native city and run off to Thebes. He is, instead, determined to remain in his cell and drink the hemlock. So, we have one event, Socrates sitting in his cell. There are two events that could follow that event, namely, running off to Thebes and remaining in Athens and drinking the hemlock. There is nothing in the first event, Socrates sitting there, that indicates or determines that it will be followed by one rather than the other. But in fact it is the drinking of the hemlock that follows. Now, Socrates complained that Anaxagoras’ account of things could not explain this sequence that actually occurs: the explanation must be in terms of Socrates “aiming at the best”. But Anaxagoras’ scheme has no model of the “best” for Socrates to aim at: there are no ideal forms and in particular no ideal form of human justice. Lacking an objective standard of virtue, Anaxagoras cannot explain Socrates’ action. So at least Socrates argues. But (here we are giving a helping hand to Anaxagoras), Anaxagoras need not deny that Socrates is aiming at the “best.” But this best is what Socrates values as the best. Anaxagoras can argue that Socrates is in fact moved towards this goal that he identifies as the best, but that there is nothing to this beyond that valuing, no objective standard in the ideal forms of things: there are no transcendent ideal forms and therefore no objective standard. At least, so he could argue: certainly, that is the way that Hume was much later to argue when he developed a similar case against the Platonists and other moral objectivists of his own day. There are no objective values but there are relative values. Anaxagoras thus has a reply to Socrates: he can explain the events that Socrates wishes to explain without any appeal to transcendent entities, either simple souls or ideal forms, entities not of this world that somehow effect a tie of necessity between events in the ordinary world. All he needs is regularity: Socrates values what he feels to be just, and we know that he regularly acts to make the world, himself in particular, be as justice, so envisioned, requires. Relative values and regularities about things in the world, including people, suffice. There is no need for objective values or forms nor is there any need for a

480 tie or causal nexus or objective necessary connection. Socrates and after him Plato and Aristotle, and then, in our own age McTaggart and Broad, and more recently Dretske, have all sought for such a tie. Fruitlessly: their claims notwithstanding, none have passed the test established by the Principle of Acquaintance. We seem to be left with the atomism of Anaxagoras, of Hume, and of the logical positivists, and, among the last, Bergmann. Interestingly, we found that Bergmann, starting as a logical positivist, came to a position on the ontology of logic which seemed in a way to do what Socrates & Co. were trying to do, namely, give a tie that would bind the separable facts of our sensible experience into necessary unities. This came when he argued that besides the particular facts that we find in the world as we experience it, there are also general facts. But when we examined those arguments, one based on the inferences of logic and one based on acquaintance, we found that neither of them withstood scrutiny. We found both these rooted in a mistake that Bergmann made about the “relation” – the apparent “relation” – of intentionality. Thinking of it as a real relation, he naturally concluded that there really were general facts that were given to him in our perceptual experience and that he had no choice but to grant them an irreducible ontological status. But we also argued that an alternative account of intentionality, in which, as Bergmann originally held, whatever its grammatical form, it is not a real relation, would eliminate that structural reason for thinking that there are general facts.101 So we are back to the metaphysics of logical positivism from which Bergmann began his ontological journey. And here we must say that there is a certain point to what Socrates was arguing against Anaxagoras. Here I do not mean the point about value. Anaxagoras’ ontology, unlike Socrates’, lacks objective value. Anaxagoras’ ontology does have souls and they do move bodies, just as do Socrates’ souls. But unlike Socrates’ souls, the minds or portions of mind that move individual things on Anaxagoras’ scheme do not transcend the world of ordinary experience. They consist of separable parts and we can see Anaxagoras’ struggling with language to describe them. They are fiery, or like fire, he says.102 Their parts are in rapid motion – we might recall Hume’s comment on selves that “they are nothing but a bundle or collection of different perceptions, which succeed each other with an inconceivable rapidity, and are in a perpetual flux and movement.”103 Or we might recall Moore’s point that mental acts are diaphanous or transparent.104 However the fiery bits go together, the separable parts of mind or noûs, they do separate themselves from other bits of noûs, and direct some at least of the corporeal groupings or bodies which emerge out the primordial mixture. Like Socrates’

481 souls, these groupings of fiery bits give bodies direction: they are purposive. But there are no transcendent forms that constitute a set of objective values. For Anaxagoras, if one may put it this way, there are valuings but no values, that is, no objective values, but there are things that are valued, these are the goals towards which mind or noûs directs the body which it moves. Or, in other words, there are in Anaxagoras no objective values, but there are relative values, the ends towards which the bodies are directed. This is the view of Hume, of the logical positivists, and of Bergmann.105 It is not this that I mean when I say that there is a point to what Socrates says. I have in mind, rather, the point about the need for a tie. Without such a tie, without objective necessities, nothing can be explained. This is the point that Plato and Aristotle, and Broad and McTaggart and Dretske are all trying to get at. Just this sort of tie is absent from Anaxagoras – at least from what we know of Anaxagoras. That at least was the reason that formed the basis of Socrates’ criticism of Anaxagoras. Such a tie of objective necessity is also absent from the Humean world of the logical positivists and of Bergmann. This tie or necessary connection that is missing from the worlds of these philosophers is just that real connection that distinguishes those patterns or regularities that are laws and those that are “mere” regularities, mere “accidental” generalities. In fact, as we have argued these are entities with which we are not acquainted in the world of our ordinary sensible experience. They are excluded from our ontology by the Principle of Acquaintance. But at the same time we seem to need them in order to distinguish those patterns which are lawful or necessary and those which are accidental and merely contingent. We seem, after all, to be driven, like Socrates (or Dretske), to find a world outside the world of ordinary experience, the world that transcends our ordinary world, but the world in which we find the tie that we must grasp if we are truly to have the reasons which explain things in the ordinary world. And so Hume gives his first definition of ‘cause’ as regularity: ontologically there is no distinction between laws and accidental generalities. But then he immediately goes on to acknowledge, along with Broad and McTaggart and Dretske, and of course Socrates, that after all one must distinguish two sorts of regularity, those that are causal and those that are merely accidental. A violation of an accidental generality is not impossible, but an event that violates a causal law is somehow impossible. So Hume asks that question, “What is our idea of necessity, when we say that two objects are necessarily connected together?”106 There is, then, after all, a necessity for which we must account. Hume saw this. It could of course not be some sort of objective necessity: all regularities are, objectively, equally contingent. That is what Hume just established.

482 Hume concludes that therefore such necessity as there is must be a subjective necessity. Just as there are no objective values but there are relative values, so also, although there are no objective necessary connections, there are relative necessities, subjective attitudes of felt necessity that distinguish those generalities that we take to be lawful from those that we take to be merely accidental. It is clearly possible, I suggest, to have much of Bergmann’s philosophy, e.g. almost all his philosophy of science,107 even if we give up generality. Some logical positivists were prepared to abandon the Humean concept of causation as they attempted to account for the required necessity. Carnap was to compromise on this issue when he introduced the notion that predicates could be introduced by “reduction sentences” to make a place in the language of science for disposition terms like ‘soluble.’108 Others were to introduce unanalysed modalities, objective causal necessities,109 when faced with the issue of justifying the assertion of counterfactual conditionals such as “If I were to heat this water, it would boil” which can be justified by “water, when heated, boils”, where I am not justified in asserting “If this coin [which is a loonie] were in my pocket, it would be silver” even though it happens – that is, happens accidentally – to be the case that “All the coins in pocket are silver.”110 Bergmann saw the way out: the difference between a statement of law and a mere accidental generality is one of context, that is, psychological context, the way we use the generality in our inferences.111 It is not that the counterfactual conditional (%) If a were F then it would be G is justified by the fact that the regularity (%%) All F are G is objectively necessary, but rather that we count (%%) as necessary because we in fact use it to support the assertion of (%), where if we didn’t so use it we would count it to be a “mere” accidental generality. Hume had made this point already. He had given his first definition of ‘cause’ as regularity. This was causation or lawfulness understood objectively. And objectively there was no necessity: objectively there is no difference between laws and accidental generalities. But for all that, Hume recognized, as Bergmann was later to recognize, that there is a necessity that must be accounted for. Hume provides this account in his second definition of ‘cause’ as “An object precedent and contiguous to another, and so united with it in the imagination, that the idea of the one determines the mind to form the idea of the other, and the impression of the one to form a more lively idea of the other”112 If C’s cause E’s, then if I observe a C, that is, in Hume’s terms, have an impression of it, then there is an association in the mind such that I am led to expect an E, that is, I come to have the idea of an E. In this case, on the basis of the regu-

483 larity I am predicting that the C I observe will be E. But I may also simply suppose that something is a C: in this case I have the idea of a C. In this case, the established association leads me to the idea that this something is also an E. If this something were C then it would be E. In other words, I use the casual regularity to make contrary-to-fact assertions. The causal regularity thus describes not only what is actual but what must be in the actual (predicting what must be) and placing limits on what is possible (contrary to fact assertions). Thus, a violation of a causal law is not only contrary to fact but something that is not possible. What is crucial is that this necessity is not objective but rather is subjective: a casual law simply is a regularity towards which we have the psychological attitude of being prepared to use it to predict and to make contrary to fact assertions. An accidental generality is, in contrast, one that we do not use to predict or to make contrary to fact assertions. This is our idea of necessity: it is rooted in the pragmatics of law assertions. Of course, some people assert that If a black cat were to cross my path then bad luck would ensue Here they are connecting the supposed events by the generality that Whenever a black cat crosses one’s path then one will have bad luck. This is superstition. But these people are asserting a contrary-to-fact conditional on the basis of a generality. That makes the generality lawlike, just as Water, when heated, boils is lawlike because we use it to assert conditionals like If this water were to be heated, then it would boil The regularity about water is reasonable. The one about cats is superstition. We therefore, as Humeans, do not have an objective criterion for distinguishing those generalities which are reasonably taken to be lawlike from those which are not, that is, those which are superstition. And there is a distinction here to be drawn: there is surely a distinction – an objectively valid distinction – between science and superstition. But the objective necessary connections of Plato and Aristotle, of Broad and McTaggart, and of Dretske provide just such an objective criterion. Surely, then, we have been too hasty in our dismissal of those ontologies. Otherwise, do we not fall into a scepticism about which generalities are laws and therefore reasonable to assert and those generalities which are not laws and which are therefore unreasonable to assert? Do we not fall into a subjectivistic relativism? But there is a way to distinguish science from superstition without invoking any mysterious objective necessary connection. In point of fact, the gener-

484 alization about water is well supported by empirical evidence, the generality about black cats is not. The objection depends upon confusing the notion of treating a generality as lawlike with the notion of reasonably treating a generalization as lawlike. When we can assert a generalization on the basis of evidence that conforms to the norms of good scientific practice, then we can reasonably treat it as lawlike, and use it to predict and to assert contrary-to-fact conditionals, and therefore to explain. Whether or not a generality is lawlike is a subjective matter, but whether or not the evidence we use to justify asserting the generality is good scientific evidence is an objective matter. The necessity of a law is subjective, the reasonability of such necessity is objective.113 We do not fall into an irrational subjectivism and scepticism about laws after all. The account of causal connection may be Humean, but that should be no cause for alarm: no subjectivistic scepticism looms before us. Indeed, given that the claim that there is an objective causal tie or necessity is one that cannot be found to hold in the world of everyday experience, the only world we know, it is the defender of objective necessities who falls into scepticism. The Humean viewpoint may be second best, given what the defender of objective necessities hopes for, but since that hope cannot be fulfilled, the second best is the best that we can do: the reasonable person will simply decide that the Humean position is the only reasonable one to adopt.

Bergmann began as a positivist, accepting a Humean position on laws. For a variety of reasons he came to accept the claim that over and above particular or individual atomic facts there are irreducible general facts. He even suggested that without such general facts there could be no laws. We have argued that this irreducible general fact is in effect a return an Aristotelianistic world of objective necessary connections. But we have also examined his reasons for introducing generality as an existent into one’s ontology, and have concluded that they are in fact unsound. So the Humean account of laws and causation stands: the earlier positivist metaphysics emerges after all as defensible.

485 Endnotes to Study Twelve

1.Gustav Bergmann, “Generality and Existence,” in his Logic and Reality (Madison, WI: University of Wisconsin Press, 1964), pp. 64-84. Compare his Realism (Madison, WI.: University of Wisconsin Press, 1967): “Generality and existence are the two subsistents represented by the universal and existential operator, respectively.” (p. 94n). But the arguments for these two entities are in “Generality and Existence.” In his posthumous New Foundations of Ontology, ed. W. Heald (Madison, WI: University of Wisconsin Press, 1992), Bergmann argues that there could not be any laws of nature if generality were not in the world: “If [generality] were not in the world, there could not be any laws of nature.” He goes on to remark that “Ever since Hume, philosophers have been so absorbed by the question of how, if there are such laws, we may come to know them, that they have become oblivious of this necessary, although to be sure not sufficient, ontological ground for there being any.” (p. 173) But again, the arguments are essentially those of “Generality and Existence,” so the latter are what we shall concentrate on. It should be noted, however, that the analysis of generality as an entity changes to a very novel account in New Foundations. It is perhaps worth noting that Hume is not concerned just with the issue of how we know generalities to be true; as we shall see, below, he is also interested in the ontological issue. In fact, as we shall also see, he is in effect arguing that generality in Bergmann’s sense does not exist. In that sense, Bergmann is right: for Hume, generality as such does not exist and there are therefore no laws of nature, there are only regularities, patterns among the particular facts that make up the world. It should perhaps be noted that for Bergmann the entities he says are subsistents all exist: subsistence is a mode of existence. Or rather, this is so by the time of Realism. In “Generality and Existence” subsistents are there, in the world, but they are not fully existents. The point is relevant for many aspects of Bergmann’s thought, but not for what we are about: we may take it that, for all intents and purposes, generality is there, in the world, and in that sense does exist. It should be remarked that Heald’s Introduction to New Foundations is important for anyone interested in Bergmann’s later thought. See also P. Butchvarov, “Bergmann and Wittgenstein on Generality,” Metaphysica, 7 (2006), pp. 123-145; and “Metaphysical Realism and Logical Nonrealism,” in Richard Gale, ed., Essays in Metaphysics (Oxford: Blackwell, 2002), pp. 282-382. 2.B. Russell, “The Philosophy of Logical Atomism,” in his Logic and Knowledge, ed. R. C. Marsh (London: Allen and Unwin, 1956), pp. 175-282. 3.Gustav Bergmann, “Logical Positivism,” in his Metaphysics of Logical Positivism (New York: Longmans, Green & Co., 1954), pp.1-16, at p. 2.

486

4.For a systematic and historical treatment of Bergmann’s thought, see the very insightful study by Herbert Hochberg, The Positivist and the Ontologist: Bergmann, Carnap, and Logical Realism (Amsterdam and Atlanta: Rodopi, 2001). 5.Plato, Phaedo, in Plato, Five Dialogues, trans, G. M. A. Grube (Indianapoliis, IN: Hackett, 1981). 6.Phaedo, 97c1 ff. 7.See G. S. Kirk and J. E. Raven, The Presocratic Philosophers: A Critical History with a Selection of Texts (Cambridge: Cambridge University Press, 1957), Chapter XV, pp. 362294. 8.See Kirk and Raven: Anaxagoras writes (according to the fragments) that “But before these things were separated off, while all things were together, there was not even any colour plain; for the mixture of all things prevented it, of the moist and the dry, the hot and the cold, the bright and the dark, and of much earth in the mixture...”. (p. 368, fragment 496). Or again, “... this [initial] rotation caused the separating off. And the dense is separated from the rare, the hot from the cold, the bright from the dark and the dry from the moist.” (p, 373, fragment 503) And so we can get flesh and bones, but the presence of the “dark” allows for shadows and the “bright” allows for rainbows: the world is as much phenomenal as it is corporeal – though this more recent terminology is no doubt misleading. The point is that, roughly speaking, the world of Anaxagoras’ ontology is the world as given to us in our sensible experience of it. Theodor Gomperz puts it this way: for Anaxagoras, “[t]he nature of objects is such as the senses perceive ….” (Theodor Gomperz, Greek Thinkers, 4 vols., trans. Laurie Magnus [London: John Murray, 1901], vol. 1, p. 210). 9.Simplicius (who is a major source on Anaxagoras) at one point suggests that Anaxagoras, like Plato and the Socrates of the Phaedo, “supposes a two-fold world, one intelligible and the other (derivative from it) perceptible...” – a world of rational forms of things and a world of perceptible things of which these are the forms (see Jonathan Barnes, Early Greek Philosophy [Harmondsworth, Middlesex: Penguin, 1987], p. 228). But as Kirk and Raven point out, the most that the passages from Anaxagoras that Simplicius quotes can support is that perhaps Anaxagoras held (like Anaximenes) that there is a plurality of contemporary worlds (see Kirk and Raven, p. 379f). 10 Theodor Gomperz writes: “The great objects of nature [e.g. the sun] were no longer divine in his [Anaxagoras’] eyes: they were masses of matter, obedient to the same natural laws as all other material aggregates whether great or small …. There was only a single point in his theory of the formation of the firmament in which he deserted his mechanical and physical principles to assume an outside intervention …. [This was] that first shock which set in motion the process of the universe that had hitherto been in repose …” (Greek Thinkers, vol. 1, p. 217). The sun was no longer Helios the god, but simply an “ignited stone.” Noûs, whatever its exact metaphysical or physical role, “had completely released him [Anaxagoras] from the old mythological fetters,” (ibid.)

487

11.See Kirk and Raven: “And the things that are mingled and separated and divided off, are all known by Mind. And all things that were to be, all things that were but are not now, all things that are now or shall be, Mind arranged them all...”. (p. 373, fragment 503) 12.Thus, Anaxagoras writes (so we are told) that “...Mind, which ever is, is assuredly even now where everything else is too, in the surrounding mass and in the things that have been either aggregated or separated.” (Kirk and Raven, p. 374, fragment 506) 13.This is how Kirk and Raven (p. 375) read this fragment: “And when Mind initiated motion, from all that was moved separation began and as much as Mind moved was all divided off; and as things moved and were divided off, the rotation greatly increased the process of dividing.” (p. 373, fragment 504) 14.It is perhaps misleading to suggest, as do Kirk and Raven, The Presocratic Philosophers, that “mind, like matter, is corporeal...” (p. 375). The term ‘corporeal’ suggests that mind is something more bodily than simply consciousness. Their earlier remark that “Anaxagoras ... is striving... to imagine and describe a truly incorporeal entity” (p. 374) seems to be more just. The point is that these things which are incorporeal (“fiery”) can causally affect and causally interact with the things of flesh and bones that are more fairly described as material and are in that respect like things corporeal. 15.For an elaboration of the Socratic position, see F. Wilson, Socrates, Lucretius, Camus – Two Philosophical Traditions on Death (Lewiston NY and Queenston ON: Edwin Mellen Press, 2001), Chapter Three. 16.Here we have what has been called, not unjustly, the “natural philosophy” of the Phaedo. For more on this, see Wilson, Socrates, Lucretius, Camus, Chapter Three. For the basic argument that what we have here in the Phaedo is the pattern that is later worked out as the pattern for the explanation of all ordinary events, see G. Vlastos, “Reasons and Causes in the Phaedo,” Philosophical Review, 78 (1960), pp. 291-325; and R. G. Turnbull, “Aristotle’s Debt to the ‘Natural Philosophy’ of the Phaedo,” Philosophical Quarterly, 8 (1963), pp. 131-143. 17.See Wilson, Socrates, Lucretius, Camus, p. 121ff. 18. This appeal to unanalyzable activity as a metaphysical category means that such explanations are inevitably teleological (see J. Barnes, The Presocratic Philosophers, 2 vols. [London: Routledge and Kegan Paul, 1979], vol. 2, p. 108f). Plato and Aristotle of course accept this implication. Anaxagoras rejects it – that is Socrates’ objection to Anaxagoras, that in his explanation there is no reference to noØs or to the Good, i.e. the end the activity is striving to instantiate. 19.Cf. F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies (Toronto: University of Toronto Press, 1999), Study One; and “Science and Religion: No Irenics Here,” Metaphysica, 7 (2006), pp. 159-170.

488

20.This account of explanation, in other words, makes causality to be regularity; it is in other words, essentially Humean. Jonathon Barnes makes the same point; see his The Presocratic Philosophers, vol. 2, p. 108ff. 21.See F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies, Studies One and Two. 22. F. Dretske, “Laws of Nature,” Philosophy of Science, 44 (1977), pp. 248-68. 23.Dretske, “Laws of Nature,” p. 262. 24.Ibid., p. 263. 25.D. M. Armstrong, What Is a Law of Nature? (Cambridge: Cambridge University Press, 1983). 26.M. Tooley, “The Nature of Laws,” Canadian Journal of Philosophy, 7 (1977), pp. 66798. 27.J. Brown, The Laboratory of the Mind (London: Routledge, 1991). 28.John Earman, “Laws of Nature: The Empiricist Challenge,” in R. J. Bodan, ed., D. M. Armstrong (Dordrecht, The Netherlands: D. Reidel, 1984), p. 221n21. 29.B. Russell, Inquiry into Meaning and Truth (New York: W. H. Norton, 1940). 30.I have substituted ‘human’ where Russell has ‘man’. 31.Russell at this point uses ‘human’ instead of the earlier term ‘man’. 32.Russell, Inquiry into Meaning and Truth, pp. 116-117. 33.G. Bergmann, “On Non-Perceptual Intuition,” in his Metaphysics of Logical Positivism (New York: Longmans, Green & Co., 1954), pp. 228-231. Earman seems to have overlooked this important essay, which makes the point he is making; at least, it is not in the bibliography to Earman’s essay. Bergmann develops the same point in a rather different context in his essay on “The Ontology of Edmund Husserl,” in his Logic and Reality (Madison,, WI: University of Wisconsin Press, 1964), pp. 193-225. Earman misses this discussion too. 34.I have previously developed the following criticism of these philosophers in F. Wilson, Laws and Other Worlds (Dordrecht, The Netherlands: D. Reidel, 1986). See also F. Wilson, Logic and the Methodology of Science in Early Modern Thought, Study Two.

489

35.See C. D. Broad, Examination of McTaggart’s Philosophy, 3 vols. (Cambridge: Cambridge University Press, 1933), vol. I; and J. M. E. McTaggart, The Nature of Existence, 2 vol. (Cambridge: Cambridge University Press, 1921), vol. I. 36. Neither Dretske nor Earman seem to have searched the literature as thoroughly as good scholars might. 37. John Stuart Mill, System of Logic, Bk. II, Chapter ii, § 2; volumes 7 and 8 in his Collected Works, ed. J. Robson (Toronto: University of Toronto Press, 1974), vol. 7, p. 174. 38. Broad, p. 223; McTaggart, p. 275. 39. Broad, pp. 223-4. 40. Broad, p. 227; cf. McTaggart, p. 275. 41.Dretske, “Laws of Nature,” p. 227. 42.David Hume, Treatise of Human Nature, ed. L. A. Selby-Bigge, second edition ed. P. Nidditch (Oxford: Oxford University Press, 1978). 43.J. Barnes, The Presocratic Philosophers, vol. 2, p. 109, notes, with Hume, that volition, willing, though the model for Aristotelian teleological explanations, is in fact but a special case of Humean regularities. 44.Hume, Treatise, p. 159. 45. Hume, Treatise, p. 172. 46.L. Wittgenstein, Tractatus Logico-Philosophicus, trans. D. F. Pears and B. McGuiness, with Intro. by Bertrand Russell (London: Routledge and Kegan Paul, 1961). References are by numbered propositions. See also P. Butchvarov, “Bergmann and Wittgenstein on Generalitty,” for references to, and comments on, this and other parts of the Wittgenstein corpus. 47.The emphases throughout are Wittgenstein’s. 48.Bergmann, “Logical Positivism,” p. 2. 49.Bergmann, “Logical Positivism,” p. 14. 50.Cf. Bergmann, “Logical Positivism, Language, and the Reconstruction of Metaphysics,” in his Metaphysics of Logical Positivism, pp. 30-77, at p. 50; and “The Revolt against Logical Atomism,” in his Meaning and Existence (Madison, WI: University of Wisconsin Press, 1960), pp.39-72, at p. 57.

490

51.Bergmann, “The Revolt against Logical Atomism,” p. 44. 52.Bergmann, “The Revolt against Logical Atomism,” p. 45. 53. McTaggart, p. 274. 54. McTaggart, p. 239. 55.They do receive a definition of sorts in Lewis Carroll’s poem on “The Hunting of the Snark,” but it is only what one calls an “implicit definition,” which is no definition at all – we learn some syntax for these terms but no semantics: they remain without any reference in terms of what we are acquainted with, and are therefore really and simply meaningless. See the study “Implicit Definition Once Again,” below. But the poem is a good poem: Lewis Carroll knew that a poem could use meaningless concepts, defined only implicitly, and still be a good poem, where a description or linguistic picture of the world cannot use such a concept and still be meaningful. The imagination can fill things in in a poem; that is not permitted in a description of the world. 56.Brown, Laboratory of the Mind, p. 83. 57.Armstrong, What Is a Law of Nature? p. 104. 58.For more on this notion, see G. Bergmann, “Outline of an Empiricist Philosophy of Physics,” American Journal of Physics, 11 (1943), pp. 248-258, pp335-342. (This was reprinted in H. Feigl and M. Bordbeck, eds., Readings in the Philosophy of Science [New York: Appleton-Century-Crofts, 1953].) See also F. Wilson, “Empiricism and the Epistemology of Instruments,” The Monist, 78 (1995), pp. 207-229. 59.This example, interestingly enough, is from Hume; see his Treatise, Bk. I, part iii, sec. 12, p. 132ff. It should make interpreters of Hume hesitate before they read him as a phenomenalist or a subjectivist or a sceptic. For discussion of these points in detail, see F. Wilson, The External World and Our Knowledge of It: Hume’s Critical Realism: An Exposition and a Defence (Toronto: University of Toronto Press, forthcoming). 60.For a discussion of these inferences in the context of empiricism, see F. Wilson, Hume’s Defence of Causal Inference, Ch. 1. 61.Those criticized by Wilson in his Explanation, Causation and Deduction, sec. 3.6, were equally unforthcoming with regard to our knowledge of the nomological connective which, in spite of that lack of knowledge, they nonetheless insist must exist. Credo ut intelligam. 62.Broad, Examination of McTaggart’s Philosophy, vol. I, p. 225. 63.Broad, Examination of McTaggart’s Philosophy, vol. I, p. 225.

491

64.As Bergmann argues in his essay on “Synthetic A Priori,” in his Logic and Reality. But Bergmann does insist that such statements are synthetic in a way that is fully compatible with logical atomism and with a Humean account of causation. 65.C. D. Broad, Mind and Its Place in Nature (New York: Harcourt, Brace, 1925), p. 102. 66.See Hume, Treatise, Bk II, part ii, sec. 1. Also Hume, Enquiry concerning the Human Understanding, in D. Hume, Enquiries concerning Human Understanding and concerning the Principles of Morals, ed. L. A. Selby-Bigge, 3rd edition, revised, P. H. Nidditch (Oxford : Oxford University Press, 1975), Sec. VII, part i. 67.Broad, Mind and Its Place in Nature, p. 102. 68.For these points, see my earlier discussion in F. Wilson, Explanation, Causation and Deduction (Dordecht, The Netherlands: D. Reidel, 1985), pp. 293ff. 69.That is, characteristics considered as universals but not necessarily as (Platonic) forms – though there are some defenders of the idea that nomological necessity derives from a relation among characteristics who take those characteristics to be universals in the sense of (Platonic) forms. Thus, Brown, Laboratory of the Mind, apparently holds that causal necessity derives from a relation among properties or characteristics which he takes to be Platonic forms; and Brown follows Broad who made the same proposal in his Examination of McTaggart’s Philosophy. The difference is of course simple. One can, like Bergmann (and like G. E. Moore), take characteristics to be universals and also hold that such universals are given to one in sense experience; but if they are taken to be Platonic forms then they are taken to be entities that are not given in sense, but only in some sort of rational or (as Broad put it) non-perceptual intuition. 70.As Bergmann would put it, the concepts that science uses must be significant, in the sense that they occur in statements of laws of nature; see G. Bergmann, Philosophy of Science (Madison, WI: University of Wisconsin Press, 1959), Chapter One. Laws determine which concepts are significant, and therefore which concepts are natural kinds. It is not the case that there are natural kinds, where this is a property of properties, distinguishing them from other properties, and such that if they appear in a generalization then that generalization is not a mere regularity but a law of nature. Laws determine which kinds are natural kinds; there are no special kind of kinds such that these kinds determine which generalizations are laws. That at least, is what is implied by the Humean account of laws. To suppose that there are natural kinds are a special sort of kind that determines which regularities are laws is to fall into a form of Aristotelianism 71.Bergmann, in his posthumous New Foundations of Ontology, offers a revised account of general facts.

492

Here Bergmann wishes to rid his ideal language of variables, since they cannot, he suggests, reasonably be held to represent anything existent, even by the ontologizer who aims to create the most luxuriant of worlds. Now, he proposes, the fact that all F’s are G’s is “built” by the function w “not just from one argument but, indifferently, from an indefinite number of alternative arguments ...from , from , and , and so on.” (p. 235, I have modified Bergmann’s notation slightly) The function w thus effects a many-one mapping from the pairs or 2-tuples , etc., onto the general fact. The ontological analysis of the general fact is not “(x)(Fx e Gx)” but rather w[a, Fa e Ga] or any of its “variants” such as w[b, Fb e Gb]. Bergmann suggests that “Each [of these variants] is an alternative assay of the one and same general fact.” (p. 232) The mind, in having before it, the general fact also has before it the sentence ‘(x)(Fx e Gx)’ as its text. In the text there is no hint, or perhaps only a hint, of the multiplicity of assays. In thinking the general fact one also thinks the text. As he puts it, “One cannot believe, or doubt, or remember, and so on, any generality without perceiving the appropriate words.” (p. 204, his italics) This part of the more general thesis that “all awarenesses, except primary Perceivings and Imaginings (and undoubtedly some ‘Feelings’), are inseparable from their texts.” (p. 234) All this is important for what Bergmann is about in New Foundations. One of these things is to get rid of the variables. For what is conventionally represented by ‘(x)(Fx)’, the singular fact Fa won’t do rather than the 2-tuple because “w Fa” does not make it explicit which constituent of the atomic fact it is that the quantifier operates upon (which is the job customarily done by the quantified variable). The quantifier must both retain the atomic fact and also recognize as it were the constituent being generalized over. So the generality operator w must operate on both, that is on the 2-tuple . (p. 168) But all this makes clear that the general fact does, still, satisfy what we have called the four crucial features that something must fulfill if it is to count as an irreducible causal tie. Indeed, the new assay makes it even more clear than before that the general fact satisfies these four crucial features. So the changes in Bergmann’s assay of general facts do not affect what we are about. 72.Bergmann, “Generality and Existence,” pp. 69-70; Russell, “The Philosophy of Logical Atomism,” p. 236. 73.Russell, “The Philosophy of Logical Atomism,” p. 236. 74.Actually, the Tractatus uses the Greek letter ‘>’ and then the same letter with a bar over it. I couldn’t duplicate the latter with my word processor, so I made a substitution which does the required job. 75.I have slightly altered the translation. Pears and McGuinness have the translation, “A proposition is a truth-function of elementary propositions.” This English translation is slightly misleading. The German is “Der Satz ...” which has the meaning of “the proposition ...”, that is, “any proposition...” and not simply “a proposition...”, as the translators would

493 have it. Compare the phrase “The whale” in “The whale is a mammal”: the latter is understood as “any whale is a mammal” or “all whales are mammals”. 76.Russell, Inquiry, p. 94. 77.Russell, Inquiry, p. 95. 78.Russell, Inquiry, p. 320. 79.Ibid. 80.Russell, Inquiry, pp. 319-320. 81.H. Hochberg, “Negation and Generality,” in his Logic, Ontology and Language: Essays on Truth and Reality (Munich: Philosophia Verlag, 1984), pp. 296-312, at p. 303ff. 82.This latter Russell is of course Russell at an earlier stage. 83.Bergmann, “Generality and Existence,” p. 70. 84.Ibid. 85.Bergmann, “Generality and Existence,” p. 71. 86.Ibid. 87 Just as we perceive negation when we perceive (as Sartre suggested we do) that someone is not in the room, even though negation is not an entity in the world; see the discussion in F. Wilson, “Grossmann on the Categorial Structure of the World,” in this volume, above. 88.Bergmann, New Foundations of Ontology, p. 234. 89.Bergmann, New Foundations, p. 219. 90.Bergmann, New Foundations, p. 61. 91.See the study below, “Effability, Ontology and Method.” 92.See for example, Bergmann, “Professor Ayer’s Analysis of Knowing,” in his Metaphysics of Logical Positivism, pp. 214-227. 93.See G. Bergmann, “Realistic Postscript,” in his Logic and Reality, 302-340; and so “Acts,” ibid., pp. 3-44. For a critical discussion of this journey that Bergmann made from logic as linguistic to logical realism, see H. Hochberg, The Positivist and the Ontologist: Bergmann, Carnap, and Logical Realism. See also W. Heald’s Introduction to Bergmann’s New Foundations of Ontology. 94.Cf. the study above, “Grossmann on the Categorial Structure of the World.”

494

95.Cf. W. Sellars, “Some Reflections on Language Games,” in his Science, Perception and Reality (London: Routledge, 1963), pp. 321-358. See also the studies below, “Marras on Sellars on Thought and Language,” and “The Aboutness of Thought.” 97.Russell makes more or less the same point in his Inquiry into Meaning and Truth: what is important about a general proposition is that it gives a prototypical form for statements of individual fact. “When we judge ‘A is a man but not mortal’, we accept ‘A is a man’ but we reject ‘A is mortal.’ The various acts of this kind, putting B, C, etc. in place of A, all have something in common; what they have in common is a belief expressed by the words ‘some man is not mortal.’ When we reject this belief, we are in a state expressed by the words ‘all men are mortal’.” (Inquiry, pp. 319-320) And again, Russell remarks that, when I believe that everything is F, then I believe that Fx is true for all values of x. a may be such a value, but I may never have heard of a. Thus, “the belief that [Fa] is one belief, and the generality is part of that belief. Moreover, it is intensional in the sense that I can have the belief without knowing all [individuals] there are. As soon as I understand the [word F], [and] the subject-predicate form, ... I have everything, except generality, that is required for understanding [(x)(Fx)].” (Inquiry, p. 317) 98.As Russell was to put it in his Inquiry into Meaning and Truth (p. 319), the elementary propositions form a sort of object language describing the world as we experience it, and the general proposition is as it were meta-linguistic, in a language that comments on that object language. 99.Let us agree here that we will not worry here about the meaning of negation, that is, about whether there are negative facts as well as positive facts. For this issue, see the study above, “Grossmann on the Categorial Structure of the World.” 100. John Stuart Mill, System of Logic, Bk. II, Ch. iii, § 3, in vol. 7 of his Collected Works, ed. J. Robson (Toronto: University of Toronto Press, 1974), p. 186. 101.I shouldn’t be taken, when I say this, that I am saying that the alternative account of intentionality that we have proposed is exactly what Bergmann had in mind when he first defended a “positivistic metaphysics of consciousness”: Bergmann’s suggestion was different, clearly so, from our alternative. See G. Bergmann, “A Positivistic Metaphysics of Consciousness,” Mind, n.s. 54 (1945), pp. 193-226. It must be said that Bergmann is not entirely pellucid in his account of meaning or intentionality. For a clear discussion of Bergmann’s development, see H. Hochberg, The Positivist and the Ontologist. 102.As Kirk and Raven, The Presocratic Philosophers, note, “Anaxagoras in fact is striving, as had several of his predecessors, to imagine and describe a truly incorporeal entity” (p. 374). 103.Hume, Treatise, p. 252.

495

104.G. E. Moore, “The Refutation of Idealism,” in his Philosophical Papers (London: Kegan Paul, Trench, Trubner and Co., 1922). For comment on Moore’s argument and on its continuing importance, see the study above, “Moore’s Refutation of Idealism.” 105.For the latter, see Bergmann’s essay, “Ideology,” in his Metaphysics of Logical Positivism, pp. 300-326. 106.Hume, Treatise, p. 155. 107.See G. Bergmann, Philosophy of Science (Madison, WI: University of Wisconsin Press, 1956). 108.See R. Carnap, “Testability and Meaning,” Philosophy of Science, 3 (1936), pp. 419471, and 4 (1937), pp. 1-40; and also C. G. Hempel, “The Concept of Cognitive Significance,” Proceedings of the American Academy of Arts and Sciences, 80 (1951), pp. 61-77. 109.Cf. A. Pap, Semantics and Necessary Truth; An Inquiry into the Foundations of Analytic Philosophy, with a foreword by Brand Blanshard (New Haven: Yale University Press, 1966). 110.The connection between laws and contrary-to-fact conditionals was emphasized by R. Chisholm, “Law Statements and Counterfactual Inference,” Analysis, 15 (1955), pp. 97-105. 111.Bergmann, “Comments on Professor Hempel’s ‘The Concept of Cognitive Significance’,” in his Metaphysics of Logical Positivism, pp. 255-267, at pp. 262-263. 112.Hume, Treatise, p. 172. 113. Derrida does not see this while Hume did, as we argued above in the study “Hume and Derrida on Language and Meaning.”

Thirteen Human Action and a Natural Science of Human Being* The philosophy of action is now perhaps less active than it was a few years ago, but it is of continuing interest. Initially, this interest arose out of a philosophical attempt to establish the impossibility of a science of human being, that is, a science in the same way in which the natural sciences such as physics are sciences. The idea was that descriptions of intentional actions could not be reduced to descriptions of bodily movements, and that therefore there are aspects of human behaviour which prevent one from having a natural science of human being. Such was the argument offered by such writers as A. I. Melden and Charles Taylor, it took its lead from the writings of Peter Winch on the social sciences which in turn took its inspiration from the work of the later Wittgenstein. One can argue that the work of the latter is less opposed to the idea of a science of human being than Winch and others have thought, but his writings are sufficiently ambiguous and often sufficiently obscure to make the issue at lest debatable. In any case, often coupled with the idea of the simplicity or irreducibility of intentional behaviour was some theory about how certain actions could be complex (e.g. turning on the light by flicking the switch). Some of these latter theories, e.g. Goldman’s, are so ontologically rich as to cause philosophical indigestion while others, e.g. Davidson’s, are so nominalistic and ontologically spare as to cause philosophical starvation. The work on the theory of action of R. Tuomela1 must be placed in this sort of context. In his study he has developed a theory of action, including complex action which is reasonable in its ontology and scientific in spirit. Indeed, Tuomela shows – as conclusively, in my opinion, as these things can be shown – that a reasonable theory of human action can be developed within the framework of a natural science of psychology, and that the fact that human action is intentional has no anti-scientific implications. It is, to be sure, a point that such * Originally appeared as a Review of R. Tuomela, Human Action and Its Explanation, Dialogue, 21 (1982), pp. 571-577

498 psychologists as E. C. Tolman argued some eight decades ago, but it bears repeating – there are always those ready to argue the uniqueness of human being and its being above the natural order – those, in other words, who yearn for the remnants at least, of the human soul – , but it especially bears repeating now in the light both of the above-mentioned philosophical arguments about the theory of action and of more recent developments in the philosophy of science. At the heart of Tuomela’s discussion is a theory of the nature of cognitive mental events and processes. Central to this theory are certain inferences from observed behaviour to unobserved entities. These inferences, Tuomela argues, are based on observed fact, that is, observed human behaviour and on dispositions to behave, but the general framework for such inferences is justified, he also argues, by his reliance upon a deductivenomological account of explanation and a theory of psychological concept formation in the spirit of scientific realism. The whole is redolent of the thought of Wilfrid Sellars. In detail there is a great deal of reliance upon work that the author and others, such as J. Hintikka, have published elsewhere. In fact, the latter sort of reliance makes Tuomela’s study somewhat idiosyncratic, and less self-contained than it might reasonably be. Which is too bad, since its main thrust in no way depends upon the peculiarities of his own views or those of his favourite colleagues. In what follows I shall sketch the outlines of Tuomela’s views, sufficiently enough, I hope, to show their general acceptability. But I shall also try to show that certain limitations, or rather oversights, in his theory of concept-formation lead Tuomela into certain, in a sense relatively minor, infelicities, not so much with regard to actions but with regard to the mental states that cause and explain actions. Crucial to Tuomela’s programme is the point that behaviour is often classified in terms of its function, that is, in terms of its effects.2 Consider the biologically protective withdrawal response, say to loud, sudden, sharp noises, that James, Tolman and others took to be the characteristic fear response. To say a movement is protective is to classify it in terms of its consequences: a protective movement is one that in the context tends to lessen the probability of harm. In the primary instance, the harm is biological, and protection is secured, basically, by withdrawing from the stimulus. But of course, that stimulus, the noise say, does not always elicit a noiseprotective movement. What the noise elicits, more precisely, is the disposition to respond noise-protectively, and the actual movement is elicited only if some further stimulus is present. Thinking in terms of the old formula

499 R = f(S) we expect to be able to give, in terms of the stimulus, necessary and sufficient conditions for the response. In our simple example, then, to say that the noise is the stimulus that elicits the tendency to respond noiseprotectively is to assert a lawful regularity – call it (A) – asserting something to the effect that, for all persons and a stimulus of the relevant sort, that stimulus is necessary and sufficient for that dispositional response.3 Suppose we know that Beverley is now disposed to react noiseprotectively. For example, we might know this by seeing the disposition actualized, by seeing that Beverley is reacting noise-protectively (which, one should note, is by virtue of the functional classification to know a law that applies to the movement that Beverley actually makes, a law describing the effects of that sort of movement in that sort of context). From this knowledge that Beverley is now disposed to react noise-protectively we can use the law (A) to explain in a deductive-nomological fashion the response, that is, the disposition to react noise-protectively. That is, knowing the effect, the response, one can use (A) to infer the cause, the stimulus; and one can then use this knowledge of the presence of the cause, the stimulus, to explain ex post facto but deductive-nomologically , here again via (A), the presence of the effect, the response.4 Most fear reactions, unlike that to a loud noise, are learned. Consider James’ example (for which he claimed no originality) of the baby who learns to fear burning candles when he or she burns his or her hand as he or she, attracted to the dancing light of the burning candle, reaches out to that light and receives a painful burn on the hand. The presence of the candle flame, unlike the presence of the loud sharp noise, it is not necessary and sufficient for the presence of the withdrawal tendency that constitutes the candle-protective response; rather, it is necessary and sufficient if and only if the infant has previously been burned by the candle. This also is a law – let us call it (B). The concept which appears in it, that of being previously candle-burned, is historical in the sense that it is a property exemplified in the present but defined in terms of past states. If we confirm the regularity of Beverley that he or she tends to react candle-protectively when he or she sees candle flames, if we discover this pattern in his or her behaviour, then we can, via (B), infer that he or she has previously has undergone previously the relevant learning experience. We can then explain the patterning of Beverley’s behaviour in terms of the learning. And, of course, if we observe the learning then this knowledge will enable us to predict, via (B), the future patterning of Beverley’s behaviour. Furthermore, if we limit ourselves to persons who have previously

500 been burnt, then it will be a lawful (though conditioned to these persons) regularity (B*) that for all such persons and times, seeing a candle will be necessary and sufficient to elicit as a response the tendency to candleprotective behaviour. We can use (B*) to explain the behaviour of persons to which it applies. Such explanations, however, insofar as they are based on a conditional regularity, will be more gappy or imperfect than those based on the unconditional regularity (B). The latter does not, of course, make them any the less explanations.5 Now, both the laws (A) and (B) are gappy, in the sense that they leave out details of the relevant processes. In (A) the noise is mentioned and its proximity. But the details of the process by which the sound is transmitted from the source to the ear are omitted. Taking them into account, we come to distinguish the noise itself, the distal stimulus, from the effects it has on the sense organ, the proximate stimulus. Moreover, we have reason to believe that the proximate stimulus, through a physiological process, brings about a fairly specific neurological state in the brain, this brain state being an underlying necessary and sufficient condition for the tendency, elicited in the first instance by the noise and then by the proximate stimulus, to react with noise-protective behaviour. If “Nxt” represents that x is in this necessary and sufficient neurological state N at time t, then the relevant law is that (C) for all persons and all times t, Nxt is necessary and sufficient for the presence in x of the tendency to behave noiseprotectively. But this is misleading. We do not know a law like (C) mentioning a specific neurological state. Rather, what we have reason to believe is true is that there is a law of the sort (C), a law of that general logical form. That is, what we have reason to believe is the existential hypothesis (C*) that there is a unique g such that g is a kind of brain state and for all persons x and all times t, gxt is necessary and sufficient for the presence in x of the tendency to behave noise-protectively. This law justifies our introducing a theoretical concept or hypothetical construct referring to the neurological state it claims to exist. We let, say, the term “N^” refer to the state that (C*) asserts to exist. This permits us to introduce as equivalent to (C*), the law (C**) that for all persons x and all times t, N^xt is necessary and sufficient for the presence in x of the tendency to behave noiseprotectively. [The term ‘N^” is of course a definite description; the logic of such terms was laid out by Bertrand Russell many years ago.] Given (C**), the object of research will be to identify the type of state to which the term “N^” refers. The result of such research will be an identificatory hypothesis of the sort (C^) that N = N^. The law (C) follows from (C^).

501 If we now know that Beverley is at present disposed to act noiseprotectively, then we can use (C**) to infer that that neurological state N^ is now present in Beverley. We can then turn about and use this to explain, again via (C**), and in deductive-nomological fashion, the presence of the disposition. This latter sort of explanation, while certainly not tautologous (the need for the factual existential hypothesis shows that it is not in this way vacuous6), is always ex post facto – of at least it is so until an identificatory hypothesis (C^) can be confirmed, the hypothetical construct “N^” replaced by a term “N” referring to its specific neurological embodiment, and (C**) replaced by (C). But for all the ex post facto nature of any explanation of the presence of the disposition in terms of the presence of N^, it does not fail to be an explanation – provided one’s philosophy of science permits the introduction of hypothetical constructs like N^, and their use in laws like (C**). In a somewhat similar fashion, hypothetical constructs can be introduced to represent the effects in the nervous system of learning experiences – the traces in the brain of earlier learning experiences and earlier stimuli. For a period of time, psychologists eschewed on methodological grounds constructs of the sort N^. Technically speaking, there is nothing terribly problematic about them. In the simplest case, such constructs are definite descriptions. That these are successful, that is, that they actually refer to entities or to kinds of entities, is guaranteed by existential hypotheses of the sort (C*). The latter are generic, and involve quantification over species, that is, sorts events and states. The logical apparatus is that of second or even higher order logic, but there is no special difficulty in that. What certain psychologists argued was that there was little reason to expect fruitful results from research guided by existential hypotheses like (C*). Skinner, especially, so argued, and Bergmann and Spence did too, though in a much more sober fashion. The disagreement was an old one. David Hartley made extensive reference to such hypothetical constructs in his Observations on Man – the earliest book that can recognizably be identified as a work in (the science of) psychology. Joseph Priestly thought that was rather useless and removed it in his edition of Hartley (he removed the theology, too), leaving the world with a textbook on associationism. James Mill, in his Analysis of the Phaenomena of the Human Mind, followed Priestly’s Hartley in reckoning physiological constructs and speculation were of little use in the learning theory of associationism. Herbert Spencer, in his Principles of Psychology, reckoned otherwise, and physiology reentered psychology once again. And so on into the twentieth century, with

502 D. O. Hebb championing such constructs and Skinner arguing their uselessness. Sometimes their champions were correct, sometimes those who dismissed them: it was all a matter of just how useful such speculation could be at each stage of the development of the science – sometimes it was and sometimes it wasn’t. But generally more ink than was necessary was lost in these disagreements. And often things got mixed up by the introduction of spurious metaphysical concerns. Thus, terms like ‘behaviourist’ were (and are) sometimes used as terms of praise and sometimes as terms of abuse. But everyone – those, anyway, who are committed to developing a science of human being along the lines of a natural science like physics – everyone who was so committed recognised, and recognizes, that laws like (A) are gappy and that the variables intervening between the stimulus and the response will have ultimately to be explored. The disagreement was, or is, on research prospects: is it fruitful now to pursue this line of research? Tuomela aims not so much at research into the laws of psychology but to develop an account of human action that can show “how possibly” our commonsense framework can be fit into a psychology that aims to be scientific. For this task, allusion to hypothetical constructs cannot be avoided. It is well, therefore, that his philosophy of science is rich enough to permit the introduction of hypothetical constructs referring to the (as-yetunlocated) intervening variables – this is his “spirit of scientific realism” – and he also recognizes that the ultimate justification of these of these constructs depends upon the progress of science in identifying the specific embodiments of the constructs he proposes. He indicates certain trends in contemporary, and in particular in cognitive psychology that suggest it might now be reasonable once again to pursue research guided by a theory involving hypothetical constructs referring to as-yet-unidentified physiological states intervening between stimulus and response. Particularly if one can speculate that these structures are analogous in various ways to computer models of learning. Linguistic items are classified in terms of their functions. The meaning of a sign is a matter of its linguistic role, and two signs are linguistically the same, that is, have the same meaning, just in case that they play the same linguistic role, where, following Sellars,7 Tuomela interprets the idea of a linguistic role sufficiently broadly as to include all the regularities, capacities and tendencies that describe the various world-word, wordword, and word-world transitions. Linguistic behaviour is pattern-governed behaviour – of which James’ infant’s learned response provides the simple,

503 but not misleading model. Underlying these patterns are (hypothesized) neurological states, or, rather, complexes of such states. Important aspects of the linguistic roles that items can play are the word-world connections. (Some) sentences of a language describe the world. Among such sentences are some describing behaviour. If a sentence of the latter sort is given a special form – if it is an imperative – then its function is to bring about or cause the behaviour it describes. The sentence ‘Beverley does X’ describes the fact that Beverley does or is doing X, that is, is performing the action of bringing it about that X. That is just description. But if I utter that sentence with imperative force, “You [ = Beverley], do X”, i.e. utter something like “Beverley make it be the case that Beverley does X,” then this saying it in the imperative mood causes it to be the case that Beverley does X. A statement of intention is a self-directed imperative or at least functions like one: my saying “I shall [vs. will] do X”, other things being equal, tends to bring about the behaviour it describes. Expressing my intention to do X is causally relevant to my actually doing X; the sentence having that function relative to my behaviour is precisely what distinguishes it from a prediction about my behaviour, which merely describes my behaviour but does not in itself causally function to bring about the behaviour that it describes. But of course it is not quite so simple. It is not so much Beverley’ actually asserting “I shall do X” that tends to bring about Beverley’s doing X; rather, for the latter to come about, it is sufficient that there be present in Beverley the tendency to assert the intention to do it. It is the present disposition (d) if Beverley is questioned, then Beverley utters imperatively an ‘I shall do X’ that tends to bring about Beverley’s doing X. Here, we of course assume when speaking of Beverley uttering imperatively an ‘I shall do X’ we are saying not just that Beverley utters those noises but we are also classifying that item as used on that occasion functionally, in terms of its linguistic role; that is, we are taking it for granted that the item is taken as having a certain meaning. If we now make the usual assumptions about persons who have learned the language and the usual existential hypotheses about the underlying neurological states, then we can assert that there is an underlying state, call it (d^) that is lawfully connected to that state according to the regularity (E) that the presence of (d^) in Beverley is necessary and sufficient for the presence of the disposition (d). We then have the following causal tie between the brain state and the action: (F) for all (such) persons x and all (such) times t, if d^xt then, and only then, Beverley does X. On the basis of (F) we can say that the hypothetical construct (d^) refers to the brain state that is ex-

504 pressed by the person’s saying “I shall do X”. At least, that is a reasonable meaning to assign to the idea of language being used so as to “express” an inner state. Then, using (F) we can explain the behaviour of the individual in terms of the inner state that is expressed by the behaviour. Tuomela skilfully defends this as the basic pattern of purposive causation. The model is elaborated in various directions. Besides inner states of intention, we want inner states of belief. These will be the hypothesized inner neurological states connected on the model of (E) with dispositions to assert sentences. These sentences describe states of affairs in the world; in the case of the simplest sentences, those of atomic form, the sentences represent or picture the states of affairs they describe. This representative or picturing function is secured by the world-word connections of the language. In this case the sentences are used not imperatively but assertively – not in the imperative mood but in the indicative mood. Sentences asserted in the indicative mood on the one hand express an inner neurological state of belief but also function to bring about in others a belief of the same sort. Asserting an imperative brings about or causes actions, asserting an indicative brings about or causes a belief.8 This apparatus enables one to deal with more complex practical syllogisms9: I shall achieve X Doing Y is sufficient for X therefore, I shall do Y therefore, straightway, Y is done by the speaker The two premises are taken as actually or potentially used to express hypothetical inner states of intention and belief, respectively. The first conclusion, too, is so understood. The connection between the two premises so understood and the first conclusion is causal. It is a true generality about human behaviour that “When one wills the end, one wills the means, so far as one recognizes the latter”; it describes a regular connection in human thought. To be sure, it is by no means exceptionless; it is, like most of our knowledge of human behaviour, gappy. Still, that does not prevent it being used, under normal conditions, to give explanations of human thought processes.10 However, we can say that, even if the generality does not hold strictly universally, it ought to: it describes how thought ought to be. Understood this way, as an ought-to-be statement, it functions as a rule that causes us (often) to adjust our own behaviour so as to conform to it, and

505 causes us to train or to raise others, especially new language users (children), that their thought will conform to the rule. Qua rule, the generality tends to bring it about that exceptions to n will be minimized.11 As for the second conclusion in the practical syllogism above, it is neither an inner state nor anything essentially linguistic. It is, rather, an act or activity, a non-linguistic behavioural response. The connection between the first conclusion and the second, is again causal, again gappy and with exceptions, and again a regularity that ought to hold without exceptions. As for the action Y, provided that the means-end belief is true, this action will have the function of bringing about (the end) X. So we can functionally classify the action also as a Y-bringing-about-X action. In this way we come to an account of complex actions. Y itself may be a functional description of an action. But one will eventually get to a description of responses simply in terms of movement or ranges of movements – e.g. raising one’s arm – or even: one’s arm going up. These descriptions will be descriptions of basic actions. A sympathetic and extensive treatment of the theory of action along these straight-forwardly scientific lines can be found in Tuomela’s study. It is worked out in sufficient detail for us to see how it can accommodate, in principle at least, the complex sort of planning that enter into so much of human action. His treatment makes perfectly pellucid how much of the hermeneutical tradition in the social sciences can be incorporated without serious distortion into the empiricist-behaviourist tradition. And he shows how to make good scientific sense out of so much of what is the exasperatingly opaque and exasperatingly metaphysical about basic actions and, in the case of complex actions, about how more complex actions are “generated” out of the less complex. What about mental states? Events of this sort undoubtedly exist: we experience them, we know their effects on our own behaviour, and we cite them as we explain the behaviour of others. Tuomela does not avoid mentioning them: he is no metaphysical behaviourist. His discussion is instructive. Like many, perhaps most, philosophers who take science and its methods as giving us what knowledge we have of the world around us, and of ourselves as part of that world, Tuomela treats mental events as events that are at least parallel to, and perhaps even (contingently) identical with neurological states. The minimal parallelistic assumption for, to continue with our example, the mental state of fearing the noise that has just in-

506 truded itself into one’s consciousness, where “Mxt” represents x being in this mental state at t, would be given by the regularity (G) that for all persons x and all times t, Mxt if and only if d^xt. There would be a similar parallelistic connection for every sort of mental state. Such a parallelism has been part of scientific commonsense since the time of the Cartesians and has been the basic framework for psychology since David Hartley’s Observations on Man. Whether the connection is, as (G) suggests, simply a regularity, or involves a stronger tie of (contingent) identity (if there be such a thing), is more a dispute in ontology than in the philosophy of psychology. Suffice it to say, one, that even if the connection is an identity, it would be a contingent one, rather than literal sameness12; and, two, any connection of identity would logically entail a regular connection of the sort (G). For most purposes one can get on in the philosophy of psychology with no need to push deeper than the recognition of the factual and parallelistic relation between mental states and brain states. In particular, once the parallelistic connection is noted then it is clear that there is no need to mention mental states when explaining human behaviour.13 In explaining and predicting human behaviour, four sorts of variables are relevant: behavioural, environmental, physiological and mental. If parallelism is true, then any reference to a mental state can, by a law like (G), be replaced by reference to a brain state, salva veritate, and without loss of predictive power vis-à-vis behaviour. Parallelism ensures the compatibility of the claim that an objective natural science is human being is possible with the claim that mental states exist and are causally relevant to what we do and say. Now, mental states are, we all know, private. That means thirdperson reference to them must always be by definite description. To that extent they are like the hypothetical constructs that we use to refer to the brain states underlying our behavioural dispositions. But to that extent alone. Consider the sort of brain state we called “N^”, the brain state that underlies the tendency to react noise-protectively. The introduction of the concept referring to this brain state is justified by the existential hypothesis (C*). This hypothesis asserts the existence of a kind of brain state. It is specified indirectly as the kind that causes the tendency to noise-protective behaviour. Because it is thus specified we can assert of, say, Beverley, when he or she is tending to react noise-protectively, that there is present in him or her a brain state of the sort that causes a reaction of that kind. The latter assertion leaves indeterminate not only the location of the brain state (if it actually is localized – an old issue we may safely ignore) but also the

507 specific properties of the brain state: the latter, too, are referred to indirectly, by means of a definite description – a definite description of a type or sort and not a definite description of an individual thing. This is necessary because in point of fact we do not know – specifically – what sort of brain state it is that causes the relevant behaviour. Science must still discover which identifying hypothesis of the sort (C^) is true; only then will we know specifically the sort of brain state that causes behaviour of that kind. In the case of mental states, however, we do know specifically what kinds cause what behaviour or rather behavioural tendencies. To be sure, this is not because the other fellow’s mental states are presented to us – they aren’t, of course, they are after all private. But from our own case we know what kind of mental state causes what kind of behaviour. Using these concepts of specific sorts of mental states, we can form definite descriptions that refer to the other fellow’s mental states. The parallelistic laws of the sort (G) establish that these definite descriptions will be successful.14 If this is so, then I think that one of the claims Tuomela makes in this otherwise careful study must be wrong. This is a thesis that he has about the nature of mental states, his Sellarsian thesis about their meaning or intentionality. Tuomela construes mental states as analogous to linguistic entities in the following sense: the meaningfulness of mental states, their “aboutness”, what Brentano called their intentionality, is to be understood in a fashion exactly parallel to the behaviouristic analysis of linguistic meaningfulness in terms of linguistic roles. The meaningfulness of mental states thus turns out to be a dispositional property of items which are intrinsically non-meaningful. Tuomela thus adopts Titchener’s theory about the meaningfulness of mental states. On this view, one cannot know the meaning of a mental state unless one recognizes the presence of a dispositional property. However, the research of the Würzburg psychologists, and others, showed that it is impossible on empirical grounds to maintain that when we know the meaning of a mental state we recognize the presence of a dispositional property.15 Rather, the meaning is manifest: we know it immediately and not inferentially as would be the case if the meaning were simply dispositional.16 It is like seeing sugar dissolving, something manifest, and knowing it is soluble, which is a dispositional property, that is, one the presence of which is known only its being actualized, that is, by being given a manifest property, and otherwise only by inference. The meaningfulness of mental states may be analogous to the meaning of items in overt speech, i.e.,something dispositional, but we know the presence of such a

508 disposition through the presence of a manifest property, a manifest property that is wholly present in our conscious experience. That is, the meaning is carried in the case of mental events by the presence in consciousness a non-sensory imageless thought.17 There is little doubt, I expect, that Tuomela would argue that this is to infer from the way things are presented to us in experience to the way they ontologically must be, and would them go on to argue that such an inference is fallacious, embodying what he, following Sellars, calls the “myth of the given.” I do not think that this response will save his position, however. There is a difference between sugar dissolving and sugar being soluble. The former is a manifest, the latter a dispositional property of the sugar, and no matter how one comes to construe ontologically the manifest properties of things, one will not be able to obliterate this distinction between what is manifest and what is dispositional: however one analyzes dissolving, there will always be a difference between this and tending to be this. Behaviourally, meaning is dispositional, and mentally, meaning is manifest, and no amount of talk about the myth of the given is going to eliminate this distinction. Tuomela’s error is that of thinking that this could be otherwise. Having said this, however, it should also be said that Tuomela’s error about mental states hardly affects the main substance of his book. That, to repeat, argues in detail that the theory of action can in detail be given a scientific analysis. The theory of action is derived from our everyday accounts and explanations of human behaviour. Tuomela’s study is a “how possibly” explanation: it shows how possibly a scientific account of human behaviour can fully accommodate the theory of action and our ordinary image of human being of which it is a part. Tuomela thus shows how our ordinary explanations are compatible with the idea of a natural science of human being. Demonstrating this is Tuomela’s no mean achievement. As to the extent that science will confirm or, as Skinner has suggested and as has been suggested by critics of so-called “folk psychology,” will systematically modify and perhaps even invalidate this ordinary and obviously gappy image of human being, this is an issue that Tuomela does not raise, but in any case is an issue that the progress of science alone will decide: it is clearly not an issue to which philosophical debate and analysis is relevant: or, rather, philosophical analysis shows this much, that the issue here is not philosophical but one that science alone can decide.18 Tuomela, by not discussing the issue, recognizes this fact.

509 Endnotes to Study Thirteen

1.R. Tuomela, Human Action and Its Explanation (Dordrecht, The Netherlands: D. Reidel, 1977). 2.Compare G. Bergmann, “Purpose, Function and Scientific Explanation,” Acta Sociologica, 5 (1962), pp. 225-238. See also F. Wilson, Empiricism and Darwin’s Science (Dordrecht, The Netherlands: Kluwer, 1991), Ch. Six. 3.Such a law, given our current limited knowledge, will be “gappy” in the sense of J. L. Mackie, “Causes and Conditions,” American Philosophical Quarterly, 2 (1965), pp. 246-264, reprinted in E. Sosa, ed., Causes and Conditionals (London: Oxford University Press, 1975). Or, what amounts to the same, in the terminology of G. Bergmann, it will be a piece of “imperfect knowledge”; see his Philosophy of Science (Madison, WI: University of Wisconsin Press, 1956), Ch. Two. 4.For detailed discussion of ex post facto explanations of this sort, see F. Wilson, Explanation, Causation and Deduction (Dordrecht, The Netherlands: D. Reidel, 1985). Ex post facto explanations of this sort are common. Think of our inference from the presence of flu symptoms to the presence of the flu bug, and our then explaining the presence of the symptoms by citing the presence of the flu bug. Ex post facto explanations like this are often cited as calling into question the deductive-nomological model of scientific explanation. Explanation, Causation and Deduction defends the reasonableness of such ex post facto explanations and on the basis of this analysis argues that they (and other purported counterexamples of the DN model) do not after all constitute any sort of counterexample to the DN model. See also F. Wilson, The Logic and Methodology of Science and Pseudoscience (Toronto: Canadian Scholars Press, 2000). 5. “Water, when heated, boils” is a gappy law but it can still be useful in figuring out how to boil water for tea, and why this water is boiling in terms of its being heated. For an extended discussion of gappy laws and explanations, see Wilson, Explanation, Causation and Deduction. 6.For exactly the same reason it is not tautologous to assert that “The present mayor of Toronto is a present mayor of Toronto”. This is easily seen given Russell’s account of definite descriptions. For, on this account, “The F is a F” is true if and only if the F exists, that is, is true if and only if it is true that “There is exactly one F” – which is clearly not a tautology. 7.Wilfrid Sellars, “Empiricism and the Philosophy of Mind,” in his Science, Perception, and Reality (London: Routledge and Kegan Paul, 1963), pp. 127-196; “Notes on Intentionality,” in his Philosophical Perspectives (Springfield, IL: Charles C. Thomas, 1967), pp. 308-320; and “Some Reflections on Language Games,” in his Science, Per-

510 ception, and Reality, pp. 321-358. 8.R. M. Hare in his Language of Morals (Oxford: Oxford University Press, 1952) made a distinction between the descriptive part of a sentence, called the “phrastic,” and the mood, called the “neustic.” One might have “That slab is here” as the phrastic, but it can be spoken or used in different moods, with different neustics, as simply an indicative sentence so that the sentence is simply fact-stating, or as an imperative in which case it causes the hearer to act, to bring the slab to the indicated place – or as an optative (“Would that that slab were here”) expressing a wish. The meaning of the phrastic is given in terms of world-word connection, but the meaning of the mood is given by the effects on others or on oneself. [Hare was interested in the imperative mood, as providing the basic account for the moral ‘ought’. The meaning of ‘ought’ is given by its imperative force, its capacity to case oneself and others to do as it is said one ought to do or be. Its meaning is given by its word-world connections rather that the world-word connections that secure the descriptive role of the phrastic. Hare’s reflections have their origin in the emotivist account of ethical language in A. J. Ayer and M. Schlick, where the meaning of ethical terms is a matter of their “emotive meaning”, and is given by word-world connections.] In Wittgenstein’s Tractatus [Tractatus Logico-Philosophicus, Intro. by Bertrand Russell, trans. D. F. Pears and B. F. McGuinness (London: Routledge and Kegan Paul, 1961)], the phrastic is the proposition that pictures states of affairs. “We picture facts to ourselves.” (Proposition 2) The proposition is determined by its syntax, which are word-word connections: “What constitutes a picture is that its elements are related to one another is a determinate way.” (Prop. 2.14) There are also rules of reference or designation, which are world-word connections: “The pictorial relationship consists of the correlations of the picture’s elements with things.” (Prop. 2.1514) These correlations establish an isomorphism between the sentence and the state of affairs that it represents or pictures: these correlations may be likened to the antennae or feelers on an insect by which it reaches out and touches the world: “These correlations are, as it were, the feelers of the picture’s elements, with which the picture touches reality.” (Prop. 2.1515) In the Tractatus “the world is everything that is the case” (Prop. 1), and the meaning of sentences is exhausted by their picturing function through which they describe the world. There is nothing else that has any meaning. In the Philosophical Investigations [trans. G. E. M. Anscombe (Oxford: Blackwell, 1958)], Wittgenstein refers to any sentence in respect of its picturing function as a “proposition-radical”: Imagine a picture representing a boxer in a particular stance. Now, this picture can tell someone how he should stand, should hold himself; or how he should not hold himself; or how a particular man did stand in such and such a place; and so on. One might say (using the language of chemistry) call this picture a proposition-radical. This is

511 how Frege thought of the assumption. (Comment added to p. 11) Wittgenstein is correct about Frege: there is the assumption and then there is the attitude of assertion which can be added to the assumption. This is the use of a sentence in the indicative mood. But as Wittgenstein notes in his remarks, the picture, the assumption, the proposition radical, has other uses: it can be used to report, in which case it is in the indicative mood, or it can be used to cause somebody to do something, in which case it is in the imperative mood. In the Tractatus, Wittgenstein allowed only for the phrastic, the proposition-radical, but in the Philosophical Investigations, he discovered the mood, the neustic, which can be added as it were to the phrastic in order to put the latter to different uses. He discovered that not only are there world-word and wordword connections in language, there are also word-world connections. It is often said that Wittgenstein, when he wrote the Philosophical Investigations, abandoned the picture theory of meaning of the Tractatus, and replaced it with the theory that “meaning is use.” It is true that some philosophers abandoned picturing, or, what amounts to the same, a reference account of meaning for an account described by the slogan that “meaning is use.” (Cf. J. O. Urmson, Philosophical Analysis: Its Development between the Two World Wars [Oxford: Oxford University Press, 1956] for a good study of those who made this move.) But this is not true of Wittgenstein. In the Investigations he introduces “use” which is absent from the Tractatus, but he does not abandon the picturing role of the latter. What he discovered was that sentences were not bare pictures but that they were pictures which also had a mood: besides pictures or proposition-radicals as phrastics there was also the neustic, which added “use”, or rather “uses”, to picturing – that is, added pragmatics (“use”) to syntax (formation rules) and semantics (rules of reference, “feelers”). It is to mark the discovery of the mood or neustic that the Investigations begins by acknowledging the imperative mood. He takes the case of builder A and assistant B. A is building with building-stones: there are blocks, pillars, slabs and beams. B has to pass the stones, and that in the order in which A needs them. For this purpose they use a language consisting of the words “block”, “pillar”, “slab”, and “beam”. A calls them out; – B brings the stone which he has learnt to bring at such and such a call. – Conceive this as a complete language. (§ 2) A calls out “Slab” which has as its phrastic or proposition-radical ‘that slab is in this place’ and has as its neustic ‘make it be the case that’ [“Make it so,” says Captain Picard], which, when said to B, causes B to bring the slab to where A wants it to be. Wittgenstein is making it clear that in any full account of the meanings of sentences, one must include the pragmatics of language; he makes it clear by showing that one

512 must recognize in the sentence the neustic as well as the phrastic of the Tractatus. What he is not doing with this example is abandoning reference for use – indeed, how could he be dong that: how could A achieve his or her end of getting the slab unless the tern ‘Slab” meant in the sense of “referred to” a slab, and unless the context made clear, that is, referred to, the place where A wanted the slab to be put down? This means that Wittgenstein does not abandon the logical atomism of the Tractatus for the holism characteristic of so-called “Oxford philosophy” in the post-WorldWar-Two period: contrary to the common view, Wittgenstein’s ontology does not change, it is just that he discovers new aspects of meaning that he previously had not noticed or had ignored. But, details aside, he could have found out about these aspects of meaning in the emotivism of positivists like Ayer and Schlick. That at least is one reading of the curiously ambiguous Philosophical Investigations. Another reading is possible, however. One must include reference or naming and picturing as a component of the meaning of statements used descriptively. But one can read Wittgenstein as arguing in the Investigations that, contrary to the Tractatus and the logical atomists, that reference is one thing and cognitive content another, and that the former does not contribute to the latter. The meaning in the sense of cognitive content is given by use, not reference, where “use” is understood in terms of linguistic function, what it brings about, its effects in linguistic behaviour – as the function of the knife, what it means to the person eating, is that of cutting meat, say, into smaller pieces. Meaning in the sense of cognitive content is given by the word-word and wordworld connections. It is the pattern into which the item fits that gives that item its cognitive content. And this is holism. To be sure, the element of picturing is still there, but that does not determine the cognitive content of linguistic items. So, on this second reading of the Investigations, the later Wittgenstein does reject logical atomism and embrace holism. For a discussion of those who did move from logical atomism to holism, from a reference account of meaning to a “use” theory, see G. Bergmann, “The Revolt against Logical Atomism,” in his Meaning and Existence (Madison, WI: University of Wisconsin Press, 1959), pp. 39-72. Bergmann reads the later Wittgenstein in the second way, as a holist with a contextualist account of cognitive meaning: see his “The Glory and the Misery of Ludwig Wittgenstein,” in his Logic and Reality (Madison, WI: University of Wisconsin Press, 1964), pp. 225-241.. Carnap made a move similar to that of the “Oxford philosophers” from logical atomism and a reference account of meaning to a holistic contextualism in his 1936 essay on “Testability and Meaning,” Philosophy of Science, 3 (1936), pp. 419-471, and 4 (1937), pp. 1-40. For discussion of Carnap’s shift, see F. Wilson, “The Notion of Logical Necessity in Carnap’s Later Philosophy,” in A. Hausman and F. Wilson, Carnap and Goodman: Two Formalists (The Hague: Nijhoff, 1967), pp. 97-225, Ch. Three.

513

The holism of Carnap and “Oxford” is in effect the doctrine that terms can acquire meaning through “implicit definitions”; about this notion, see in this volume, below, the essay on “Implicit Definition Once Again.” Sellars has recognized the importance of mood as yielding in effect modal categories in language; see his “Reflections on Language Games,” and also his “Imperatives, Intentions, and the Logic of ‘Ought’,” in H. N. Castañeda, eds., Morality and the Language of Conduct (Detroit, MI: Wayne State University Press, 1965), pp. 159-218. Sellars, however, welcomes the holistic and contextalist account of cognitive content (while not, however, dismissing a role for picturing); see his “Meaning and Inference,” Mind, n.s. 62 (1953), pp. 313-338, and “Naming and Saying,” in his Science, Perception and Reality, pp. 225-246.. Erik Stenius, in his Wittgenstein’s ‘Tractatus’ (Oxford: Blackwell, 1960), has strongly, and almost singly, emphasized the distinction of phrastic and neustic, that is, sentence and mood, and its importance for an understanding of Wittgenstein’s philosophy, and for recognizing, what is missed by other commentators such as Urmson, the continuity in his thought. For a discussion of Stenius, see G. Bergmann, “Stenius on the Tractatus,” in his Logic and Reality, pp.242-271. 9.Cf. Wilfrid Sellars, “Imperatives, Intentions, and the Logic of ‘Ought’.” 10.To repeat the point made earlier, gappiness does not imply a lack of capacity to predict and to explain: this gappy law can be used to predict and to explain just as the gappy law, “water, when heated, boils,” can be used predictively when we make tea, and can be used to explain why some water is boiling in terms of its being heated. Gappiness may well constitute imperfection in the capacity to explain (to use Bergmann’s terminology), but does not (as some have alleged) imply a total inability to explain. 11.For this account of rules and of rulish behaviour, see Sellars, “Some Reflections on Language Games” and “Imperatives, Intentions,, and the Logic of ‘Ought’.” See also F. Wilson, “Marras on Sellars on Thought and Language,” in the present volume, below. 12.Cf. M. Brodbeck, “Mental and Physical: Identity vs. Sameness,” in P. Feyerabend and G. Maxwell, eds., Mind, Matter, and Method (Minneapolis, MN: University of Minnesota Press, 1966), pp. 40-58. 13.Cf. L. Addis, “Ryle’s Ontology of Mind,” in L. Addis and D. Lewis, Moore and Ryle: Two Ontologists (The Hague: Nijhoff, 1965), pp. 1-101; and also F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill (Toronto: University of Toronto Press, 1990), Ch. Eight.

514

14.Cf. F. Wilson, “Why I Do Not Experience Your Pains,” in M. Gram and E. Klemke, eds., The Ontological Turn: Essays in Honour of Gustav Bergmann (Iowa City: University of Iowa Press, 1974), pp. 276-300. 15.This research is admirably summarized in G. Humphrey, Thinking (New York: John Wiley and Sons, 1963), Chapters I-IV. In particular, Würzberg is defended against Titchener’s criticisms in Chapter IV; see especially p. 122ff. Oddly enough, Tuomela makes no reference to this or any other part of the body of research on this topic – but, then, neither does Sellars. G. Bergmann, however, certainly made clear these connections in his important study on “Intentionality,” in his Meaning and Existence, pp.3-38. 16.See in the present volume the essays on “The Aboutness of Thought” and “Effability, Ontology and Method.” 17.Wittgenstein came to recognize the fact that meaning is inseparable from imageless thought. Thus, compare the following remarks from the Philosophical Investigations: When someone says the word “cube” to me, for example, I know what it means. But can the whole use of the word come before my mind, when I understand it in this way? Well, but on the other hand isn’t the meaning of the word also determined by this use? And can’t these ways of determining meaning conflict? Can what we grasp in a flash accord with a use, fit or fail to fit it? And how can what is present to us in an instant, what comes before our mind in an instant, fit a use? What really comes before our mind when we understand a word? – Isn’t it something like a picture? Can’t it be a picture? Well, suppose that a picture does come before your mind when you hear the word “cube”, say the drawing of a cube. In what sense can this picture fit or fail to fit a use of the word “cube”? – Perhaps you say: “It’s quite simple; – if that picture occurs to me and I point to a triangular prism for instance, and say it is a cube, then this use of the word doesn’t fit the picture.: – But doesn’t it fit: I have purposely so chosen the example that it is quite easy to imagine a method of projection according to which the picture does fit after all. The picture of the cube did indeed suggest a certain

515 use to us, but it was possible for me to use it differently. (§ 139) The sound ‘bell’ has a meaning for one who understands the language English that is different from the meaning or lack of meaning for one who does not speak English. The meaning for the English speaker is manifest and is not attached to this sensory element, it could be attached to a different or somewhat different element. But for the non-English speaker, no such meaning appears in consciousness. Again, there is an image in Beverley’s mind, it is of his or her mother, but a different though no doubt similar image could have the same meaning, and in any case that meaning is manifest, not inferred, manifest in an imageless thought. These sorts of examples were investigated by the Würzburgers. Wittgenstein uses the example of an image of a cube carrying different meanings and of the same meaning carried by different, though related, images. The point is the same: when a word is used by a speaker of the language, the “whole use” of the word is before the mind; it is present as a manifest property and not as a dispositional property the presence of which is not manifest and would have to be inferred. Wittgenstein here recognizes this point, recording his discovery of imageless thoughts. And it seems it was indeed a discovery. Certainly, in the Tractatus, thought was taken to be imagistic, on the model of sentences. See Proposition 5.5421: It is clear ... that ‘A believes that p’, ‘A has the thought p’, and ‘A says p’ are of the form ‘“p” says p’: and this does not involve the correlation of a fact with an object, but rather the correlation of facts by means of the correlation of their objects. Not only is the model that of sentences in overt speech, but also the mood or propositional attitude is dismissed. We have already noted Wittgenstein’s discovery of the latter; see note 8, above. We have now also seen that Wittgenstein discovered that the simple model won’t do as an account of thought: he also came to see the need for imageless thoughts, and, in fact, clearly came to recognize their existence through the experiencing of them. No doubt he could have come to recognize the existence of these entities sooner had he read some of the literature on the topic, had he not been so pathologically adverse to looking at what others had written on a topic. 18.Some, e.g. the Churchlands, have suggested the whole language of the mental that we use will, as science progresses, change in their meaning – these terms will go from meaning such things as qualia to meaning electromagnetic waves or, perhaps, photons, emitted by congeries of atoms, and from meaning phenomenal thoughts to meaning neurological states. The world we are now acquainted with will become the world of science with which we are not now but will become acquainted. Indeed, those acts of acquaintance, our human consciousness, will become material. See Paul Churchland,

516 “Reduction,, Qualia, and the Direct Introspection of Brain States,” Journal of Philosophy, 82 (1985), pp. 8-28, and “Eliminative Materialism and the Propositional Attitudes,” Journal of Philosophy, 78 (1981), pp. 69-70. But such a position is simply silly, as J. T. Stevenson points out in his “Sensations and Brain Processes,” Philosophical Review, 60 (1960), pp. 505-510. Not even Sellars goes quite this far; see his “Philosophy and the Scientific Image of Man,” in his Science, Perception, and Reality, pp.1-40. The argument is that when we know a phenomenal quality like red we are noticing not the property red as we experience it but rather discriminating an emitted wave of light of a certain wave length or a group of photons or whatever the reducing theory of physics says is “really” there. The experiencing of red is itself reduced to a neurological state, that is what the experiencing “really” is. In general we have it that in theoretical reduction in science, F-ness in the reduced theory reduces to G-ness in the reducing theory just in case G-ness turns out to have all the causal qualities had by F-ness. As Churchland puts it, “... F-ness reduces to G-ness just in case that the ‘causal powers’ pf F-ness (as outlined in the laws of [the reduced theory] are a subset of the ‘causal powers’ of G-ness (as outlined in the laws of [the reducing theory])”. (“Reduction, Qualia and the Direct Introspection of Brain States,” p. 11) So, heat reduces to atomic motions because such motions have all the causal properties in the reducing realm as are had by G-ness in the reduced realm. Similarly for the colour quality red and for the consciousness quality of experiencing. And just as the concept of heat is eliminated through the reduction in favour of the concept of molecular motion, so the concept of the quality red is eliminated in favour of a concept regarding electromagnetic waves or photon, and so too the quality of experiencing is eliminated in favour of some concept or other of a neurological state of the brain. But we still use the term ‘red’, that is, use the concept of red the phenomenal quality when we give our observation reports, that is, as the reduction would have it, the reports of our discrimination of the presence of electromagnetic waves of a certain wavelength. Because of this reporting function, the concepts of phenomenal qualia seem uneliminable. But, Churchland argues, there is no reason to suppose we cannot train ourselves to use the language of the reducing theory to record our discriminations. That is, there is no reason we cannot train ourselves to use in our reporting the relevant sentences of the reducing theory. This is Churchland’s “central suggestion”: Consider the possibility of learning to describe, conceive and introspectively apprehend the teeming intricacies of our inner lives within the framework of a matured neuroscience, a neuroscience that successfully reduces, either smoothly or roughly, our common-sense folk psychology. Suppose we trained our native mechanisms to make a new and more detailed set of discriminations, a set that corresponded not to the primitive psychological taxonomy of ordinary language, but to some more penetrating taxonomy of states drawn from a completed neuroscience. And sup-

517 pose we trained our selves to respond to that reconfigured discriminative activity with judgments that were framed, as a matter of course, in the appropriate concepts from neuroscience. (“Reduction, Qualia and the Direct Introspection of Brain States,” p. 16) When our concepts drawn from neuroscience acquire this reporting function we shall have given to the concept ‘G’ from the reducing theory all the functions, that is, all the uses, of the term ‘F’ from the reduced theory, that is, our commonsensical “folk psychology.” So F-ness disappears completely into G-ness, for the concepts of qualia and for the concept of experiencing (and also for our concepts of other forms of consciousness). As Stevenson points out, the conclusion of the argument is silly. Does Churchland really think that the world as we experience it, the qualia and our feelings and our experiencing, all our conscious life, will simply disappear? Not even Skinner thinks this. But whatever Churchland really thinks, Stevenson is clearly correct: our experience is here, in the world, and that is all there is to it. As for ontology, anyone who starts with a Principle of Acquaintance clearly admits this world into his or her ontology. And, once again, that is all there is to it. But not quite: there is Churchland’s argument. It has a false conclusion. It must therefore be unsound. The issue is, just where is the error? The argument clearly depends upon the premise which asserts that F-ness in the reduced theory becomes G-ness in the reducing theory just in case that G has the causal qualities of F. In other words, G comes to mean in the sense of have the linguistic function what F means just in case that the regularities describing the functions (though not the reference) of the latter become regularities describing the functions of the former. Meaning, that is, cognitively relevant meaning, is given by the word-word connections in which a concept occurs: saying that meaning is determined by causal properties describing its function is to say that meaning is a matter of the regularities that give its word-word connections. And meaning, that is, cognitively relevant meaning, is not given by reference. The argument depends on the premise that cognitively relevant meaning is given by the implicit definition that a concept has by virtue of its appearing in a theory and not by its reference. The point to be made is that this assumption must be rejected. It rejects an ontology based on a Principle of Acquaintance by making meaning be determines not by reference, that is, reference to entities given to us in our experiencing of things, but by implicit definition by the theories in which they occur. But any account of meaning in terms of implicit definition is to be rejected; see “Implicit Definition Once Again” in this volume, below. In fact, Churchland gives no reasonable defence of this concept of cognitive meaning upon which his argument depends. One presumes that he has in mind such defences as that of Sellars in his “Meaning and Inference.” But Sellars’ ar-

518 gument fails; see Bergmann, “The Revolt against Logical Atomism,” and also, in the present volume, above, “Acquaintance, Ontology and Knowledge.” Churchland’s case fails for the same reason that that of Sellars fails.

Fourteen Marras on Sellars on Thought and Language* Wilfrid Sellars talks about our “Rylean” ancestors.1 What he is discussing in this way is the possibility of a purely behaviouristic language, that is,2 one in which the basic concepts refer to observable overt behaviour of persons, and in which all other concepts of the language are explicitly defined in terms of these basic concepts.3 If physicalism is correct, then the basic concepts of the language will be of the same sort as apply to inanimate objects. This would mean that all mentalistic terms, and specifically the crucial term ‘means’, will be explicitly definable physicalistically.4 Sellars’ story is an exemplification of his belief that physicalistic behaviourism is true. Ausonio Marras claims5 that Sellars claims that Sellars has been unable to avoid introducing irreducible mentalistic concepts. He claims that, when Sellars attempts to state his case, concepts are introduced which are incompatible with the physicalistic behaviourism that Sellars is trying to state. Now, I happen to think, with Quine and Bergmann, that behaviourism and physicalistic behaviourism is the view of all reasonable persons. This view of all reasonable persons has unfortunately come under attack from various quarters. One might note Chomsky’s attacks – one cannot say criticisms, for Chomsky’s remarks cannot be so honoured.6 Marras’ remarks are of a philosophically more worthy class. I think they are wrong. But they are wrong in interesting ways. The basic idea with which the physicalistic behaviourist begins is the utterance of sentences, overt verbal behaviour, – where the “sentences” are to be understood simply as noises, a point to which we shall have to return. If a person x utters ‘a is F’ at time t, let us say that ‘Pxt’. This property is, let us suppose, elicited by x being in situation S at t, or, for short, Sxt. If x exhibits the overt behaviour then he or she eo ipso exhibits the disposition to such behaviour. But of course, the disposition may be present even if it is not, as one says, actualized. In the latter case, we would have P*xt, * Originally appeared as “Marras on Sellars on Thought and Language,” Philosophical Studies, 28 (l975), pp. 91-102.

520 where this is defined as short for ‘if Sxt then Pxt’. As defined, ‘P*’ holds of x only at a moment t. We may call any such disposition a “dispositional property.” In addition, it might also be true that P**x, where this is short for ‘(t)(P*xt)’. When such a generality – which may be more or less restricted – holds over a period of time we may say that the individual is exemplifying a dispositional trait.7 Finally, it should be mentioned that situations, e.g. learning situations, can serve as stimuli which bring about the presence in persons of dispositional traits.8 Thus, there can be dispositional traits to acquire dispositional traits. It is with this sort of apparatus that Sellars, and any physicalistic behaviourist, proposes to talk about all human behaviour, including all verbal and meaningful behaviour. Any physicalistic behaviourist who is not also a metaphysical behaviourist will not deny the existence of mental states.9 Metaphysical behaviourism is the sort of view that Broad once characterized as “silly,”10 which indeed it is. We are therefore not surprised to find Sellars talking about mental states. He also adopts a specific thesis about their meaningfulness and intentionality, and this thesis is closely connected to another thesis of his that these entities are in some sense “theoretical.”11 Neither of these latter two theories need be accepted by one who accepts Sellars’ views on physicalistic behaviourisn and the existence of mental states. The question to be asked in the abstract is what relation must obtain if12 (I) we are to be able to infer mental states from behaviour (as we can in fact do), if (ii) mental states are able to explain causally the behavioural states of persons, and if (iii) we are also able to maintain the truth of physicalistic behaviourism that we do not need to appeal to mental states in order to predict (and explain) human behaviour? The simplist sort of answer which is not altogether implausible is that the mental states must be connected parallelistically to behavioural states, according to the schema (L) (x)(t){Mxt if and only if P*xt]13 If one knew the dispositional property P* were present, e.g. by observing its actualization in the behaviour P, then one could infer the presence of the mental state. Conversely, the presence of the mental state M together with this law enables one to infer and explain the presence of the dispositional property P*. At the same time, one can also explain the presence of P* in terms of the purely behavioural law which we called the dispositional trait P**, and more completely in terms of the acquisition of this trait. The concrete question is whether this abstract answer to our three-part question does in fact obtain. Which is to ask whether the connection between mental

521 states and bodily states are in fact of the parrallelistic sort (L). There are a number of qualifications which must be made, but it would seem that the answer is indeed affirmative, or at least all the evidence points in that direction and none in the negative direction. Sellars also accepts this point.14 Ausonio Marras’ argument challenges parallelism, specifically Sellars’ parallelism, suggesting that there is negative evidence for this claim: he proposes that the answer to part (iii) of our three-part question, above, is negative, that, in other words, we cannot avoid appealing to mental states in any attempt to predict human befaviour. An affirmative answer of part (iii) of our question means that, if parallelism is true then a physicalistic reconstruction of mental states will always be possible: one will be able to find a behavioural state which is parallel to the mental state. That means terms referring to mental states will always, in principle at least, be replaceable by terms which are physicalistically defined and which refer only to behavioural states.15 And insofar as the parallelistic connections (L) are not analytic,16 it will be logically possible at least to have the concept of meaningful speech and meaningful overt behaviour prior to the concept of thought. But Marras denies the latter: if he is correct then there can be no meaningful behaviour prior to thought. That means that the point at which Marras is attacking Sellars is on the truth of parallelism, and on the possibility of having an objective physicalistic science of human behaviour. In order to evaluate Marras’ views on this matter, we must turn to Sellars’ physicalistic reconstruction of ‘meaning’. Sellars understands ‘x means y’ in terms of ‘x has the same meaning as y’. This in turn is understood as ‘the role which x plays in its language is the same role as the role which y plays in its language.’ Finally, the ‘role’ which a linguistic item plays is defined, as Marras quotes, “in terms of the uniformities and propensities which connect the utterance with (1) other utterances (at the same or different levels of language), (2) with the perceptible environment, and (3) with courses of action (including linguistic behaviour).”17 Thus far, Sellars’ reconstruction of ‘meaning’ is physicalistic. And once he gets ‘meaning’ then there is not really much by way of principle that would suggest he couldn’t define all the other psychological concepts in terms of it. If he is to be criticised on the grounds that such a physicalistic reconstruction is not possible, as Marras proposes, then this is the point at which to get him. So we are not surprised to see that Marras goes after him right here, and tries to show that Sellars cannot define ‘meaning’ without presupposing the psychological

522 concepts he is hoping to define in terms of his ‘meaning’. The way in which he goes after Sellars is by taking up a point which Sellars makes much of, and which moreover is true. This point is that the uniformities to which one appeals in defining a linguistic role “... are not mere uniformities, for they are grounded in rules in a way most difficult to analyze, but which involves the causal efficacy of rule expressions.”18 Marras argues that insofar as the relevant behaviour is rule-governed Sellars has to bring in the psychological concept of something’s being intentional, in the sense of intentional action, and thereby invokes the notion of ‘meaning’ of which he is trying to give a physicalistic reconstruction. The essential question as I see it is whether or not one can give a physicalistic reconstruction of the idea of rule-governed behaviour, or, more generally, of action and intentional action. I think one can, and if this is so then Sellars can be defended. Indeed, Sellars himself hints at the defence in the passages we have quoted. For what he says is that the uniformities are not “mere uniformities”. But that is not to say that they are not uniformities. And insofar as they are uniformities, the physicalistic definition of ‘meaning’ that he proposes will go through. So: can one give a behaviouristic account of rule-obeying behaviour? Can one give a behaviourisitic account of how it is that the rules themselves are “are engaged in the genesis of the moves”?19 The point I wish to make can be best approached, I think, by the following example. The example is deliberately framed mentalistically. If this cannot subsequently be translated into a purely physicalistic statement, then Marras will be correct and physicalistic behaviourism will be shown to be false. But I shall propose that such translation is possible – and that Sellars holds it to be possible – and that therefore physicalism is defensible, in principle at least. Suppose that Beverly wills his or her arm to go up. Then (1)

Beverley volits that his or her arm go up

correctly describes Beverley’s state. Beverley being in the state described by (1) has the causal consequence that his or her arm goes up, and that he or she comes to be in the state described by (2)

Beverley’s arm goes up20

where the causal uniformity accounting for the event described by (2) fol-

523 lowing the event described by (1) is given by21 – more or less roughly – (3)

For any person, when that person volits that his or her arm go up then it goes up22

The volition, which (1) describes Beverley as having, has as its object Beverley’s arm going up. Beverley is not thinking of the uniformity (3), though it is the case that this uniformity describes the causal efficacy of the volition. The volition is “engaged in thee genesis of the move” even as its active role is completely described by the descriptive uniformity (3). But we are not yet at “rules”. Consider the case where Beverley feels that is wrong to lie – where the feeling is conative in nature, tending to move one to act23 : (4)

Beverley feels that it is obligatory that (it be the case that) for any person if he or she is in a situation in which he or she can choose between lying and not lying, then he or she chooses not lying.

Beverley recognizes that he or she is a person who falls under the rule and that he or she is in a relevant situation: (5)

Beverley recognizes that he or she is a person in a situation in which he or she can choose between lying and mot lying

which yields that (6)

Beverley feels that it is obligatory that (it be the case that) he or she not lie

and straightway (7)

Beverley performs the action of not lying

where these states are connected by the following sort of uniformities24: (8)

For any person, when that person feels that something generally ought to obtain, and recognizes that

524 the rule applies to him- or herself in the present situation, then he or she feels that he or she ought to do the something in question and (9)

For any person, when that person feels that he or she is obliged to do something, then he or she does it25

(9) is, of course, just a special case of (3). The rule is26: (10)

It ought to be the case that, for any person, if he or she is in a situation in which he or she can choose between lying and not lying, then he or she chooses not lying

which has this descriptive content27: (11)

For any person, if he or she is in a situation in which he or she can choose between lying and not lying, then he or she chooses not lying28

(10) is not, of course, really a rule until it has been, as one now says, internalized: it is not really a rule until the appropriate conative feeling has as it were become attached to it. (4) describes the state of affairs of Beverley having internalized the rule. To put it differently, it describes his or her awareness of the rule as a rule. (8) and (9) describe the causal efficacy of the (internalized) rule (19). This rule, or, more accurately, Beverley’s awareness of the rule, is engaged in the genesis of the move, viz., the move or action described by (7). Moreover, the move described by (7) is an instance of the descriptive content (11). Thus, if the whole (linguistic) community internalizes the rule, that is, if each member conatively feels that it is obligatory that the descriptive content (11) be true, then, given that (8) and (9) obtain, it will be the case that (11) is in fact true. Thus, the uniformity (11) is true, but it is not a mere uniformity. For the fact that it obtains is to be explained in terms of the rule which has (11) as its descriptive content.29 More explicitly, it is to be explained in terms of the fact that the rule has been universally internalized, and the higher level uniformities (8) and (9) which describe the causal efficacy of the internalized rules.30

525 And now we simply replace the mentalistic aspects by the corresponding verbal aspects, the parallel dispositions to say certain things, so that, for example, “Beverley feels that it is obligatory that p” is replaced by “Beverley tends to say ‘It is obligatory that p’.”31 This done, we can apply the schema to verbal behaviour. Here the rules will be those of language rather than moral rules, rules of syntax, semantics and pragmatics. But the point is the same, be they rules of morality of rules of syntax – they have the force of being generally obligatory. So, the moral rules of our example will simply be replaced by the norms of language. What (8) and (9) then describe is the linguistic role that the modal operator ‘it is obligatory that’ plays in language, namely, its motivating role.32 What Sellars is saying when he says that the uniformities in terms of which one defines the notion of a “linguistic role” are not mere uniformities is that the notion of ‘meaning’ in its essential linguistic sense is inapplicable to language-like behaviour unless there is in the language-like behaviour an item which plays in that set of behaviour patterns the role which is played by ‘it is obligatory that’ in our language, where this role is described by uniformities of the sort (8) and (9). The point is that there is no reason to suppose that this crucial feature, the role of ‘it is obligatory that’, cannot be handled physicalisticially. And if this is so, then Marras’ criticism is wrong. Notice that, in our example, Beverley is not merely conforming to the rule (19). He or she would be merely exemplifying a pattern of behaviour if (7) were true without having been a causal consequence of (4). (7) constitutes a case of obeying a rule because it is a consequence of (4). But there is no rule which brings it about that (8) and (9) are true in the way in the rule (10) qua internalized brings it about that (11) is true. So, obeying a rule is itself a case of pattern-governed behaviour – unless, of course, there are higher-level rules, but even then one must simply exemplify a (“mere”) pattern at some still higher level. And at that level language ceases to have a dimension that it has at lower levels. But this is hardly troublesome for Sellars. His definition of ‘meaning’ in terms of ‘linguistic role’ will be unaffected. And since it will remain unaffected, his physicalism remains unscathed. I think Marras goes wrong in the following way, if I may be permitted a speculative diagnosis. He thinks that if you are going to have rulegoverned behaviour at all, then you have to have it at every level. But in that case one will never come to a rule which is engaged in the genesis of the moves. Before any rule can get going, it must so to speak be activated by a higher-level rule. But that will require one at a still higher level. And

526 so on. The infinite regress precludes our ever getting started. So, in order to prevent the regress, he introduces another factor, a non-linguistic factor, an ultimate meaning-giver, the psychological element which is irreducible in physicalistic terms. He introduces our old friend the active self (of Platonic ancestry) that erupts into the physicalistic process and thereby precludes the possibility of a naturalistic science of human being. If I am correct, then another way out is simply to deny that in order to have rule-governed behaviour one must have rule-governed behaviour at every level. In that way one can save the physicalistic behaviourism as the view that every reasonable person ought to adopt. I mentioned earlier that Sellars has views on mental states that one need not accept. I had in mind his thesis that mental states insofar as they mean or intend are to be understood linguistically in the very same way in which language taken physicalistically is understood. Some reflections and comments on this point are in order. Sellars analyzes thought on the model of language: the aboutness or intentionality of thought is the same sort of thing as the intentionality or aboutness of language.33 This means, for Sellars, that the intentionality or aboutness of thoughts is to be analyzed into dispositions and uniformities with respect to those entities which correspond to sounds and marks. This language, or, rather, “language”, Sellars refers to as “mentalese.” In mentalese in order to identify a thought as about something it is necessary to locate a non-intentional item among a pattern of non-intentional items, just as we do with overt speech. The meaningfulness is given by the pattern, the “linguistic role,” and not by the items which make up the pattern, just as in the case of overt verbal behaviour it is noises and marks which are in themselves meaningless but which carry meaning, where this meaning is a matter of being part of a certain pattern. In particular, one must have world-word connections, rules of reference, that coordinate marks to objects, and rules of syntax that establish that combinations of marks reflect combinations of objects– rules, in other words, that establish that there is an isomorphism between basic sentences and combinations of objects (that is, facts).34 The same holds for mentalese: in thinking there are basic components, which are sensory contents, and combinations of these entities, like combinations of noises and marks in overt speech, are isomorphic to the facts and states of affairs that the represent. The meaningfulness of these sensory contents is not intrinsic, they are not by the own nature inten-

527 tional, just as marks and noises in overt speech are not by their own nature intentional. The meaningfulness of the sensory contents in mentalese is given, rather, by the pattern, the “linguistic role,” just as in the case of overt verbal behaviour it is meaningless noises or marks become meaningful only through being parts in a certain pattern or sets of patterns. This view of the meaningfulness of thought is that of Titchener.35 The items which carry the meaning are non-intentional sensory contents. Titchener’s views are essentially those of the associationists, from Berkeley onwards, and, as Boring points out, are continuous with the behaviouristic theories of meaning of subsequent psychologists such as Holt and Tolman.36 It is best to note in this context that it is one thing to accept the behaviouristic account of meaning as applied to behaviour and another thing to apply it to thought. One can hold that the contextualist account of meaning is good behaviourisitics but that it cannot be applied to mentalistics, or at least, if it is so applied, then certain qualifications must be made. And it seems to that this is precisely what one ought to say about thought.37 If Sellars – and Titchener – are correct, then to identify a thought as being about something, as meaning or intending something, it is necessary to identify the pattern from which its meaningfulness derives. This pattern connects the sensory content, which is in itself meaningless, to other sensory contents, past, present and future, to the extra-mental world, and to bodily movements. The pattern as such is exemplified not just by the specific content which is presently in consciousness, the content in which we are now interested, but by this content and all other items to which it is connected by the pattern. But these other items are not all in consciousness simultaneously. So neither is the pattern which gives meaning to the content which is wholly present in consciousness. Titchener explicitly recognized this, and suggested that the meaning was carried unconsciously, by some appropriate physiological state in the brain.38 If the dispositional and lawful connections were not themselves wholly present, these could at least (he believed) be correlated (after the fashion of (L)) with some (unconscious) physiological state.39 In this way Titchener hoped to account for the paucity of sensory contents in the case of thought, that is, the thought that occurs in the higher mental processes. Indeed, he saw this as his contribution to the context theory of meaning, as his improvement of that theory over the usual associationist account deriving from Berkekey. This move of Titchener is ingenious. It is the move that Sellars makes.40 But it won’t work. For it misses the important fact that meaning is wholly present in consciousness: when we think that p there is wholly pre-

528 sent to us in consciousness an awareness that the thought is one that means p – this knowledge is, to repeat, wholly present to us and not present either dispositionally or physiologically, neither of which are present in any way in consciousness. This very simple point was made by the Würzberg psychologists associated with Oswald Külpe. Titchener had his students work on this problem, and they performed some remarkable feats of introspection, but they were never able to make sensory contents into imageless thought: meaning simply was wholly present and that was all there was to it. Present the sound ‘bell’ to someone who knows English then there is in his or her consciousness an awareness that this means bell, the thing with the characteristic sound, but present the same thing to someone who does not know English then such an awareness will be quite absent. Different images can carry the same thought: the image one has of one’s mother may be different on different occasions but in each case there is present in consciousness an awareness that the sensory content is an image of one’s mother: the sensory content differs but the thought – the imageless thought – is the same.41 Phenomenology requires that mind – intentionality – meaning – be analyzed in such a way that there is wholly present in conscious an imageless non-sensory something that constitutes our knowledge that we are thinking of this rather than that.42 Sellars’ account fails for precisely not squaring with this phenmenological fact.43 The mental, thought, is irreducible.44

529 Endnotes to Study Fourteen

1.See W. Sellars, “Empiricism and the Philosophy of Mind,” in his Science, Perception and Reality (London: Routledge and Kegan Paul, 1963), pp. 127-196; see in particular p. 178.. 2.Cf. G. Bergmann, “The Contribution of John B. Watson,” Psychological Review, 63 (1965), pp. 265-76. 3.Sellars, in his essay “Scientific Realism or Irenic Instrumentalism?” (in his Philosophical Perspectives [Springfield, IL: Charles C. Thomas, 1967], pp. 337-369) refers (p. 342) to what Bergmann, “Outline of an Empiricist Philosophy of Physics,” (in H. Feigl and M. Brodbeck, eds., Readings in the Philosophy of Science [New York: Appleton-Century-Crofts, 1953], pp. 262-287), calls the “empirical hierarchy” of concepts, where Bergmann explains that these are operationally and explicitly defined concepts that have observation terms as their undefined basis. On this ground I take Sellars to be accepting the indicated notion of a purely behaviouristic language. Of course, for Sellars there is more to the conceptual meaning of a concept than in the definition: the laws in which it appears serve to define, that is, as one says, “define implicitly,” and thereby adds to its conceptual content (cf. Sellars, “Comments on Maxwell’s Paper,” in H. Feigl and G. Maxwell, eds., Current Issues in the Philosophy of Science [New York: Holt, Rinehart and Winston, 1961], pp. 183-191; and also his “Is There a Synthetic A Priori?” in his Science, Perception and Reality [London: Routledge and Kegan Paul, 1961], pp. 298-320). But Bergmann’s distinction (in his Philosophy of Science [Madison, WI: University of Wisconsin Press, 1957], pp. 48-67) between meaning in the sense of definition (or, in the case of primitive undefined terms, rules of reference) and significance (laws in which the concept occurs) is one which Sellars (in his own way) accepts (see his “Theoretical Explanation,” in his Philosophical Perspectives, p. 325), so that the latter’s views on “implicit” definition are compatible with the idea of a purely behaviouristic language in which all nonprimitive terms are explicitly operationally defined. See his “Empiricism and the Philosophy of Mind,” in his Science, Perception and Reality, pp. 127-196, at p. 185, p. 186, p. 189. For more on Sellars on meaning as “implicit definition”, see note 32, below. 4.On the basis of Sellars’ correspondence with Chisholm (W. Sellars and R. Chisholm, “Intentionality and Mental,” [in H. Feigl, M. Scriven and G. Maxwell, eds., Minnesota Studies in the Philosophy of Science, vol. II [Minneapolis, MN.: University of Minnesota Press, 1958), pp.507-39]), it may be thought that Sellars disagrees with this and holds that ‘means’ cannot be explicitly defined physicalistically (cf. also “Empiricism and the Philosophy of Mind,” p. 180). But see below; it is argued that Sellars is perfectly clear, or, at least, as pellucid as Sellars can be, on the physicalistic definability of ‘means’.

530

5.A. Marras, “Sellars on Thought and Language,” NoØs, 7 (1974), pp. 152-163. 6.Herbert Spencer said sufficient on the subject of the compatibility of empiricism, in the sense of a metaphysics and epistemology (though not in the sense of a theory of learning, as in associationism in Mill, Spencer and Skinner) and innatism or nativism (as an account of our learning or acquisition of language), to settle the issue, one would have hoped, for all time. See H. Spencer, The Principles of Psychology, 2 vols. (New York: Appleton, 1888), vol. ii, Ch. xiv, “The Perception of Space,” pp. 178-206. Chomsky has read widely, but with evident selectivity – may one suspect bias? or perhaps even malice aforethought? – and so quite possibly is not acquainted with Spencer’s work. For more on Chomsky, see the Appendix to “Language and (Other ?) Objects,” this volume, below; and also the Appendix to “Implicit Definition Once Again,” this volume, below. 7.Cf. G. Bergmann, “Dispositional Properties and Dispositions,” Philosophical Studies, 6 (1955), pp. 77-80. 8.Cf. D. Seivert, “Historical Cross-Section Laws and Austin’s Scepticism in ‘Other Minds’,” Philosophy of Science, 37 (1970), pp. 146-52. 9.Cf. L. Addis, “Ryle’s Ontology of Mind,” in L. Addis and D. Lewis, Moore and Ryle: Two Ontologists (The Hague: Nijhoff, 1965), pp. 1-101. 10.C. D. Broad, Mind and Its Place in Nature (London: Routledge and Kegan Paul, 1925), pp. 5-6. 11.Sellars, “Empiricism and the Philosophy of Mind,” sec. xv. See also his “The Language of Theories,” in his Science, Perception and Reality, pp.106-26, in sec. iv. 12.This is not to say, contrary to the way Fodor expresses, more boldly than reflectively, the matter, that mental states must be connected – logically speaking – to physical states. See J. Fodor, An Introduction to the Philosophy of Psychology (New York: Random House, 1968), p. 51. Rather, what we are asking about is what must be true if.... Fodor’s way of putting it makes the thesis of bevaviourisn a necessary truth – which it isn’t – where our way of putting it makes it a contingent truth, but a contingent truth whose truth is necessarily consequent upon another contingent truth. Compare also F. Wilson “Why I Do Not Experience Your Pains,” in M. Gram and E. Klemke (eds.), The Ontological Turn: Essays in Honour of Gustav Bergmann (Iowa City: University of Iowa Press, 1974), pp. 276-300; and also F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill (Toronto: University of Toronto Press, 1990), Ch. Eight. 13. Cf. Addis, “Ryle’s Ontology of Mind,” p. 43ff. Also D. Seivert, “Historical Cross-

531 Section Laws and Austin’s Scepticism in ‘Other Minds’,” and F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill, Ch. Eight. There is much talk nowadays about mental states “supervening” on brain states. The term derives from John Stuart Mill, who used it to make the same point in his Introduction to the Second Edition of his father’s Analysis of the Phaenomena of the Human Mind. Many current writers on the topic, e.g. Jaegwon Kim, think the idea is new: it isn’t. The talk about possible worlds which often accompanies these discussions adds nothing of philosophical interest to the topic: it’s all just jazzed up – and obfuscated – discussions of parallelism. 14.Cf. Sellars, “Empiricism and the Philosophy of Mind,” pp. 186-7; and his “Philosophy and the Scientific Image of Man,” in his Science, Perception and Reality, pp. 140, at p.31, where he remarks that “...there are cerebral processes ... which run parallel to conceptual thinking (and cannot be identified with it)...”. Sellars’ commitment to a form of parallelism comes out quite clearly in his comments on theories and reduction. Reduction requires an isomorphism. See his “The Language of Theories,” p, 124; his “Empiricism and the Philosophy of Mind,” sec. xii; and his “Theoretical Explanation,” [in his Philosophical Perspectives, (Springfield, IL: Charles C. Thomas, 1967), pp. 321-36], p. 330. Sellars takes this isomorphism to be a sort of identity, which is (like supervenience) but a strong form of parallelism. See his remark that “...the inner states which as sensations are conceived by analogy with their standard causes [that is overt behaviour and dispositions to such behaviour], are in propria persona complex neurophysiological episodes in the cerebral cortex...” (“Philosophy and the Scientific Image of Man,” pp. 34-35.) Whether one takes the strong form or a weaker form is a different matter than whether one opts for one form or other of parallelism. Sellars’ opting for the strong form is a matter of his specific views on mental states, rather than on his views as to whether mental states, whatever they are, are connected parallelistically to behavioural states. For a discussion of mere parallelism vs. identity, see M. Brodbeck, “Mental and Physical: Identity vs. Sameness,” in P. Feyerabend and G. Maxwell, eds., Mind, Matter, and Method (Minneapolis, MN: University of Minnesota Press, 1968), pp. 40-58. 15.See Sellars, “Empiricism and the Philosophy of Mind,” sec. xiv. 16.Sellars has the parrallelism an identity, but it is a contingent rather than a formally analytic truth in the sense that the possibility of identification is something that is discoverable through experience; cf. his “Imperatives, Intentions,, and the Logic of ‘Ought’,” in H. N. Castañeda, eds., Morality and the Language of Conduct (Detroit, MI: Wayne State University Press, 1965), pp. 159-218, at p. 384. 17.Sellars, “Notes on Intentionality,” in his Philosophical Perspectives, pp. 308-320, at p. 310.

532

18.Sellars, “Notes on Intentionality,” p. 310. 19.Sellars, “Reflections on Language Games,” in his Science, Perception and Reality, pp. 321-58, at p. 341. 20.Sellars, “Thought and Action,” in K. Lehrer, ed., Freedom and Determinism (New York: Random House, 1966), pp. 105-40. p. 110. 21.Sellars, “Reflections on Language Games,” p. 329; “Thought and Action,” pp. 108109; and “Empiricism and the Philosophy of Mind,” pp. 186-187. 22.This is only “roughly” because the generality (3) embodies only what has been called “gappy” knowledge or, what Bergmann has called, equivalently, “imperfect knowledge”. (3) is gappy or imperfect knowledge because it holds only under certain conditions; in this case, the persons must have learned the behaviour patterns (3) describes. Cf. Sellars, “Reflections on Language Games,” p. 329, and “Empiricism and the Philosophy of Mind,” p. 161. For Bergmann on imperfect knowledge, see his Philosophy of Science, Ch. Two. The term ‘gappy’, describing the same thing, comes from J. L. Mackie, “Causes and Conditions,” American Philosophical Quarterly, 2 (1960), pp. 245-64. For a more detailed discussion of gappy knowledge, see F. Wilson, The Logic and Methodology of Science and Pseudoscience (Toronto: Canadian Scholars Press, 2000). 23.In Hume’s terms, Beverley has a feeling of moral disapprobation, and this passion is conative, moving him or her to action in the appropriate circumstances. 24.Sellars, “Imperatives, Intentions, and the Logic of ‘Ought’,” pp. 208-209. 25.Again, (8) and (9) are gappy. But in spite of their gaps, or, as Bergmann would say, their imperfection, they are nonetheless still laws and can be used in explanation and prediction. “Water, when heated, boils” is gappy, but useful when we wish to make tea, and provides an explanation when we are asked why that kettle, there, on the stove, is boiling. For these points, see F. Wilson, The Logic and Methodology of Science and Pseudoscience. 26.Sellars, “Imperatives, Intentions, and the Logic of ‘Ought’,” p. 180ff. 27.Sellars, “Thought and Action,” pp. 105-106, and “Imperatives, Intentions and the Logic of ‘Ought’,” p. 168. Also his “Empiricism and the Philosophy of Mind,” p. 146ff, and p. 150ff. 28.Sellars, “Imperatives, Intentions, and the Logic of ‘Ought’,” p. 194ff and pp. 208209.

533

29.The causal efficacy of the rule is of course acquired. This acquisition is a matter of learning. Sellars deals with these matters in terms of the distinction between ought-tobe’s and ought-to-do’s. See his “Language as Thought and Language as Communication,” Philosophy and Phenomenological Research, 29 (1968), pp. 506-527. Also his “Reflections on Language Games,”, p. 324ff, and “Empiricism and the Philosophy of Mind,” p. 188. 30.Cf. Sellars, “Imperatives, Intentions, and the Logic of ‘Ought’,” p. 210ff’ and also his “Thought and Action,” pp. 138-139. 31.Sellars, “Empiricism and the Philosophy of Mind,” p. 161ff, p. 179f. 32.Cf. Sellars, “Reflections on Language Games,” p. 350ff. Note that for Sellars, this is not to say that ‘ought’ has motivational rather than conceptual meaning. Here his views are to be distinguished from those of the emotivism of several of the positivists, such as Ayer and Schlick. For the latter, conceptual content is given by the semantic rules of reference for the descriptive vocabulary. ‘Ought’ does not in this sense for these philosophers refer. Rather, its use on the one hand expresses (rather than describes) a mental state of feeling and on the other hand causes others to feel the same way. For Ayer and Schlick it expresses the (affective) mental state described by “I like that” and as it were urges you to like it too. This makes the view of these emotivists into a crude hedonism, for which they can rightly be criticized. They would have done better to endorse the more plausible view of Sellars (and Hume and R. M. Hare) that to say that something ought to be expresses the conative mental state of feeling that something is obligatory: we have a logic of imperatives rather than a logic of liking. One can then distinguish (as Ayer and Schlick cannot) between interest and duty. But to say this is not yet to allow that ‘ought’ lacks conceptual content: on the positivist/imperativist account it still lacks such content. On Sellars’ view it does not. For, on Sellars’ view, conceptual content is not given by the rules of reference, as it is for the positivists/imperativists, (that is, the world-word connections), but rather by the wordword connections (rules of inference) it has with other words in the language (conceptual scheme). For ths view of conceptual content, see Sellars, “Meaning and Inference,” Mind, n.s. 62 (1953), pp. 313-338, at p. 336, for a particularly clear statement of this (implausible) position. See also Sellars, “Is There a Synthetic A Priori?”, p. 316ff. Since ‘ought’ has word-word connections, it will have conceptual meaning in addition to its motivational meaning which is primarily a matter of word-world connections. The view that conceptual content is given by rules of reference (world-word connections) rather than word-word connections (rules of inference), see the Appendix to the essay on “Acquaintance, Ontology, and Knowledge” in this volume, above. Sellars’ view may be said to be one in which the rules of inference give meaning to the terms they govern by “implicitly defining” them. For criticism of this doctrine of “implicit definition,” see the essay “Implicit Definition Once Again,” in this volume, below. Also F. Wilson, “The Notion of Logical Necessity in Carnap’s Later Philosophy,”

534 in A. Hausman and F. Wilson, Carnap and Goodman: Two Formalists (The Hague: Nijhoff, 1967), pp. 97-225, Ch. Three. Sellars, “Meaning and Inference,” appropriates some views of Carnap. Carnap himself moved towards the non-positivist doctrine of “implicit definition” in his 1936 essay “Testability and Meaning,” Philosophy of Science, 3 (1936), pp. 419-471, and 4 (1937), pp. 1-40. This essay presents his notion of a “reduction sentence”, which is a version of the doctrine that terms receive their meaning from their inferential connections to other concepts, in other words, a version of the notion of “implicit definition”; this notion of a “reduction sentence” together with his notion of “A-truth” and “Meaning Postulates”, a later version of the same doctrine of “implicit definition”, are critically evaluated in Wilson, “The Notion of Logical Necessity in Carnap’s Later Philosophy,” Ch. Three. If one holds that conceptual content is given by the rules of reference, as the early positivists held, then ‘ought’ has expressive and motivational meaning but lacks referential content, and we have again the view of the early positivists, in this case their emotivism. But it is an emotivism made more plausible by denying [with Hume and Hare], its simplistic hedonism. I would suggest that what people found shocking in Ayer’s emotivism was not its denial of cognitive meaning to the language of morals, but rather its simplistic hedonism. Hare’s discussion of imperatives was a version of emotivism, but one made more human – and more acceptable – by its retrieval of what was moral about our language of morals, namely, our feelings of moral approbation and disapprobation. Sellars’ views on the language of morals, when shorn of the doctrine of “implicit definition,” is a plausible version of the imperativism of Hare – and Hume. Like the positions of the latter two thinkers, it becomes a plausible and defensible position in moral theory. 33.Sellars, “Empiricism and the Philosophy of Mind,” p. 186ff, and “Notes on Intentionality,” in his Philosophical Perspectives, pp. 308-20, at p. 309ff. 34.Cf. Sellars, “Naming and Saying,” for this account of language. Sellars places this in the context of a nominalisitic construal of the objects to which the basic terms of the language refer. But the same points hold if one holds that the world consists of particulars with properties construed realistically as universals. The basic point is the same: the rules of language (world-word and word-word connections) establish an isomorphism between combinations of marks or noises on the one hand and combinations of objects on the other. In Wittgenstein’s metaphor, the basic sentences picture facts and states of affairs. For other aspects of these issues, see also Sellars’ “Notes on Intentionality.” See also the essay on “The Aboutness of Thought”, above. As for Sellars’ nominalism, see F. Wilson, “Universals, Particulars and Tropes:

535 The Role of a Principle of Acquaintance in Ontology,” in the present volume, above. 35.See the Titchener pieces in R. Herrnstein and E. G. Boring, eds., A Source Book in the History of Psychology (Cambridge, MA: Harvard University Press, 1966); see in particular, p. 193. For further discussion of these issues, see also G. Bergmann, “Intentionality,” in his Meaning and Existence (Madison, WI: University of Wisconsin Press, 1959), pp. 1-38. 36.Boring in R. Hernstein and E. G. Boring, A Source Book in the History of Psychology, p. 193, p. 197. 37.See the essays in this volume on “The Aboutness of Thought” and “Effability, Ontology and Method.” 38.See Titchener in R. Hernstein and E. G. Boring, A Source Book in the History of Psychology, p. 195. 39.Sellars, “Philosophy and the Scientific Image of Man,” p. 31f. 40.See Sellars, “Philosophy and the Scientific Image of Man,” 31f. 41.G. Humphrey, Thinking (New York: John Wiley and Sons, 1963) summarizes excellently the whole debate in Chapters I-IV. In particular, Würzberg is defended against Titchener’s criticisms in Chapter IV; see especially p. 122ff. Compare also the discussion of this episode in the history of psychology in F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill, Ch. Eight. 42.See also the essays in the present volume on “The Aboutness of Thought,” above, and, below, “Effability, Ontology and Method.” 43.Sellars rejects such an appeal to phenmenology, to acquaintance: “It is a mistake,” he tells us, “to suppose that we must be having verbal imagery – indeed, any imagery – when we ‘know what we are thinking’ – in short to suppose that ‘privileged access’ must be construed on a perceptual or a quasi-perceptual model” (“Empiricism and the Philosophy of Mind,” p. 178). This is his well-known attack on what he calls the “myth of the given.” But be what we may as philosophers, we cannot simply deny that the world is as we experience it to be: what is present to us in experience is experienced and must be held to be as it is experienced. In any case, Sellars’ attack on the “given,” his argument that it is a “myth” to take seriously what we experience, fails: the Principle of Acquaintance can be defended – see the essay on “Acquaintance, Ontology and Knowledge,” in this volume, above.

536

44.This is Brentano’s essential point against the associationists, be they the British associationists or the physiologically oriented associationists of Wundt’s laboratory, among whom one must count Titchener. See F. Brentano, Psychologie von empirischen Standpunkt, 2 vols. (Hamburg: Fzelix Meiner, 1955), vol II, p. 190ff. Insofar as this is the main starting point of Bergmann’s study on “Intentionality” [in his Meaning and Existence, pp.3-38], that essay must stand as one formulating a truth to be accommodated, whatever the problems are with Bergmann’s own solution to some of the problems he discussion (see Wilson, “Effability, Ontology and Method” in the present volume, below). The discussion in Bergmann’s essay can be ignored only at the of ceasing to do serious philosophy. Sellars always had Bergmann in mind in his discussions of intentionality (see his “Notes on Intentionality,” and also his “Naming and Saying”), though he often does not mention Bergmann. His students seem to have taken this sort of silence as a reason for ignoring Bergmann’s work.

Fifteen Effability, Ontology and Method: Themes from Bergmann’s Ontology* Now let the generall Trumpet Blow his blast, Particularities, and pettie sounds to cease. 1

Bergmann has proposed an ontology that contains as entity many find exceedingly strange. That entity is particularity. And in fact Bergmann, too, seems to find it strange. He proposes a phenomenological method in ontology, one based on a Principle of Acquaintance, and he holds, as he therefore should, that particularity is presented, that he is acquainted with it. At the same time he also holds that it is ineffable, and that its presence in a particular is an unsayable state of affairs, and further that it is something which is not a thing but yet is not nothing. The aim of this study is to explore Bergmann’s method in order to see whether we cannot get a better grasp of this entity, particularity. Specifically , we will try to see whether it is not, after, effable. But actually, this disagreement with Bergmann on the effability of particularity is really, as we shall see, THREE DISAGREEMENTS. In what follows we shall first discuss Bergmann’s philosophical method and gradually move to the case of particularity and our three disagreements. Our discussion of particularity will focus on three late papers by Bergmann.2 These essays raise many issues of great importance,3 and in picking a few that centre around particualrity it should not be concluded that the others do not deserve discussion. It was Kant’s suggestion, and before him Locke’s, that before undertaking research in ontology we ought to investigate our tools, to see what sort of job they can do, what sort of world they can reveal to us. That is, or so goes the suggestion, epistemology precedes ontology. Now, the maxim that one should check one’s tools before one starts work may sound advice for a carpenter, but I cannot think of worse advice for a * Originally appeared in Philosophy Research Archives, 9 (1983), pp. 419-470.

538 philosopher. Another maxim, equally appropriate for the philosopher: investigate carefully the nature of the material before you try to design a tool to work upon it; that is, for the philosopher, ontology must precede epistemology. But even this maxim is misleading in the case of ontology. For, to do ontology we must know what we are talking about. Epistemology thus seems to be both presupposed by and also justified by ontology. But that apparent circle should not surprise us: it is but one way of noting the delicacy of the metaphysical-ontological-epistemological enterprise. One point, however, is clear, and that is the starting point of this enterprise: the starting point of ontology is the world as we experience it, and the starting point of epistemology is our experiencing of the world, and, in fact, these two starting points cannot be separated. Our experiencings of the world from which epistemology takes its start are themselves part of our experience, and part of the world as we experience it, and therefore part of the world that is the starting point of ontology. As for our ways of experiencing: they are manifold. Among the ways of experiencing the world, that we encounter in our experience, are our loves and hates, our perceivings and our rememberings, our sensings and our thinkings, our believings and our disbelievings, our expectations and our hallucinations, our scientifically methodical researches and our occult fantasizings. No doubt we are agreed that the justified procedure for extending our knowledge of the world is to start from perception and memory and to explore the world by the methods of science: that is the pathway to truth. But that judgment – that perception and science are the measure of what is – presupposes that certain basic moves in ontology have already been made. It presupposes that the world that we experience has a certain ontology and that our various ways of experiencing it have been examined, located in their place in the world as we experience it, and their capacity to testify to the truth subjected to criticism and evaluated. All of which means that the starting point of all ontology can be none other than the phenomenological description of the world as we experience it and of our experiencings of that world. As Bergmann once put it: “Ontology is phenomenological. The rest is merely science. Such is the nature and such are the limits of human knowledge.”4 To suggest otherwise – to suggest as, for example, Jay Rosenberg has done – that science begins with epistemology, and that epistemology is exhausted by the methods of science5 – is simply to avoid all the serious work of philosophy. It is, of course, true that among the things that we find in experience,

539 in the world we experience, are hallucinations, or, more generally, falsehoods. In fact, perhaps almost all our ordinary perceptions are false, as Locke, Hume, and other critical realists have, on the basis of the inferences of science, so cogently argued. That, however, is something we find out about the world we experience: recognizing that fact is one way of experiencing the world. Moreover, that atoms or whatnot that are there where our perceptions are false we know to be there because of what we discover through experience, and are parts – though perhaps unperceived parts – of the world we experience. For, our experience shows us that there are, in the world we experience, many parts that we do not perceive. Why should atoms not be among the unperceived parts of the world of our experience? But be all that as it may, the point is that, to say that ontology is phenomenological is not to deny the fact of error: error and its correction are processes that must be described by any phenomenological description of experience, that is, of the world that we experience and of our experiencings of that world. A full pheomenological description of the world that we experience cannot but include a description of falsehood and errors. This being so, it means that what is given to us in experience – what is presented to us in experience – is sometimes what, later, or perhaps even at the moment, we decide is not the case. It follows that what is given to us, what is presented to us, is not given or presented incorrigibly. It is a myth created by foundationalists that what is given is always given incorrigibly. Or perhaps better, it is a myth that was created by those who are concerned to refute something they call the “myth of the given.” There is, of course, one sort of experience, often referred to as sensing, or, by some, who wish to include inner awareness, as direct awareness, which is, in some sense, error-free, or, at least, is such that it makes no real sense to speak of correcting it. What is given in the kind of experience called sensing and inner awareness are often in turn referred to as phenomena, or as phenomenal entities. There is a sense, then, in which phenomena might be said to be given incorrigibly. But phenomena are only some among the many sorts of entity that are presented, and sensing and inner awareness are only two of the many ways of experiencing the world. Moreover, phenomenology or the description of the world that we experience is not the description of phenomena. Thus, while, on the one hand, when we offer a phenomenal description of something, what we describe must, in some sense of ‘must’, exist, on the other hand, when we offer a phenomenological description of something, it simply does not follow that what is thus described exists.

540 In addressing ourselves to the philosophy of Gustav Bergmann it is important to be clear on this point. When one criticizes a philosopher, it is best to be clear on the method which that philosopher proposes to use in his or her attempts to solve those problems and perplexities that are typically philosophical. Rosenberg is an example of one who has failed to follow this maxim. Bergmann proposes that the core of ontology – and of epistemology – is phenomenology.6 Ontology begins with what is presented to us, or, as he also says, synonymously, with what we are acquainted with.7 What we are acquainted with we know8 – in the sense that we know what we are presented with, what is given to us: we can describe it phenomenologically. But, of course, we do not thereby know it in the sense that in knowing it we know incorrigibly that it exists. Acquaintance in this sense is not incorrigible, and it is not what some foundationalists at least have called “direct awareness,” where the latter is so direct that it incorrigible. To be sure, Bergmann does speak of “direct awareness,”9 which is a way of experiencing that includes at some of those sorts of experiencings that others have called “sensings.” “Direct awareness” is therefore in a way incorrigible: in some sense of ‘must’, what is given in direct awareness must exist. But contrary to what Rosenberg suggests,10 when Bergmann speaks of ontology as phenomenological, or beginning with what one is acquainted with, he is not speaking of direct acquaintance in the foundationalist’s sense nor of some incorrigible starting point. Failure to appreciate this point means that Rosenberg almost everywhere misunderstands Bergmann, and it throughout his essay vitiates most of his critique of the latter. To say that ontology begins with phenomenological description is not to say that the philosopher does not argue. Description does not exclude dialectics. Thus, Bergmann regularly argues that certain entities that are picked out in his phenomenological descriptions solve certain philosophical problems, and that other philosophers cannot solve these problems without invoking these entities. Such arguments must carefully be distinguished from transcendental arguments. In the latter sort of argument, one argues that certain things that exist within experience could not exist unless certain entities outside the world of experience also existed. For example, Plato argued that the things within the world of experience may often correctly be said to be the same, but that no entities within the world could account for that fact, and that, therefore, there must be entities outside the world of experience that account for the sameness of experienced things. These objects are the Forms. These are not simply unexperienced parts of

541 the world we experience, likes specks of dust too small to see, or atoms, which are even smaller; rather, the Forms transcend the world of ordinary experience. But then, if we are to identify things as the same, then we must have some knowledge of these Forms. This provides the basis for a further transcendental argument on Plato’s part, an argument to the effect that there must exist a special sort of form of knowing, which Plato calls “reminiscience,” that gives us knowledge of these Forms. Needless to say, this sort of knowing is not something that is described when one gives a phenomenological description of the world of experience, and, specifically, of the various modes of experiencing that are given to us in the world of experience. In contrast to Plato, Bergmann systematically picks out certain entities in the world of experience, and then argues that these entities solve the problem of sameness that Plato thought could be solved only by introducing transcendental entities. Bergmann argues concerning entities in the world of experience; Plato argues for the existence of entities that transcend the world of experience. In fact, Bergmann proposes a method that conforms to a Principle of Acquaintance: not only is phenomenology the starting point of ontology-epistemology, it is also the ending point, in the sense that no entity is introduced into one’s ontology unless it is given in experience,11 that is, unless it is picked out in one’s phenomenological description of the world of experience. Rosenberg has wrongly concluded that, since Bergmann introduces arguments for his ontology, therefore he has abandoned phenomenology and the Principle of Acquaintance.12 This is connected to another error in Rosenberg’s discussion of Bergmann. Phenomenological description is of phenomena; phenomena are known incorrigibly; what is incorrigible is known non-inferentially; only the corrigible requires inference; but argument is inference; so, Bergmann’s arguing testifies that the knowledge he claims to have is not incorrigible and therefore not phenomenological. At least, that is what Rosenberg seems to think. However, that such reasoning is wrong is clear. One should perhaps add, however, that philosophers are human too, and are quite capable of convincing themselves that they are not acquainted with what is in fact presented to them. In this sense, dialectics may contaminate one’s phenomenology. Is there any guarantee that anyone’s phenomenology is not contaminated? The answer is No: there is no guarantee. The best we can do is to be ever vigilant. As Allaire once put it, “One cannot do more, one must not do less.” 13 This does spot, however, an important role that dialectical arguments

542 can play in the philosophical enterprise. Philosopher B might well succeed in directing the attention of S to a certain entity, but S’s own philosophical reflections convince him or her that there entities are philosophically irrelevant. Or perhaps S’s reflections may lead him or her to a different philosophical description of those presented entities. Thus, for example, some would, for their own reasons, refer to as “places” what others call “particulars.” Suppose B is among those who speak of “particulars” and that S is among those who speak of “places.” In such circumstances, B’s dialectical arguments have the role of aiming to convince S that the entities to which B is attending have a certain philosophical relevance. Moreover, it is even possible that S’s reflections have had the result of convincing him or her that the entities to which B is attempting to direct his or her attention do not exist; in that case S’s reflections will have convinced him or her that he or she is not presented with what is in fact presented to him or her. In this latter sort of case, B’s dialectical arguments will have the further aim of challenging those arguments that have led S to deny the phenomenological claim of B. In this way, B’s argument does aim to convince S to, as it were, introduce a “sort’ of entity into his world, and not just his ontology but the world as he or she pheomenologically describes it. But again, such dialectical argument is not a form of transcendental argument. It is not an attempt to introduce into his or her ontology an entity which is not presented but is rather is an attempt to get one to recognize an entity which is presented.14 None of this is puzzling, but – again – clarity in these matters is important if we are to understand correctly one who insists upon a phenomenological method in philosophy. “All awareness is propositional.” Bergmann has made this phenomenological claim a central pillar of his ontology and epistemology.15 It is a claim that is undoubtedly correct. In thought, in perception, in sensing, what is given to us are complexes . In other words, in order to give a phenomenological description of the whole of what is given to us in a perceiving, or in a thinking, or in any sort of experiencing, one must use a sentence. If we say that the whole of what is given to us in an experiencing is the intention of the experiencing, then the intention of an experiencing is always propositional, represented by a sentence, and not by a name standing alone. But, of course, to say that – to say that the whole of what is given to us in an experiencing is a complex – is to say that the parts of such complexes are also given. Otherwise they would not be complexes. So, we are presented with complexes and also with the parts that are in those complexes. As Bergmann once put it, “To be presented (to me) is the

543 same as to be ‘in’ the intention of an act [of experiencing](of mine).”16 This of course requires that we be able to distinguish the parts of the complexes that are given in experience. We must be able to attend to the parts – not, naturally, as separate entities (though they may be separable), but at least as parts. Otherwise what grounds would we have for including them when we give a phenomenological description of the complex that is the intention of the experiencing? All this is obvious enough. And it is equally obvious that, for one like Bergmann, who proposes to philosophize in accordance with a Principle of Acquaintance, the experiencings through which we distinguish the parts of the presented complexes must themselves be ordinary sorts of experiencings – experiencings that are themselves experiences and not forms of experience, like Plato’s “reminiscience” of Forms, that are introduced into one’s epistemology solely on the basis of transcendental arguments. Moreover, if the complex I am presented with is a red square, then I can clearly distinguish the colour and the shape. That makes it obvious that the experiencings in which we distinguish the parts of presented complexes are ordinary and are themselves experienced. So, when Bergmann insists that the parts of presented complexes are also presented, he is not thereby introducing a new sense of ‘presented’ or ‘acquaintance’, contrary to what Rosenberg suggests,17 and certainly not any kind of acquaintance that is incorrigible in the foundationalist sense. Now consider a red square and a red circle. (Assume that they are the same shade of red.) In each of these two complexes we can distinguish a colour. The colour in each is what we represent in our phenomenological description by the word ‘red’. The word ‘red’ is applied to each of the two complexes because the colour in the one is indistinguishable from the colour in the other.18 It is this indisdinguishability that accounts for ‘red’ applying to both complexes. That is, it is this fact that accounts for the sameness of the two complexes. Assuming that the distinct are distinguishable, it follows that the colour of the one is the same as, not distinct from, the other complex. The colour red is distinguishable from the shapes. It is also, of course, distinguishable from other shades of colour. It is thus distinct from all these other entities, But the red in the one complex is not distinguishable from, is the same as, the red in the other complex. A distinguishable entity that solves the problem of sameness by virtue of the fact that, as the same entity, it can recur in several complexes, is traditionally called a universal.19 Similar arguments will of course show that the other colours, the various shapes, and so on, are all universals. And we are therefore acquainted with universals: they are presented parts of presented complexes.

544 Again, none of this is problematic. It is worth noticing, though, that in the discussion one moves quickly from phenomenological description (“the red in the one complex is indistinguishable from the red in the other”) to dialectical argument (“the fact of this indistinguishability provides the solution to the problem of samenss”) to categorizations that are justified by those dialectical moves (“the red in the two complexes is a universal”, “since wee are presented with red, which is a universal, we are therefore presented with universals”). The point is that such a transition does not imply that one has therefore somehow abandoned phenomenology for philosophy, where the latter is contrasted to phenomenology, perhaps identified with transcedental ontology: the entities described at the end of the dialectical argument are precisely those that are described in the phenomenological description with which one begins. Bergmann has, of course, made these points about properties and universals.20 The above discussion is misleading only in the brevity of the dialectic that it incorporates. For in fact, the dialectic can and must go on much longer than the above sketch suggests. Nor does Bergmann neglect that dialectic. He regularly was concerned to argue that those who deny that there are universals are mistaken. In particular, he was concerned to argue that the nominalism of one like Quine was mistaken.. Let us say that the sentences which describe the complexes with which we are presented describe facts. Then, for Quine, facts are “ultimate and irreducible”: that facts with which we are presented, in perception at least, are not really complex, they have no distinguishable parts. That makes Quine a “fact ontologist” rather than a “thing ontologist,”21 that is, rather than one who, like Bergmann, holds that presented facts have distinguishable parts to which one must appeal in order to solve various philosophical problems. As Quine puts the point: One may admit that there are red houses, roses, and sunsets, except as a popular and misleading manner of speaking, that they have anything in common. The words ‘houses’, ‘roses’ and ‘sunsets’ are true of sundry individual entities which are houses, roses and sunsets, and the word ‘red’ of ‘red object’ is true of each of sundry individual entities which are red houses red roses and red sunsets; but there is not, in addition, any entity whatever, individual or otherwise, which is named by the word ‘redness’, nor, for that matter, by the words ‘househood’ ‘rosehood’, and ‘sunsethood’. That the houses and roses and sunsets are all of them red may be taken as ultimate and irreducible, and it may be held that McX is not better off, in point of real explanatory power, for all the occult entities which he posits under such names as ‘redness.’22

545 Suppose that we are presented with two red spots. We describe one using the sentence ‘This is red’ and the other with the sentence ‘That is red’. For Quine, ‘red’ is a common name. We need not rehearse the problems inherent in this notion,23 nor in the similar notion of a predicate being “true of” something.24 The crucial point is that the common name ‘red’ applies non-arbitrarily to this and to that. That is, there is something about each of the two spots as presented that justifies applying ‘red’ to both, as they are presented (which may not be veridically), rather than, say, ‘red’ to the one and ‘green’ to the other.. This something is, of course, the fact to which Bergmann also appeals, namely, the fact that, with respect to colour, the two spots are indistinguishable, that is, the same. Nonetheless, the spots are two: they are distinguishable. Yet at the same time, according to Quine, neither do the two spots have any distinguishable parts. So, according to Quine, each spot is an unanalyzable and irreducible entity that is both the same and different from the other spot. The two spots in their simplicity, the pair and nothing else, ground the two relations of sameness and difference. Since the pair itself grounds these relations, they have been called internal relations. The nominalist has two internal relations, one for difference – here he or she agrees with the realist – and one for sameness – where he or she disagrees with the realist, who accounts for the sameness in question not by another internal relation, but by a shared part. Now, we contrast the complexity of a sentence with the simplicity of a name, that is, of a subject term or a predicate term. In this sense, for Quine the sentences ‘This is red’ and ‘That is red’, used to describe the spots, are misleading in their complexity. The spots, being simple, would be more perspicuously represented by names rather than by sentences: then the grammatical simplicity would reflect the ontological simplicity. Still, the names ‘this’ and ‘that’ won’t do, since that will serve only to represent the difference of the two spots. Rather, the names must be such as to, by themselves, represent both the sameness and the difference. The closest we could come would be ‘thisred’ and ‘that-red’, or, more generally, ‘red1’, ‘red2’, ‘red3’, ..., and so on.25 Simple things which, without internal complexity, are both the same and different from each other, and are most perspicuously represented by internally complex names, have been referred to as “perfect particulars” by Bergmann and more recently by others as “tropes.” Thus, the phenomenologist uses sentences to describe presented facts, but these facts are held by the fact ontologist to be unanalyzable, and turn out, upon his or her view to be tropes. The doctrine of tropes is the core idea of one version of

546 nominalism, the thesis that there are no universals. A fact ontology like Quine’s is therefore nominalistic. The disagreement between Quine and Bergmann, between the fact ontology of the former and the thing ontology of the latter, is whether the presented facts are really are simple, or whether such facts have distinguishable parts. But consider Quine’s examples, red houses, red roses, and so on. Or, more simply, a red circle and a red square. These would be named by names like ‘this-red-house’ or ‘this-red-square’ or ‘this-redround’. And surely, as a matter of phenomenological fact, no matter the device used to represent the red house or the red square in language, whether it be by (internally complex) names or by sentences, it is still the case that the red house and the red square are presented as complex and that by virtue of which ‘red’ applies to any such complex is part of and is presented as part of that complex. The nominalist uses names like ‘red1’ or ‘this-red’, but these names have an internal complexity that represents the phenomenologically given complexity of what those names are used to represent.26 Construing the sentence ‘This is red’ as the complex name ‘thisred’ cannot deny the complexity of the fact represented by the sentence; all it does is represent it one way rather than another. This far from exhausts the dialectic of the universals vs. tropes issue, but it does suffice to sketch, however broadly, certain crucial features. I have done this, not for its own sake, but rather to try to illuminate some other aspects of Bergmann’s ontology. Consider once again our red circle. This is a whole, and is a complex whole. This whole has a colour, it has a shape, and it has an area. The colour so to speak fills the area that the shape as it were surrounds. This area is part of the presented complex, and it too is presented. Now, it is clear about areas that, at least so far as we can tell, for each perceived complex there is exactly one area that is part of it, and that the same area is never part of two distinguishable complexes. This is a general fact about perceived complexes: colours, shapes, and indeed all parts of such complexes, save areas, recur, while areas, alone among the parts of complexes, do not recur. If the problem of difference is that of discovering that by virtue of which two presented complexes are correctly judged to be two, then, as Bergmann argues,27 dialectically, areas can reasonably be claimed to solve that problem: one complex is distinguishable from another complex just because it contains an area distinguishable from every other part in it and distinguishable from every part, including the area, that is in the other complex. But something which is in a perceived complex, and which is not

547 a universal, and which solves the problem of difference, is traditionally called a particular. So areas are particulars.28 And since areas are presented, so are particulars: particulars are given in perception. Such at least is the case, which does not seem unreasonable, that Bergmann makes. However, we need also to note that red (that is, the specific shade of red that we are presented with) turns out itself to be complex: red has as a part the feature of being a colour – being a colour is a quality which red has, just as red is a quality of the spot we are considering. This distinguishable feature of (the shade of) red is shared by (this shade of) red with all other shades of colour; it recurs in blue, green, scarlet, etc. Similarly, the property of being a shape recurs in all the specific shapes, that is, in all the square shapes, in all the circular shapes, in all the pentagonal shapes, and so on. We are, in short, distinguishing properties and properties of properties. Now, just as colour is a distinguishable property that recurs in, and is common to, all colours, so there is a distinguishable property that recurs in, and is common to, all areas, to wit, the property of being an area. This property is presented to one, just as the property of being a colour is presented to one. It is by virtue of this property that we can identify areas as areas; it is this property that solves the problem of sameness with respect to areas, that is, it is this property that accounts for the facts that all areas are the same, in respect of being areas, just as the property of being a colour accounts for the fact that all colours are the same, in respect of being colours. It is by virtue of this property of being an area that we can formulate the rule for sentences like ‘This is red’ that the subject term is to be used to refer to particulars, i.e. areas. In such sentences, the subject-term, ‘This’, refers on each occasion of its use to a distinguishable area, but nonetheless on each occasion when it is used, it is used to refer to an area, some area or other. Or, if we use the customary standardization instead of the tokenreflexive device, where each ‘a1’, ‘a2’, ‘a3’, etc., names a distinct particular, i.e. area, then the fact that each of the ‘ai’ refers to an area is reflected in each of these names having that a-shape in common, while the fact that each refers to a different area is reflected in each such name having a different subscript. If we decide so to speak that the property that is shared by all and only particulars is the property of particularity, then, since areas are particulars, and since that property which all and only areas have in common is the property of being an area, it follows that the property of particularity is the property of being an area. The basis for identifying particularity with

548 the property of being an area is, of course, the previous dialectical argument that areas are particulars: once the case is made that areas are particulars, that is, are the entities that solve the problem of difference, then the identification follows trivially. And it also follows, of course, that, since areas are presented and are presented as being, all of them areas, the property of being an area is presented, so also we are eo ipso presented with the property of particularity. The case for particularity as an entity in one’s ontology that I have just presented is, of course, just the case that Bergmann presented in his classic discussion in his “Ineffability, Ontology, and Method.” He first29 considers a certain complex, to wit, a green spot. Then, within the fact that the spot is green he distinguishes two entities, the spot and the character that is the colour of the spot. He next proceeds to put these entities into the appropriate dialectical context. On the one hand, this shows the relevance of the phenomenological distinction to the solution of philosophical problems. On the other hand, it has a strategic importance in convincing others of the relevance of these entities or even that they are presented with them. Bergmann next turns to the main point of the essay: “When I know that this is a green spot, I also know that (1) the spot is an individual, (2) the color is a character ... How could I know all this if it were not, in some sense, presented to me?”30 The rhetorical question simultaneously makes the phenomenological point – individuality or particularity, and universality, are presented – and introduces the dialectical context – how could I know this unless ... ? But this does not mean that Bergmann has here slipped from phenomenological description to transcendental argument. Rosenberg has completely misunderstood this passage: “Where the PA would lead us to expect reportage – ‘This and that are presented’ – what we find instead is clearly an argument.”31 Why instead? We find phenomenological reportage and an argument. Specifically, we have a dialectical argument. This argument is noted by Rosenberg, who describes it in this way: “It is, in fact, a transcendental deduction in the style of Kant, for its major premise is that something is known and its conclusion that, in order for such knowledge to be possible, something else must be the case.”32 Dialectical, certainly, but transcendental, certainly not. And certainly not Kantian. Rosenberg does get correct the premise and the conclusion of the dialectical argument: if one is to identify particulars as particulars, then there must be a feature of particulars that recurs. Phenomenological description has already yielded such a feature. What the dialectical argument reveals is the relevance of this fea-

549 ture to the philosophical problem of sameness: what is it about things that permits one to identify them as the same? that is, the same sort of thing. The argument does not have as its conclusion, however, that some nonpresented entity exists or must exist. And, of course, it would have to have the latter sort of conclusion if it were to be what it is not, a transcendental argument, after the fashion of Kant (though he was hardly the first one to use such a style of argument – Plato and Aristotle also come to mind!). So Rosenberg’s charge33 that Bergmann is here violating PA and the phenomenological method is just mistaken. As for his question, “With adequate dialectics available, why the putative phenomenological appeal at all?”34 The answer is obvious: dialectical argument can merely implicitly define its terms, give them a syntax; alone it can give them no content, no semantics. For those terms and those arguments to acquire content, they must be hooked up to the world – to the world, the world that is given to us in our ordinary experience. To achieve such a hook-up, to give our dialectics content, is the task of phenomenology. Phenomenology without dialectic is empty; dialectic without phenomenology is blind. The students of Husserl make clear the former, the students of Sellars the latter. Bergmann was later to point out35 with respect to the particularparticularity connection a significant difference between it and the connection between particulars and universals. Any particular is not only distinguishable from but is also separable from any universal. If ‘a1’ names some particular, then we can think of a1 as separate from any specific colour, shape, pitch, etc. For example, we can think of a1 as separate from the property red. That is, we can form the thought that would be expressed by the sentence: ~(a1 is red) Indeed, we can attend to a1 itself without thinking of any specific colour, shape, etc., at all; we can think of a1 in isolation from all such universals. For, we can from the thought that would be expressed by uttering the sentence: (›x)( x = a1 ) which is the thought that Hochberg and Bergmann once, not inappropriately, dubbed “the concept of a1”.36 However, the particular is not in these ways separable from particularity: these points – about which more later – Bergmann expresses in the claim that a1 is inseparable in thought from the property of particularity.37 This, as Bergmann points out, is a crucial phenomenological difference between particularity, on the one hand, and (other) universals on the other.

550 This “can’t” is a fact about our capacities to think certain things. This fact finds its reflection in the language we use to describe the world, or, at least, in any such language that is systematically developed. Consider the difference between a1 and a1 is red That a1 is a particular is expressed by the shape of the name, while that a1 is red is represented by the juxtaposition of two names to form a sentence. What cannot be in thought separated from particulars, we represent though an internal complexity in the name itself; what is with a particular but can be thought apart from it, we represent by means of a sentence. An inseparable feature of the particular is represented in the logical form of its name: a separable feature is represented by a sentence. The limits of the thinkable are reflected in the logical forms we ascribe to things. This move concerning the logical forms of names must carefully be distinguished from a similar move by the nominalist, e.g. Sellars.38 As we have noted, the latter represents facts by internally complex names such as red1 which may be said to make red the logical form of the name. But for the nominalist the claim is that there are no distinguishable parts in the fact represented by ‘red1’ or ‘this-red’. Here the notation represents indistinguishability, the indistinguishability of the red and the this in the this-red. From our or Bergmann’s (non-nominalist or realist) perspective, however, the particular and particularity are distinguishable, and what the notation represents is inseparability. For the nominalist, the two “parts” are indistinguishable, for Bergmann the two parts are genuine parts, and are distinguishable, but are inseparable: the contrast is between indistinguishable, and distinguishable but inseparable. In adopting a Principle of Acquaintance as his primary rule of method in ontology, Bergmann is identifying what exists, in the sense of ‘exist’ relevant to ontology, with the thinkable.39 Moreover, he goes on to point out, there are aspects of the thinkable that are not sayable.40 In locating ontology within the realm of the thinkable rather than the realm of the sayable, Bergmann is, of course, disagreeing with the author of the Tractatus. For a sentence to be part of the sayable, then, as the Tractatus would have it, it must be possible both to assert that sentence and to deny it. Now, clearly, in some sense of ‘can’t’, we can’t deny a necessary truth; to the contrary, we must assert it. Furthermore, also in some sense of ‘can’t’, we can’t as-

551 sert or deny an ill-formed sentence. The author of the Tractatus wished ‘can’t’ to be absolute; he wanted its sense to be univocal. This led him to assimilate necessary truths to formation rules and contradictions to illformed sentences.41 This, however, is clearly wrong: while we can’t even entertain an ill-formed sentence without assenting to its falsity, we, quite clearly, can entertain a contradiction without assenting to its falsity – as , for example, when we say that an argument with contradictory premises conforms to a criterion of validity that an argument is valid just in case that, if its premises were true, then its conclusion would also be true. Moreover, while we can’t deny a necessary truth, we certainly can say it. So that can’t be the sense of ‘can’ or ‘possible’ in which the “possibility of assertion and denial” defines the limits of the sayable. The point has to be that, while we can’t deny a necessary truth, we can entertain its negation. This suggests the following criterion for unsayability: if merely entertaining ~p requires us to assert p then p is not sayable. Bergmann connects the sayable and the separable.42 Consider the sentence (1) This is P This ought to be denied only when that to which ‘P’ refers is not that, or part of that, to which ‘This’ refers; which is to say that (1) ought to be denied only when This and P are separate. Conversely, if merely entertaining the separateness of This and P requires us to assert that they are not separate, then (1) is not sayable. Notice, though, that even if (1) is not sayable, it is nonetheless thinkable, in the sense that we can believe it, have it as an intention as a species of experiencing, and in the sense that we can entertain its negation. Now, Bergmann holds that (2) This is a particular cannot be said, since the particular which is the referent of ‘This’ is inseparable from particularity. But, as we have argued, (2) represents the same presented complex that (2') This is an area represents. It follows that neither will (2') be sayable. Given the definition of what is sayable, this is not problematic. It does, however, raise problems about phenomenological descriptions. For, on the one hand, we do want to distinguish particularity from particulars, and to do this we must use (2) or sentences to the same effect; while, on the other hand, it would seem that we cannot use sentences like (2) = (2') to say anything. In order to resolve this problem – which seems to cut to the heart of the idea that ontology is

552 at bottom a phenomenological enterprise – we must look more closely at Bergmann’s account of thought. We may begin our discussion of Bergmann on thought by considering the schema (3) +p,Mp of his classic essay on “Intentionality.”43 ‘p’ is any well-formed sentence; ‘M’ represents the “means” of intentionality; ‘+ p ,’ is the name of a character – a propositional character or thought – and specifically it is the name of the thought that means the state of affairs p. A sentence of the form (3), Bergmann proposed in that essay, is analytically true while one of the form +p,Mq where ‘p’ and ‘q’ are different sentences is analytically false. This analysis may be seen as deriving, under certain dialectical and phenomenological pressures, from some reflections of Frege. Frege took the standard analysis of thought at his time to be this: the meaning of a sentence ‘p’ consists of certain sense ideas in the mind – a sort of vulgar Lockean position. Frege44 also assumed a sort of nominalism with respect to sense “ideas.” These are apparently images such that the same sense idea could not be in two different minds.45The meaning of a sentence ‘p’ therefore cannot be public. But of course it is: several people can think the same thing. This leads Frege to advance a transcendental argument for a Third World, neither mental nor sensible, of meanings or senses.46 When a mind understands a sentence, it is related to the nonmental meaning that is the sense of the sentence: it grasps as it were the sense of what is said. Since senses are non-mental, the nominalistic problem is avoided: several minds can be related to the same sense. Meanings are therefore public. But they are also problematic. Let a person express what he or she is experiencing by uttering, say, ‘That chair is red’. We usually take this sentence to, on the one hand, express the experiencing, and, on the other, to describe the fact or state of affairs that is experienced. The experiencing expressed is of the fact described: the experiencing is mental, and the mind relates to, or, as Bergmann would say, following Brentano, intends the state of affairs described. This seems obvious, but it is in fact, on Frege’s view, not the way the world is. On this view, there is, on the one hand, the person’s experiencing. This experiencing is expressed by the utterance of the sentence. In uttering that sentence, the mind grasps its sense. This sense is not the fact or state of affairs experienced, nor the experiencing of it, but rather an entity in Frege’s Third World. Thus, on the other hand, what the experiencing relates to directly is not the red chair but

553 the sense of the sentence. So the sense or proposition stands as it were between the experiencing and what is experienced. The result is a version of representationalism. The sceptical problems that always accompany the introduction of transcendent entities immediately appear.47 These difficulties are obvious. But the more profound criticism is the phenomenological: Fregean senses or propositions are simply not presented to one.48 The Fregean sense solves the problem of the publicity of thought, but only by introducing entities which we do not experience either in our ordinary sensible perception of the world or in inner awareness, or, in other words, in our ordinary experience of things: Fregean senses are not of the world of ordinary experience, but are rather of the Third World.49 To avoid the sceptical problems consequent upon introducing these entities of the Third World, what one requires is that the meaning of the sentence be public, in the sense of being sharable by several minds, or by the same mind on several occasions, and that it also be found in experience. Consider again the experiencing that we express by uttering (a token of) the sentence (type) ‘That chair is red’. The experiencing is an event that – details aside – causes one to assert, or, at least, be disposed to assert the sentence ‘That chair is red’.50 Now consider a situation in which we see the same chair again. Once again we have an experiencing that causes one to be disposed to assert ‘That chair is red’. Here is a causal relation between the experiencing and the disposition to assert that sentence. What is relevant to this causal relationship is not the red chair nor a relation of the experiencing to the red chair.51 For, after all, both occasions may be hallucinatory (recall Macbeth’s dagger), and objects that are not actual, objects that seem to be real but in fact are not actual but merely possible, do not enter into causal relations. What is relevant to the causal relationship must be an element in such experiencings by virtue of which they both cause one to be disposed to utter the same sentence. But: “same effect, same cause.” The feature in the experiencings which causes one to be disposed to utter the same sentence must be such that the experiencings are of the same sort. So long as one is not a nominalist, this sort, this quality of he experiencings, can be in many experiencings, both mine and others. It is, in that sense, sharable, and therefore also in the relevant sense public – and also, unlike Fregean senses, is something, a quality, found in experience. One may therefore identify this quality or property of experiencings as the proposition that is the meaning that one expresses when one utters or is disposed to utter the sentence ‘That chair is red’. By rejecting Frege’s nominalism, one can locate a feature in experience that permits one to solve the problem he pro-

554 posed to solve by introducing transcendental entities. The transcendental move – and the consequent radical scepticism – is parasitic upon the nominalism, and once the latter is rejected one is in a position to solve Frege’s problems within the framework of the Principle of Acquaintance. It was G. E. Moore who first saw this clearly, when he identified propositions with properties – construed as universals – of acts of experiencing.52 Bergmann takes up this Moorean idea. He begins with the notion of the text of an awareness (experiencing): If one asks the proper question of one who has an awareness while he [or she] has it, one elicits a certain answer. If, for instance, somebody points at the tree while I am looking at it and ask me what I see, I shall say “This is a tree”. This statement is the text of my awareness.53

The text represents the state of affairs that is the content of the awareness. As for the sentence that is the text of the awareness, we may give its meaning in three different senses of ‘meaning’. In the sense of context (“meaningc”), the meaning of a sentence is, as Sellars has put it, ...understood in terms of the uniformities and propensities which connect utterances with (1) other utterances (at the same or different level of language), (2) with the perceptible environment, and (3) with courses of action (including linguistic behaviour). I say uniformities, but the uniformities are not mere uniformities, for they are grounded in rules in a way most difficult to analyze, but which involves the causal efficacy of rule expressions.54

Elsewhere he refers to (1) as “word-word” connections”, (2) as “worldword” connections, and (3) as “word-world” connections. Sellars so explicates ‘maning’ that the conceptual meaning or conceptual content of a sentence or expression is given by the word-word connections in which it stands to other expressions.55 That is, the conceptual content is given by the axioms that, as one says, implicitly define it. One may well wonder whether this is a felicitous explication of the conceptual content of an expression.56 In any case, for any one who works within the framework of a Principle of Acquaintance, it is another sense of ‘meaning,’ a more restricted sense, that is relevant to explicating the idea of the conceptual content of an expression: this conceptual content will be given by the reference (meaningr) of an expression, which is a matter (though in a complicated way) of the world-word connections that govern its use. Finally, of course, an expression has as its intentional meaning (meaningi)

555 the propositional character that causes one to be disposed to utter the sentence. E. B. Titchener held that all meaning is meaningc, denying that there is any meaningi. The only contents, he argued, were sensory contents, contents that could be described using terms that refer to colours, shapes, pitches, and so on. Thoughts were held to be verbal and other sensory imagery. The meaning of thoughts was to be analyzed on the model of the meaning of sentences as meaningc.57 Sellars defends the same position, and, as he has described it, “the framework of things is an analogical one the fundamentum of which is meaningful overt speech...”.58 But this won’t do. Upon this account, the meaning of a sentence cannot be a manifest sensory property of the sentence or the verbal image of the sentence, since the same sentence may carry different meetings – the sentence ‘Dass ist rot’ has one meaning for those who know German, another for those who don’t.59 Moreover, a small change in the sound or imagery may carry a vast difference in meaning − compare ‘Our son is dead’ with ‘Your son is dead.’ So the meaning cannot be a manifest sensory property; rather, upon the Titchener-Selllars view, it must be construed as a dispositional property – construed, in other words, as the propensities that are involved in the world-word, word-word, and word-world connections that define the meaningc of the sentence. However, as the research of the Würzberg psychologists showed,60 it is impossible on empirical grounds to construe the meaning of the sentence as anything but a manifest property of the mental state: it has to be the property the presence of which causes one to be disposed to utter the sentence, and that – the fact that it has, when it is present, causal effects – implies that it be a manifest property.61 This manifest property is, of course, the one that we have called the proposition, or, more accurately, the propositional character. We have, then this account of propositional characters: when the sentence ‘p’ is the text of a thought, then the propositional character that is the meaningi of the sentence ‘p’ is that character present in mental states that, when so present, causes one to be disposed to utter the sentence ‘p’. To elucidate this further, consider a parallel. Suppose that we have a disease D which we have reason to suppose is caused by some species or other of germ, without, however, our knowing specifically which species it is. This permits us to form the definite description the species of germ the presence of which causes D Let us abbreviate this by the phrase the D-bug

556 With this definite description ready to hand, we can state the causal law that, roughly speaking, goes like this (^) The presence of the D-bug causes D This is similar to the common explanation “He has the flu; it is caused by the flu bug”. (^) is, of course, a law, a synthetic generality, just as “Flu is caused by the flu bug” is a law. It is certainly not, as some suppose, an analytic truth. For, accepting (as I think we must) Russell’s account of definite descriptions, the statement that The cause of A causes A is to be construed as synthetic. The aim of medical research is, clearly, to identify the species of germ that causes D (or the flu). That is, the research aims to find a species S of germ such that the identificatory S = the species of germ the presence of which causes D is confirmed (or, to use our more quotidien example, the species S of germ such that S = the flu bug is confirmed).62 In the case of propositional characters, such research is, for obvious reasons, unnecessary: being the causes in consciousness of the dispositions to utter sentences, these characters are things with which we are perforce acquainted. Moreover, these propositional characters have a characteristic “feel” that distinguishes them from other characters of mental states. Bergmann correctly draws out attention to what he calls the “species” of a mental act,63 a sort of character that is always present when a propositional character is present and which disposes one to utter, depending upon which one of the sort it is, a specific mood of utterance. Thus, the species “believing” disposes one to utter the sentence which is the text in the asserting mood, “disbelieving” in the denying mood, “wishing” in the optative mood, “feeling of obligation” in the resolutive mood (if the obligation is on oneself) or in the imperative mood (if the obligation is on another), and so on.64 Propositional characters “feel” different from these species just as colours “feel” different from shapes. This “feeling” is, of course, nothing other than the fact that the property of being a colour recurs in all colours and in nothing else. Similarly, the characteristic “feeling” of propositional characters is nothing other than the fact that a certain property – call it the property of being a thought – recurs in all propositional characters and in nothing else. Now, since we are acquainted with the propositional character that causes one to be disposed to utter ‘p’, let us name this character ‘"’. We are, furthermore, acquainted with the character of being a thought. And we

557 are presented with the state of affairs that would be described by " is a thought However, as with so many characters, we do not in general refer to them by names but by definite descriptions. In the case of propositional characters, the definite descriptions that we use are given in terms of their content, that is, their texts For " we could use the definite description the thought the presence of which causes one to be disposed to utter ‘p’ Here we have the causal relationship (a) The presence of the thought the presence of which causes one to be disposed to utter ‘p’ causes one to be disposed to utter ‘p’ and the identificatory relationship (b) " = the thought the presence of which causes one to be disposed to utter ‘p’ Bu this is not yet quite correct; it is not yet exactly in line with what Bergmann means by ‘propositional character.’ There are in fact many sentences that we are often disposed to utter, that do not express “propositions”, in the sense in which Bergmann uses that term. For example – it is an old example from Russell –, we are at times disposed to utter Quadruplicity drinks procrastination This disposition is clearly caused by the presence of a certain specific sort of conscious state. This sort of state is indeed a thought. Yet at the same time, we may, not unreasonably, deny that that sort of state is a propositional character: ‘This is red’ expresses a proposition in a way that the sentence about quadruplicity does not. Again, ‘Dass ist rot’ and ‘This is red’ are different sentences and no doubt express different thoughts – such are the subtle difference in meaningc between languages. Nonetheless, there is a felt sense – an “intuition” if you wish (though that is a very misleading way of speaking) – in which these two sentences have the same meaning, that is, the same meaningi. The two thoughts, while different, are also the same, and it is this by virtue of which they are the same that is the propositional character that is meanti by each of the sentences. And it is the absence of a feature of this sort that one has in mind when one says that the sentence about quadruplicity does not express or meani a proposition. In order to get a handle on these things, we can start from the side of sentences. The problem is to classify sentences into propositional (‘This is red’) and non-propositional (‘Quadruplicity ...’), and then to so classify sentences as the same that, when they are the same they “say the same”

558 only in different words. The basic idea is, of course, that two sentences “say the same” if and only if their respective world-word connections are such that they refer to or represent the same state of affairs; two sentences “say the same” if and only if they have the same meaningr. One way of effecting this classification is by means of the sort of abstraction that has been called an “ideal language”. This is the technique that has been used by Bergmann. For present purposes we may take it that Bergmann’s ideal language L has the syntax of the (unramified) system of Russell and Whtiehead’s Principia Mathematica.65 The syntax is interpreted or given a semantics in a two-stage process. First, the logical words are interpreted – the connectives in the usual truth-table fashion and the quantifiers in the usual satisfaction fashion. And second, to this interpreted syntax are added descriptive constants that are interpreted by coordinating them to presented things, that is, by establishing rules of designation by virtue of which they refer to presented things – constants of type 0 to particulars, constants of type 1 to properties of (including relations among) particulars, and constants of type 2 to properties of (including relations among) properties – and in terms of these designation rules, truth-conditions for atomic sentences are laid down. These, together with the interpretations of the connectives and the quantifiers yield truth-conditions for all sentences of L.66 A sentence of L refers to, or represents, or meansr, its truth-conditions. In order to justify this choice of L one must argue that one could in principle use L to “say what one wants to say about the world” – which in this context, that of ontology, one could, in principle, use L to provide what phenomenological descriptions one needs for getting on with the task of ontology-epistemology. In such an argument the case is made that in principle of the world can be translated into L.67 This then yields the desired classification of ordinary sentences. If such a sentence can be translated into L then it is a proposition. And if two such sentences receive the same translation in L, then they “say the same”. Thus, for example, suppose that ‘p’ is a sentences of L and that ‘S’ and ‘S*’ “say the same” just in case they both translate into ‘p’. This may happen if ‘S’ is in one language and ‘S*’ in another, say German. Or, ‘S*’ may be definitional shorthand for ‘S’. So, for example, ‘He is a bachelor’ and ‘He is an unmarried male’ would receive the same translation in L. This shows, by the way, why the transcription into L can only be “in principle”: L contains no definitional abbreviations, a point that Bergmann came to recognize only late in ontological reflections.68 But in any case, if ‘S’ and ‘S*’ both translate into ‘p’ of L, then both ‘S’ and ‘S*’ mean what

559 ‘p’ meansr; or, as we may also express it, ‘S’ and ‘S*’ both meanr the state of affairs that p. Now, the two sentences ‘S’ and ‘S*’ express different thoughts. One is disposed to utter the one sentence rather than the other because the one thought is present in consciousness rather than the other. But in both cases one is disposed to utter sentences that have something in common: both sentences meanr that p. This “something in common” is, of course, a shared dispositional property, defined in terms of relevant world-word connections. The different thoughts can causally explain the presence of this shared dispositional property only if they too have something in common. And – we need not repeat the case again – this feature must be a manifest property that is present in the thoughts. We may say that the thoughts have flavours, and that different thoughts may have the same flavour in common, much as the same colours may have different intensities or hues. If different thoughts have the same flavour, then the sentences they dispose one to utter meanr the same state of affairs, to wit, the state of affairs that is meantr by the sentence in L into which they both translate. There is a one-one relation between flavours and referential meanings. The relation of thought to sentence is one-one; that of referential meaning to sentence is one-many. The relation of flavour to sentence is one-many; but that of flavour to referential meaning is one-one.69 Now, for Bergmann, a propositional character is not identified by its actual text but rather by its ideal text.70 In our way of talking, what correlates to actual texts are thoughts, and it is flavours that correlate to ideal texts. That means, then, that the properties that Bergmann refers to as propositions or propositional characters are what we have been calling flavours. Since propositions are identified by means of their ideal texts, it follows that the relation between a proposition and actual sentences will be onemany. In particular, two sentences that are definitionally equivalent in ordinary language will express the same proposition. Bergmann in his later papers71 pointed out the importance this point has in the detailed phenomenology of perception and in the solution of various philosophical problems concerning perception. D. M. Armstrong has also exploited the same point in a slightly different context.72 We can now take up the theme of the separable, the thinkable, and the sayable. We above mentioned the characters that Bergmann calls the species of an experiencing – loving, hating, wishing, willing, remembering, perceiving, and so on. For present purposes, two species are important, namely,

560 those of believing and disbelieving. If an experiencing is characterized by the species believing and by (the character which is) the proposition + p ,, then that experiencing causes one to be disposed to utter assertively a sentence that meansr p. Similarly, if an experiencing is a disbelieving, then that causes one to be disposed to utter denyingly a sentence of the relevant sort. Now, for the atomic sentences ‘p’ of L, we in general believe – or ought to be believe – with respect to the proposition for which ‘p’ is the ideal text, only if the states of affairs that ‘p’ meansr is given in a perceptual experience. The inferential basis for knowledge comes from our perceptual experience. It is of course necessary to distinguish those perceptual experiences that are acceptable and those that are illusory. We distinguish my perception of the desk upon which I am writing from Macbeth’s perception of the dagger. Let us take for granted that such a distinction can be made. Starting from this perceptual basis, there are rules, of a more or less complicated sort, that connect these believings to believings with respect to the molecular and quantified sentences of L. These rules will be rules of language, semantic and syntactical, rules of logic, and rules of the scientific method (supposing one to be rational73). For most of these cases, the believing (or disbelieving) does, and ought to, causally result from perceptual experience. There is one group of sentences, however, for which perceptual experience is not in the same way a necessary condition for evoking the attitude of believing. I refer, of course, to the class of necessary truths. Our L contains, thus far, no terms than mean ‘possible’ or ‘necessary’; they must, therefore, be introduced. In this connection one develops the syntactical notion of a sentence L being a tautology: a sentence is tautology just in case that it holds in all possible (non-empty74) universes.75 There are problems here, but at least for the first order language the idea is unproblematic.76 It is this syntactical notion of being a tautology that explicates the philosophically problematic notion of “necessary truth.” This justifies the following rule for the use of the necessity sign ‘~’ in L: ‘~p’ is true in L if and only if ‘p’ is a tautology77 It follows immediately that, if ‘p’ is a tautology, holding in all possible universes, then ‘~p’ also holds in all possible universes and is therefore also a tautology. It also follows that ~p e p is a tautology, and therefore also ~(~p e p) We define possibility in the usual way:

561 ‘p’ is true in L if and only if ‘~~~p’ is true in L This gives p e p and ~(p e p) as further tautologies. The ideas are familiar enough, so we need not develop them further. The basic idea is due to Carnap.78 Note, however, that the introduction of the modal terms into L does not support the philosophic task of explicating the problematic modal terms;79 rather, it is a technical move that presupposes that the philosophical problems have already been solved. Carnap, of course, always more impressed by technical moves than a philosopher ought to be, never quite saw this.80 For many years, Bergmann’s L contained no modal terms. If I understand him, then he came to the conclusion that it does require those terms. He tells us that “...what is analytic ... is not an expression of IL [ideal language] but the complex it stands for,”81 which must mean that we can introduce into L, Bergmann’s IL, for any analytic sentence ‘p’, a new term, ‘...is necessary’, which, when predicated of ‘p’ to give ‘p is necessary,’ represents the objective feature, to which Bergmann refers, of the complex that ‘p’ stands for (meansr). We now want to connect these objective features of states of affairs with our believings and disbelievings. Let us therefore sketch, as armchair psychologists, some of the principles that now go under the name of the “logic of belief,” that is, the axiomatic treatment of the lawful relations among propositions and the species believing and disbelieving. For example, just as colours are incompatible, so are the species believing and disbelieving. That is, it is a law that (B1) For any experiencing a, and any proposition + p ,, if a is a+ p , and also is a believing [disbelieving] then a is not a disbelieving [believing] or, to put it equally clearly but more briefly, For any proposition + p ,, if + p ,is believed [disbelieved] then + p , is not disbelieved [believed] We also have laws that go roughly like this: (B2) For any proposition + p ,, + p ,is believed [disbelieved] if and only if + ~p , is disbelieved [believed]82 But if believing + p , always excludes believing + ~p ,, the same does not hold for another species, that of supposing or entertaining. If I believe

562 a proposition, I cannot (as a matter of law) believe its contradictory, but I can still entertain the latter. The interesting case is that of necessary propositions. For most propositions, believing with respect to them is and ought to be causally evoked through some perceptual experience. But if we have before our minds a proposition to the effect that something is necessary, then that by itself is sufficient to evoke a believing. That is, for such propositions something like the following holds: (B3) For any experiencing a and any proposition + ~p ,, if a is + ~p , then a is a believing Propositions ascribing necessity to states of affairs thus by themsevles cause one to believe them. In this way such propositions constrain our capacities to believe and disbelieve, and this is, moreover, a felt constraint: the believing that is evoked by a proposition ascribing necessity to a state of affairs is one that is felt to be imposed, one cannot resist it even if one tries, that is, there is a feeling of necessity or inevitability about it. There is, as Bergmann has said, nothing logical about logic. In this sense, there is no logical must. This is indeed so, as far as concerns the truth of such propositions, Nonetheless, it is also a fact of psychology, or of the “logic of belief,” if you wish, that propositions ascribing necessity demand belief. In this sense, logic does constrain us; in this sense, there is a “logical must.” This feeling of constraint that we are talking about is a feature of the believing itself, and such a believing has been said by Bergmann, following Brentano, to be evident.83 Thus, a proposition ascribing necessity to a state of affairs evokes an evident believing. It is no doubt clear that besides evident believings there are also evident disbelievings. If ‘p’ is a contradiction, in the syntactical sense, of L, that is, a sentence that is false in all possible worlds, then ‘~p’ holds. ‘~p’ says that p is impossible. Then an evident disbelieving will be evoked by any proposition ‘~p’ that causes one to be disposed to utter a sentence that meansr that p is impossible. We need to record two further laws in our brief “logic of belief.” These describe processes of inference. The first concerns inferences from propositions that are believed: (B4) For any propositions + p ,, + ~ (p e q) ,, and + q ,, if + p , is believed and + ~ (p e q) , is believed, then + q , is believed. This is the modus ponens inference. A similar law will hold for disbelievings yielding the modus tollens inference. And other laws will describe

563 other forms of inference, corresponding to other elementary valid argument forms. We must also recognize that we have inferences from entertainings or supposings or assumings, described by laws like: (B5) For any propositions + p ,, + ~ (p e q) ,, and + q ,, if + p , is entertained and + ~ (p e q) , is believed, then + q , is entertained As the last one that we need look at, we should note that (B6) For any proposition + ~ p ,, if + ~ p , is believed, then + p , is believed with evidence If we assume that a person believes the tautology ~ ( ~ p e p) then (B4) implies that if he or she believes ~p then he or she believes p This means that if one has the mental state a which is a believing and which is characterized by the proposition + ~ (~ p e p) , and with it a mental state which is a believing with regard to the proposition + ~ p ,, then that mental state places constraints on the attitude one can have with regard to + p ,; it will be an evident belief – which is what (B6) asserts. (B6) accounts for how tautologies come to be believed with evidence. In order to specify L we need a list (l1) of individual constants ‘a1’, ‘a2’, ‘a3’, ... each of which is interpreted to meanr a distinguishable area; and we need a list (l2) of predicate constants ‘F1’, ‘F2’, ..., each of which is interpreted to meanr a distinguishable property of areas. We may, for present purposes, ignore relations and also ignore higher order properties and relations. Now, the Principle of Exemplification holds84: for every particular, there is a property which it has, and for every property there is a particular which has it. This is reflected in the formation rule for atomic sentences of L85: complexes in the world are to be represented by strings of the form ‘Fa’, where it is understood that something has the a-shape if and only if it occurs in the list (l1) of the individual constants that are interpreted to meanr areas, ands similarly for F-shaped signs. Once the atomic sentences are introduced in this way, the other sentences are introduced in the usual way. Thus, for example, we lay it down that if ‘p’ is any sentence, and in particular any atomic sentence, then ‘~p’ is also a sentence. There is a sense in which a string ‘p’ is truth-valuable – that is, is something which has one truth value or the other – if and only if ‘p’ is a wellformed sentence of L. We may therefore adopt Strawson’s term ‘presup-

564 poses’ and say that p presupposes S just in case that the statement that ‘p’ is a well-formed sentence of L entails S Let us now suppose that we have included in our list (l2) of predicate constants a predicate that meansr the property of particularity, that is, the property of being an area. We will then have a sentence in L that says (4) a1 is an area Let (5) Ga1 be any well-formed sentence of L that contains the individual constant ‘a1’ but not the predicates that meansr the property of being an area. It is evident that (5) presupposes (4). For, ‘Ga1’ can be a well-formed sentence of L if and only if a1 is an area: the definition of ‘well-formed sentence of L’ requires that each individual constant, and ‘a1’ in particular, meanr a thing which is an area. Thus, in using any sentence (5) about a1 which does not explicitly ascribe the property of being an area to a1 we are nonetheless still implicitly ascribing that property to a1, and in that sense (5) presupposes (4). In this sense, there is never any need to introduce (4) into L, and never any need to introduce a property that meansr the property of being an area, i.e. the property of particularity. As Bergmann once expressed it, the property of particularity can be explicitly introduced into L only with futility.86 We can already see one sense in which particulars cannot be separated from particularity. Consider the proposition dubbed by Bergmann and Hochberg “the concept of a1”, that is, the proposition that would be expressed by uttering the sentence (›x)( x = a1 ) This is a special case of (5). It therefore presupposes (4). In this sense, the concept of a1 cannot be separated from a1 being an area, that is, from a1 being a particular, or having particularity. But Bergmann makes a stronger claim: the proposition that a1 has particularity can be thought but cannot be said. The argument for the futility of introducing a predicate for particularity does not establish the unsayability claim. After all, if something is sayable only with futility, it is still sayable. We must see if we can make sense of Bergmann’s stronger claim. Consider once again the sentence (4) a1 is an area This has the logical form of an atomic sentence of L. This means that (6) ‘a1 is an area’ is a non-self-contradictory well-formed

565 sentence of L is a sentence about a sentence of L, and is true if and only if (7)  ( a1 is an area) which is its reflection in L, is true. (7) itself is a well-formed sentence of L, and therefore presupposes (4). We therefore have (8)  ( a1 is an area ) e a1 is an area But, the semantics of ‘’ make (6) and (7) logically equivalent, and the definition of ‘well-formed sentence of L’ means that (6) entails (4). Hence, the antecedent of (8) entails the consequent. Thus, (8) itself is a necessary truth, and we have (9) ~ [  ( a1 is an area ) e (a1 is an area )] Similarly, for the sentence (10) ~ ( a1 is an area ) we have (11) ~ [  ( a1 is an area) e ( a1 is an area) ] Thus, for ‘a1 is an area’ we have the principle ad posse ab esse: (9') ~(pep) and for ‘~(a1 is an area)’ we have the principle (11') ~ (  p e ~p ) ab posse ad nonesse. But if ‘ p e p’ holds for any ‘p’, then by transposition so does ‘~p e ~  p’. And by substitution and elimination of double negation, so does ‘p e ~  ~ p’, or, what is the same, ‘p e ~ p’. Thus, (9) commits us to (9") ~ [ ( a1 is an area e ~ ( a1 is an area) ] Similarly, if ‘ p e ~ p’ holds for ‘p’ then so does ‘p e ~  p’, and therefore also ‘p e ~ ~ p’. Thus, (11) commits us to (11") ~ [ ~ ( a1 is an area) e ~ ( a1 is an area ) ] Let us now suppose that one has an experiencing that is an entertaining of the proposition that + a1 is an area , Assume also that we have thought seriously about these matters and have come to entertain also that (9") is true. By (B3) we will therefore believe that (9"). But, by (B4) this will generate an inference that will lead us to entertain + ~ (a1 is an area ) ,. However, by (B3) that means that we will believe + ~ (a1 is an area) , and therefore, by (B6), that we will believe with evidence + a1 is an area ,. And the latter believing is one that causes us to be disposed to assert that a1 is an area. A similar line of reasoning will show that if we entertain the proposition + ~(a1 is an area) ,, then we will be led to believe with evidence that

566 + a1 is an area ,and again be disposed to assert that a1 is an area. Compare this to several other cases. Entertain the proposition + ~p v p ,: the “logic of belief” does not bring about the disposition to assert the tautology. Nor does entertaining its contradictory. Or, entertain + ~ p ,: the “logic of belief” leads one to believe that + p ,, to believe with evidence that + p ,, and to be disposed to assert that p. But we can still entertain that + ~ ~ p ,, for which the “logic of belief” would lead us, were we to think seriously about possibility and necessity, to entertain +  ~p ,, where the “logic of belief” would not lead us to believe that + ~ p ,, nor to believe that + p ,, nor, therefore, lead us to be disposed to assert that p. In contrast, the truths that relate possibility, necessity, and states of affairs, and the constraints on thought that we call the “logic of belief”guarantee that if we entertain a thought a particular is not an area, that is, that the particular is separate from particularity, then we are constrained to believe – we must believe – that the particular does have particularity: just to think it separate constrains us to believe that it is not separate. Moreover, if a sentence is sayable, it must be possible to deny it. What cannot be denied is not sayable. But have jsut seen that the truths that relate possibility, necessity, states of affairs and the laws of the “logic of belief” constrain our thoughts, and therefore the utterance of sentences expressing those thoughts, in such a way that one cannot deny particularity of a particular. Hence, as Bergmann says, while a particular having particularity is thinkable, it is not sayable. The problem we raised above was this. Given that (2) This is a particular and (2') This is an area represent the same complex, and if what these represent is not sayable, then how can (2) = (2') be used to give the phenomenological description that is a necessary condition for solving the philosophical problems surrounding particularity? Now, as we have just laid things out, ‘p’ is unsayable provided that, whether ‘p’ or ‘~p’ is entertained, in either case one is constrained to believe that p. However, it is also clear that, even if ‘p’ is in this sense unsayable, the proposition + p ,which its utterance expresses can be thought: it can be present in an act of entertaining or supposing. Thus, even if ‘p’ is unsayable, the complex that ‘p’ represents can still be the intention of an act of experiencing. So, what (2) = (2') represents can be the intention of an act of experiencing. And we can have terms – as in (2') – that represent different parts of the complex. In short, although the state

567 of affairs that (2') represents is unsayable, nonetheless the state of affairs can be represented in language, and recorded, therefore, in our phenomenological description. Of course, what we can do with such sentences in language is severely circumscribed by their being unsayable. Language is developed not merely to give phenomenological descriptions of the intentions of our experiencings; it is also developed in order to record which among those intentions we believe actually exist in the world. Language is developed not merely to record what propositions we can entertain about the world; it is also developed to represent what we believe to be the truth about the world. Now, one of the things that is important for most of our beliefs, is that we allow for their revision. After all, there is no foundation for empirical belief. Our language must therefore provide a mechanism by which, for every empirical belief, there is a sentence which not only records that belief, but also the negation of that sentence which then be available if we change our minds. Now, although we can entertain the negation of (2') – entertain it as an hypothesis – we cannot believe it. That is why (2') is unsayable. And since we can’t believe the negation of (2'), there is no possibility for our changing our belief (2), nor therefore any need to allow for the negation of that sentence in our language. We distinguish between the unrevisable framework beliefs, beliefs like that expressed by (2') = (2), and other, revisable, beliefs. The states of affairs for which we have unrevisable beliefs are represented in language by the logical forms of the names of things, while the states of affairs for which we have revisable beliefs are represented in language by sentences about things. In normal discourse, then, concerned as it is about discovering the truth, states of affairs in which we have unrevisable beliefs can be taken for granted, represented implicitly in the form of names, rather than explicitly in the form of sentences. Representing the state of affairs that (4) a1 is an area implicitly in the form of the name a1 has an important consequence for the negation (10) ~ ( a1 is an area ) of that state of affairs. We have seen that, if (10) is a well-formed sentence then it is false. In other words, (10) is excluded from (correctly) representing any state of affairs in the world by the very formation rules of the language in which it occurs. In that sense, (10) is an ill-formed sentence. If a sentence representing a state of affairs is not sayable, then its negation is

568 ill-formed. However, even ill-formed sentences can be used, and in certain contexts they must be used. That is, in those contexts, whatever they may be, we just use sentences that would, in belief-stating contexts, be construed as ill-formed. This, by the way, permits us to make an interesting point about the nominalist. He or she use the name this-red to name the red spot, and has the red-shape, which represents the redness of the spot, as the logical form of the name, just as for us the name of the particular a1 has the a-shape, which represents the particularity of the area, as the logical form of the name.87 It is thus part of the nominalist’s language that its formation rules guarantee the truth (12) This-red is red and the falsity of (13) ~ ( This-red is red) Thus, just as for us (4) is unsayable and (10) is ill-formed, so for the nominalist will (12) be unsayable and (13) ill-formed. Moreover, a parallel sort of reasoning will lead directly to the conclusion that, in the nominalist’s language the unsayable (12) expresses an unrevisable belief. Thus, the nominalist’s commitment to a language in which names like ‘this-red’ occur amounts to a commitment to the idea that beliefs in which properties are ascribed to particulars are unrevisable. We thus see clearly that nominalists like Sellars are committed to a foundationist account of empirical knowledge. They disguise this foundationalism by adopting the peculiar language that they do, the language with the peculiar names like ‘this-red’. Nor, if our above case – which is essentially Bergmann’s – is correct, then is nothing to recommend the nominalist’s ideal language L that a philosopher might adopt. But to return to our main theme: in belief-stating contexts it suffices to represent the unsayable implicitly. But, in ontological discourse we must make explicit the structure of these states of affairs, represent their distinguishable parts in language. Which is to say, we must introduce formally into our language a predicate that represents the property of being an area, that is, the property of being a particular. For, how else could we provide, for our ontological discourse, the indispensible phenomenological description of the state of affairs of, say, a1 being an area? To be sure, the adding of the predicate ‘is an area’ to language will, in the sense we have sug-

569 gested, be an exercise in futility. At the same time, although the use of sentences like ‘a1 is an area’ will be futile with respect to conveying information – the information is already implicitly there in the fact that ‘a1' is an individual constant – nonetheless the sentence is still needed in order to make that information explicit for the purpose of solving philosophical problems.88 But if we introduce into language a predicate that permits us to form sentences like (4) a1 is an area then we will also have sentences like (10) ~ ( a1 is an area ) which in belief-stating contexts must be considered ill-formed. And if we can have sentences like the latter, then we can at least entertain propositions that they can express. We can at least conceive that (10) obtains, entertain it, even if we can’t believe it. After all, what is a proposition but a feature of consciousness that causes one to be disposed to utter the sentence said to express it? We have uttered (10) – just writing it down constitutes a way of uttering it! We have uttered it, not asserting it, but treating it hypothetically. And if we have uttered it, then we were disposed to utter it. And the disposition has a cause in consciousness, to wit, the proposition + ~ ( a1 is an area ) , that was being entertained. So the negation of (4) is in this sense thinkable. Now, on this last point Bergmann disagrees.89 THIS IS THE FIRST OF MY DISAGREEMENTS WITH BERGMANN. For Bergmann, the state of affairs that (10) represents cannot be thought: not only can it not be believed, it cannot even be entertained. And if it cannot be entertained, that is because there can be no proposition intending that unthinkable state of affairs. But a proposition just is the property the presence of which in an experiencing causes one to be disposed to utter a sentence. When a sentence is available for use, then, there a proposition is possible, and the state of affairs that they sentence represents is thinkable. That means that a predicate meaningr the property of being an area, i.e. meaningr particularity, cannot be introduced into language, not even with futility. For, once such a predicate is introduced, that permits the formation of sentences that negate its holding of a particular. It follows that, for Bergmann, one can never in a sentence ascribe particularity to a particular. Ontological discourse about particularity must therefore be non-literal.90 That can only mean that this sort of ontological discourse can only proceed by means of metaphor and image. Bergmann has often used metaphors and images to discuss ontology, and

570 more often than not these are illuminating. Some of his metaphors about particularity are of this sort. But for us, these metaphors can be unpacked into non-metaphorical language; for us, they − metaphor and image − do not form the substance of ontological discourse. For Bergmann, in contrast, metaphor is the substance of ontological discourse. But it seems odd that he should say this. Why does he not recognize that predicate that meansr the property of being an area can – as we have done – be introduced into language, and therefore that sentences negating it of a particular can be introduced, and therefore that propositions negating it of a particular can be entertained and the separation of a particular from particularity can be thought if not believed? This DISAGREEMENT WITH BERGMANN is related to MY SECOND DISAGREEMENT, which concerns the relation between a propositional character and its content. Bergmann represents this relation, as we saw, by (3) +p,Mp where ‘+ p ,’ names the propositional character, ‘p’ represents its ideal text, and ‘M’ is read as “means.” It would seem, from what we have seen above of Bergmann’s discussion, that one way of reading (3) would be as (14) The presence of the proposition the presence of which causes one to be disposed to utter a sentence that meansr p causes one to utter a sentence that meansr p Understood in this way, the sign ‘+ p ,’ is a definite description; it is a notational abbreviation for the proposition the presence of which causes one to be disposed to utter a sentence that meansr p The corner quotes would represent the second-order property of being a proposition, and the occurrence of ‘p’ within them indicates in admirably brief but perspicuous fashion the sort of sentence that the proposition in question causes one to be disposed to utter. As for the sign ‘M’, this, it is evident, has to be understood as a shorthand way of referring to the causal relation between the propositional character and that which it causes, viz., the disposition to utter a sentence of a certain sort. Bergmann, however, does not construe (3) as (14). In the first place, he construes ‘+ p ,’ as a predicate, that is, as the name of a character rather than a definite description. And in the second place, he construes ‘M’ as representing a logical connection rather than as a causal relation; it is a logical connection between a character and a state of affairs. So construing

571 these things in these ways, Bergmann can provide, he believes, and as we saw, a syntactical sense in which (3) is analytic, a necessary truth:91 a sentence of the form + ... , M ... is analytic just in case that the sentence of L that comes after the ‘M’ is identical to the sentence placed between the corner quotes. This, however, is problematic indeed. For one, the rule for introducing the propositional predicates is general, and yields such a predicate for every sentence of L. But many such sentences one is not, not ever, disposed to assert, and therefore are the text of no propositional character: these sentence correspond to what might be called unthought thoughts. However, unthought thoughts aren’t experienced and therefore can’t be predicates of L. Hence, the rule Bergmann proposes for introducing the propositional predicates introduce many predicates that do not denote, many names that do not name.92 In this way Bergmann’s rule for propositional predicates is more like the rule for introducing definite descriptions than it is like any rule for introducing names (individual constants or predicates). For two, the propositional predicates are supposed to have all the properties of (undefined) predicates. So far as concerns syntax, the only structure that predicates like F1, F2, F3, ... etc., have is the difference between the F-shape (which represents the property of being a character, i.e. universality) and the subscripts that distinguish the predicates among predicates. If the propositional characters are predicates among predicates, they should have no more structure than this. So, upon Bergmann’s of propositional character names, ‘p’ is no more a part of ‘+ p ,’ than the word ‘cat’ is part of the word ‘cattle’. This the reason that one can’t quantify into corner quotes: for, strictly speaking there is nothing (proper) inside the quotes. But in fact the propositional characters do have more syntactical structure than ordinary predicates. It is precisely the internal structure that they officially do not have that permits Bergmann to give his proposed syntactical explication of the necessity of ‘+ p , M p’: we need to know what is between the corners in order to determine the analyticity of an ‘M’-sentence.93 Of course, if the propositional “predicate” is really a disguised definite description, then its internal complexity can be understood. But in that case, it doesn’t really occur in L – recall, L contains no abbreviations – in which case the syntactical explication of the necessity of ‘+ p , M p’ is destroyed. In any case, for three, ‘M’ remains in a special syntactical category of its own, neither a relation

572 (which connects things, i.e. individuals or properties, to other things) nor a connective (which connects states of affairs to other states of affairs), but a “pseudo-relation” connecting certain things, viz. propositional characters, to states of affairs.94 Bergmann came to recognize that there are indeed a series of problems here, and came to modify his view about the relation between a proposition and the state of affairs that is its content. To see where he goes, we can begin by noting that in (3) +p,Mp there is a redundancy. Either (3') (+p,,p) or (3") (FMp) could represent that there is a connection between a character and a state of affairs. In (3') the connection shows itself in the name of the character; in (3") the connection is represented as an external relation between a character and a state of affairs. Bergmann replaced the old notation by that of (3").95 At least, that is what the notation is officially. However, it is also explained that the connection is not external to either the character or the state of affairs, but is grounded in the pair, (character, state of affairs) and nothing else.96 This means, then, that unofficially Bergmann has opted for (3'). The move that he makes here is curiously similar to that of the nominalist who insists that the two red spots are simples – can only be named – and also that their sameness is grounded in the spots themselves – something which reflects itself in the internal complexity of the names. Just as for the nominalist, the colour of the spot becomes the logical form of the name ‘this-red’ of the spot, so for Bergmann, it would seem, that a proposition has a certain state of affairs as its content becomes part of the logical form of the name of the proposition. In any case, by making the relation of a proposition to its content part of the logical form of the name of the proposition, Bergmann has the formation rules of his language guarantee that a sentence that ascribes the content of a proposition to that proposition will represent the unsayable and express an unrevisable belief, and the negation of that sentence will be ill-formed. Thus, (3) +p,Mp will turn out to be a necessary truth. Bergmann was committed to the necessity of (3). His earlier attempt to explain the necessity syntactically failed. His later, revised view attempts to explain its necessity in terms of the formation rules of the language.

573 At the same time, however, one might well question on both dialectical and phenomenological grounds whether Bergmann succeeded. Dialectically, the names of propositions will stand out from all other names as involving complex patterns of logical form. That is, is it really true that the problems surrounding the special names ‘+ p ,’ have disappeared? Moreover, phenomenologically, propositions would seem to have – as Bergmann also insisted – a simplicity that these complex names deny. To be sure, the use of sentences to refer to propositions is natural, but that does not mean that those sentences must be construed as naming the propositions. In any case, Bergmann does wish to construe (3) +p,Mp as a necessary truth. Now consider or sentence (10) ~ ( a1 is an area ) and substitute it in (3). This gives (3^) + ~ ( a1 is an area ) , M ~ ( a1 is an area) A necessary truth is a truth, and a truth can be expressed only by a wellformed sentence. But if, as we have argued, (10) is ill-formed, then so presumably is (3^), at least in respect of the occurrence of (10) on the right hand side. Hence, so long as Bergmann construes (3) as a necessary truth, he cannot allow ill-formed sentences to substitute for ‘p’, and since (3) is the only schema he has for relating a thought to its content, it follows that, for Bergmann, ill-formed sentences can never express thoughts: the illformed is the unthinkable. It is for this reason, I suggest, that for Bergmann (10) cannot even be thought, and indeed cannot even be entertained. Thus, given Bergmann’s commitment to the necessary truth of (3), it follows that ontological discourse about particularity must be non-literal. In contrast, if we construe ‘+ p ,’ as a definite description, and (3) as the causal statement (14), then there is no reason why we cannot have thought the text of which are ill-formed sentences. The problem that Bergmann has with respect to ill-formed sentences no longer arises. We now have (c) (The presence of) + p , causes one to be disposed to utter a sentence that meansr p as our construal of ‘+ p , M p’. To say that Sentence S in English meansr p is to say that Sentence S in English has the same world-word connections that ‘p’ has in L

574 So ‘p’ does not itself occur in (c), nor therefore in ‘+ p , M p’, as it does on Bergmann’s construal of the latter. What occurs in (c) is the name of a sentence,97 and so long as an ill-formed sentence has a well-formed name, then that name can occur in (c) and therefore – albeit misleadingly – in the ‘p’ slot on the right hand side of ‘+ p , M p’. Our construal of this “M” sentence in terms of (c) thus permits one to have thoughts that are expressed by ill-formed sentences. Such a thought may well, as we have suggested, be flavoured differently from those expressed by well-formed sentences, so that we group the latter together as propositional and the former as, let us say, non-propositional. But non-propositional thoughts are still thoughts. Quadruplicity drinks procrastination may be ill-formed and in that sense not a genuine proposition, but we can still think it and entertain it. The question would therefore seem to be whether Bergmann has a good reason for holding (3) to be necessary. The reason that he gives in his essay on “Intentionality” is this. He considers the state of affairs that a is green, which he transcribes into his L as the sentence ‘gr(a)’. He then remarks that ...after ‘gr’ and ‘a’ have been interpreted, ... ‘+ gr(a) ,’ must be interpreted as the name of the character which my awareness possesses if and only if it is an awareness of the ‘gr(a)’ refers to. [That is to say, he recognizes here an objection we raised above concerning the aptness of construing ‘+ p ,’ as a name. But, he proceeds to argue, this anomaly is justified.] On the other hand, we would like to say that [this] is “merely a linguistic matter” or, as I once put it, that to be an awareness of a certain kind and to have a certain content (and, therefore, text) is one thing and not two.98

Now, it seems that there is some justice in Bergmann’s argument. Taking (a) and (b) as our model, we construe + p , as the proposition that causes one to be disposed to utter a sentence that meansr p, and let "=+p, In that case we have the law that (15) " causes one to be disposed to utter a sentence that meansr p What (15) asserts is that " constrains us linguistically, in our use of language, and indeed that it constrains us in a very specific way, namely, to utter a sentence with a very specific meaningr. In this sense, being an awareness of a certain sort and having a certain content are indeed inseparable. Yet the inseparability is causal rather than ontological. But, as

575 against Bergmann, while these are inseparable, inseparability in this causal sense does not justify construing +p,Mp as necessary in the sense of analytic. It would seem that Bergmann move illegitimately from One is constrained to utter a sentence that meansr p when + p , is present to One is constrained to believe ++ p ,M p , Moreover, as Bergmann so carefully brings out, when a thought + p , is present in consciousness, there is also very often a thought of a sentence that meansr p. I suspect that these further causal relations facilitate the illegitimate move just indicated. So here, then, is MY SECOND DISAGREEMENT WITH BERGMANN: he construes ‘+ p , M p’ as analytic where I would construe it as causal. We now turn to THE THIRD POINT OF DISAGREEMENT. Not surprisingly, it is related to the first two. Bergmann introduces into his L a sign ‘…’ to express the concept of difference, or, as he calls it, diversity.99 The contrary concept expressed by ‘=’ is that of sameness. The concepts, of course, the strict concept of sameness, and not the Russell-Leibniz concept in which individuals are said to be the same if and only if they all their properties (non-relational and, if there are any, relational) in common.100 For clearly, it is possible (in the sense of: not self-contradictory) that two different individuals (strict sense) should have all their properties in common, and therefore be the same (in the Russell-Leibniz sense).101 The basic semantical rule for interpreting the names (individual and predicate constants) of L is that of “one name-one thing”. Given this rule, it immediately follows that a1 … a2 is true in all possible worlds and therefore is necessary, while a1 = a2 and a1 … a1 are false in all possible worlds and therefore are self-contradictory. These things being so, the “logic of belief” [see (B3) and (B6)] entails that if the proposition + ~ ( a1 … a1 ) , is present, then this causes one to believe with evidence the proposition

576 + a1 … a2 , and to disbelieve with evidence + a1 = a1 , We can, however, entertain both this proposition and also the proposition + ~ ~ ( a1 … a2 ) , without being constrained to believe + a1 … a2 ,. In this case, the mere entertaining of + ~ p ,does not constrain us to believe + p ,. It follows that the state of affairs represented by a1 … a2 is sayable, unlike the state of affairs represented by a1 is an area It should be noted, however, that it is only sayable with futility, since, given the rule for the semantics of L, any well-formed sentence R(a1, a2) containing ‘a1’ and ‘a2’ will presuppose a1 … a2 Diversity already shows itself in the difference of names. The diversity of two things is grounded in the things themselves – in the pair and in no third entity. Diversity is thus an internal relation. Besides the diversity of individuals, we also have the diversity of properties: red … green and the diversity of things and properties: a1 … red However, the property of particularity is not diverse from particulars.102 In this sense, for Bergmann, a particular is not different from the particularity in it. But at the same time, he also insists,103 neither is a particular the same as the particularity in it. The concepts of sameness and difference (diversity) simply do not apply to the pair { a1 , particularity } where ‘a1’ names some particular. To try to say either a1 … particularity or a1 = particularity is to try to say the ill-formed. As a consequence, in his ontological discourse about the connection between particulars and particularity Bergmann resorts to metaphors rather than literal discourse: the particular is, he tells us, “pervaded by particularity”,104 and the two, which are not really two, but not really one either, do not form a whole which is complex but

577 rather a “two-in-one”.105 It is easy enough to see, in the light of our discussion of the first two disagreements with Bergmann, why he must say that a particular and particularity are neither the same nor different, and why he must therefore resort to non-literal discourse to make his phenomenological point. For, if particularity were to be different (diverse) from a particular, say our friend a1, then we should be able to say in L a1 … particularity But in that case ‘is a particular’ would appear as a predicate in L, which would in turn mean that both the sentences ‘a1 is a particular’ and ‘~ ( a1 is a particular )’ would appear as sentences. The latter, however, is to be construed as ill-formed. But if it could appear as a sentence in L, then the state of affairs it represents would be thinkable. Or, in short, if particularity were said to be different from the particulars it pervades, then the ill-formed would be thinkable. On Bergmann’s account, it would seen that the pair { a1 , a2 } does more that by itself ground the diversity of the two particulars. It would seem that the pair also by itself grounds the sameness of the particulars, that is, the fact that they are the same in respect of being particulars. Moreover, to repeat, they ground this sameness not by virtue of a shared part – for that would make them complex – but simply by themselves and in their simplicity. It would seem, then, that the pair grounds not merely the internal relation of diversity but also a second internal relation of sameness. Nor is this idea unfamiliar to us. For, it was the nominalist who held that the pair {this-red, that-red} or {red1 , red2} by itself grounds the internal relation of difference, by virtue of which there are two red’s, and also a second internal relation of sameness, by virtue of which the two are both red’s.106 However, if – as I think it does – Bergmann’s case against nominalism goes through, then it would seem that the same case can be made against his “two-in-one’s”. Consider the “two-in-one” in which particularity “pervades” a given particular. This state of affairs would be represented – with futility, and not in a sayable way, for Bergmann – by the sentence (2) This is a particular

578 which, as we have argued, must represent, for Bergmann, the same presented sate of affairs that (2') This is an area represents. (2') is a phenomenological description. The identification of (2') and (2) is the upshot of a dialectical argument developed subsequently to the giving of this description. So the question about “two-in-on’s” arises for that which is described in (2'). Now suppose that the referent of ‘this’ in (2') = (2) occurs in a complex which also contains the colour red, and consider (16) Red is a colour The case against nominalism directs our attention to the distinguishable property referred to by the predicate ‘colour’ which is a feature of the property red and which recurs in other colours. Similarly we have the distinguishable property referred to by the predicate ‘area’ which is a feature of This and which recurs in other areas (particulars). It is just this distinguishable property of being an area that we are led by the dialectics to identify it as the property of particularity. The point to be made against Bergmann is that (2') no more represents a special sort of complex, a ‘twoin-one’ than does (16). In this sense, being an area, or, what is the same, particularity, is a property among properties, and like other properties is a distinguishable part of any entities in which it occurs or recurs. Contrary to Bergmann, there is nothing special about particularity, and there is no need to introduce a suspect category of “two-in-one’s”. But this counts as an argument against Bergmann’s ontological claims about “two-in-one’s” only if difference ( … ) is identified with distinguishability. Bergmann was at one point inclined to make this identification.107 At that point he was willing to say that a particular and particularity were diverse. Later, however, he came to hold that they are in fact not diverse. On Bergmann’s later view,108 particulars are diverse from other particulars, from universals, and from states of affairs; universals are distinct from other universals, from particulars, and from states of affairs; and similarly for states of affairs. These entities are separable from each other But particularity is not separable from the particulars that it characterizes. Neither is it different from them, Bergmann claims – that is the “two-in-one” business. The point is that Bergmann makes the inference: not separable so not different. Where for us, there is no identity of difference and separability, we identify difference with distinguishability; in contrast, Bergmann identifies difference with separability. THIS IS OUR THIRD DISAGREEMENT WITH BERGMANN..

579 Ought Bergmann’s identification be accepted? Or ought one to reject such an identification, as we do, and accept, as we also do, an identification of difference with distinguishability? Clearly, this is a matter of dialectics: one must argue that the option one chooses can solve the philosophical problems. It would seem that Bergmann’s identification does run into difficulties. I shall mention but three points. FIRST. Bergmann ends up being unable to say literally what a “two-inone” is. He must, upon his own account of the matter turn to non-literal ontological discourse, and resort to metaphors and similes that cannot be unpacked. This by itself is odd for a philosopher who, like Bergmann, locates himself in the more sober philosophical tradition that believes poetry is best left to the poets. But in any case it also seems to come into conflict with Bergmann’s philosophical method, his phenomenological method, which asserts that phenomenological description must precede all attempts at ontology. The idea of description certainly seems to be that of a literal description. SECOND. If Bergmann permits “two-in-one’s” – simples that ground internal relations of sameness and difference – at the point of particulars and particularity, then why does he not allow them at an earlier point – at, for example, the point where the nominalist introduces the “two-in-one’s” that he or she refers to by names like ‘this-red’? Peterson has raised this point in a strong defence of the nominalist position.109 If Bergmann’s reply is dialectical, then that would violate the phenomenological method. But if the reply is phenomenological, then the first difficulty arises immediately. THIRD. Bergmann must somehow recognize the distinguishability of the particular a1 and particularity. And, indeed, this he does: the “two” in the “two-in-one” are “things which are not nothings”.110 To do this he will need a sign in his language to mark this notion of distinguishability. But as soon as he does mark that notion then it will no longer be necessary to speak of two primitive internal relations of sameness and diversity. The relation of sameness between a1 and a2 will be definable, as it were, in terms of there being in the one an area distinguishable from the area in the other (with both distinguishable from the property of being an area). It would seem, then, that if Bergmann grants the phenomenological point that particularity is not nothing, that it is distinguishable from the particulars it is said to “pervade”, then there is no need for Bergmann’s problematic entities, no need for things that are “two-in-one” in a problematic sense, things that are both simple and also complex.

580 The alternative: What of the identification of difference with distinguishability that we have proposed? It at least does not run into the problem of describing what Bergmann calls a “two-in-one”: the “two-in-one” of a particular being “pervaded” by particularity is simply the complex represented by the sentence ‘This is an area’. Moreover, that sentence is as literal and as non-metaphorical a description as ‘This is red’. There is no problem with having to resort to nonliteral language. Two possible difficulties with this approach ought to be mentioned. One. If we are correct, the ultimate referent of ‘a1’ in he sentence ‘a1 is a particular’ – what Bergmann once called the determinant111 – and later came to call the item,112 – is just different, i.e. distinguishable, from other determinants of the same sort, i.e. from other particulars. But then, are not the determinants also the same as each other in that they are all determinants? The suggestion here is, clearly, that the problem of sameness arises once again, with a regress threatening. That is, if don’t make the particularparticularity complex into a Bergmann type of “complex”, i.e. a “two-inone”, then all the problems one hoped to solve by introducing particularity re-appear once more, leading to an infinite regress. Peterson has made this point.113 One can stop this regress, I think, simply by defining a determinant to be any entity which has either the property of particularity or the property of universality.114 The determinants which are particulars have in common the property of particularity. The detrerminants which are universals have in common the property of universality. And the two sorts of deteerminants have in common the characteristic of being either particulars or universals. A difficulty arises only if one insists that all questions about sameness must be answered in terms of a simple character. But there is no reason why that should be so. Two. Bergmann suggests that in its basic sense, the word ‘true’ applies to thoughts.115 He defines (t) ( + p , is true ) = [ ( + p , M p ) & p ] and (f) ( + p , is false ) = [ (+ p , M p ) & ~p If we allow both (4) a1 is a particular and (10) ~ ( a1 is a particular ) into our language, then we must say that (4) is at least potentially false; its falsity, or, what amounts to the same, (10), can at least be entertained, if

581 not believed. But (10) is excluded as false by the formation rules, and must, in that sense, be treated as ill-formed. However, the syntax of the sentences (t) and (f) is so specified that only well-formed sentences can be substituted for ‘p’. It follows that if we are to allow (4) and (10) into our language, as we are arguing we must, then we need new notions of truth and falsity defined in a further language in which (4) and (10) are both well-formed sentences. But the same sorts of general problem will arise with respect to this linguistic framework and the logical forms of its names. A regress looms, as Bergmann pointed out.116 However, it arises only if one insists that only well-formed sentences are to replace ‘p’ in ‘+ p , M p’. Bergmann does insist that only well-formed sentences replace ‘p’, but as we have argued, one need not do so, nor does Bergmann’s reason seem compelling. FINALLY, there is one further aspect of Bergmann’s idea that a particular and particularity form a “two-in-one” that I think is relevant. But to discuss it thoroughly would require that this study, which is already too long, to be considerably longer. I shall, therefore, just mention it. Bergmann refers to the “properties” of particulariity and universality as “subsistents:” Such entities “need no tie to tie them to the entities they pervade”. Similarly, exemplification is a subsistent that “needs no tie to tie it to what it ties.” These formulae are introduced as a way of forestalling Bradley-type regresses. Connected with this – as the inclusion of exemplification in the list makes clear – whether one must take facts as forming a basic ontological category, A full treatment of particularity must eventually treat of all these issues. Which illustrates once again, I suppose, that in ontology everything is connected to everything else.117

582 Endnotes to Study Fifteen

1.Shakespeare, 2 Henry VI, V, ii, l. 44ff 2.G. Bergmann, “Sketch of an Ontological Inventory,” Journal of the British Society for Phenomenology, 10 (1979), pp. 3-8; “Notes on Ontology,” Noûs, 15 (1981); pp. 131-154’ and “Notes on the Ontology of Minds,” in P. A. French et al., eds., Midwest Studies in Philosophy, 6 (1981), pp. 189-213. 3.A number of these further issues have been fruitfully explored by H. Hochberg in his illuminating study on The Positivist and the Ontologist: Bergmann, Carnap and Logical Realism (Amsterdam and Atlanta: Rodopi, 2001). 4.Bergmann, “Realistic Postscript,” in his Logic and Reality, (Madison, WI: University of Wisconsin Press, 1964), p. 339. 5.Jay Rosenberg, “The Given and How to Take It,” Metaphilosophy, 6 (1975), p. 316. Rosenberg’s comment reflects a similar position adopted by his teacher Wilfrid Sellars. 6.Bergmann, “Realistic Postscript,” p. 339, also p. 307; “Ontological Alternatives,” (in his Logic and Reality), p. 127; and “Strawson’s Ontology,” (in ibid.), p. 185. 7.Bergmann, “Ineffability, Ontology and Method,” in his Logic and Reality, p. 45. 8.Ibid. 9.Bergmann, “Realistic Postscript,” p. 309; and also “Acts,” (in his Logic and Reality), p. 5. In these passages, however, Bergmann carefully notes that not all acquaintance is direct acquaintance or sensing – something which Rosenberg (“The Given and How to Take It,” p. 306) seems to have missed, since he ascribes to Bergmann the view that all phenomenological description is description of what is given in acts of direct acquaintance. As a consequence of his failure to read such passages carefully (though he does cite the first!), Rosenberg’s whole treatment of Bergmann quite misses its target. 10.Rosenberg, “The Given and How to Take It,” pp. 305-306. 11.Bergmann, “Acts,” p. 14. 12.Rosenberg, “The Given and How to Take It,” p. 308. 13.E. B. Allaire, “Bare Particulars,” Philosophical Studies, 14 (1963), pp. 1-8, and also in E. B. Allaire et al. Essays in Ontology (The Hague: Nijhoff, 1963), pp. 14-21.

583

14.Cf. G. Bergmann, Realism (Madison, WI: University of Wisconsin Press, 1967), p. 68. 15.Bergmann, “Intentionality,” (in his Meaning and Existence [Madison, WI: University of Wisconsin Press, 1960]), p. 7. 16.Bergmann, “Realistic Postscript,” p. 308. 17.Rosenberg, “The Given and How to Take It,” p, 308. 18.See F. Wilson, “Universals and Tropes: The Role of a Principle of Acquaintance in Ontology,” this volume, above. 19.Note how this approach locates Bergmann squarely in the positivist tradition in which he first learned to philosophize. The problem of universals – how can words be general in their meaning? – is solved by such simple and unproblematic facts as that we can have two shirts that are the same colour. The point can be made by reference to the technique of an ideal language L that was employed by the positivists and which Bergmann still uses. (On this, see more below, notes 63 and 65.) Bergmann has an L with predicates like ‘Gr’ ( = ‘green’). This predicate names or refers to or denotes green. In other words, it is interpreted into that presented feature of things. (Cf. Bergmann’s “Logical Positivism, Language, and the Reconstruction of Metaphysics,” in his Metaphysics of Logical Positivism [New York: Longmans Green, 1953], pp. 60-61, p. 63.) It applies, (let us suppose) to this green shirt (which is called ‘a’) and to that one (which is called ‘b’); that is, ‘Gr(a)’ and ‘Gr(b)’ are both true – they both correctly describe the entities presented to us. What ‘Gr’ refers to is a presented feature of a and of b. This idea of interpretation into entities with which we are acquainted is taken as unproblematic. Indeed, it is taken a unproblematic not only by Bergmann but by other positivists such as Carnap. (Cf. C. G. Hempel’s discussion of the idea of an “empiricist language” in his “Problems and Changes in the Empiricist Criterion of Meaning,” Revue Internationale de Philosophie, 2 [1952-53], pp. 52-53.) What Bergmann does is take this unproblematic idea and use it to explicate the traditional claim that there are universals. That is, Bergmann argues, dialectically, that by adopting this ideal language he can solve/dissolve the traditional problem (and others, e.g. analyticity) where others cannot. (Cf. Bergmann, “Logical Positivism,” in his Metaphysics of Logical Positivism, p. 8; also his “Particularity and the New Nominalism,” in his Meaning and Existence, pp. 91-105) The language and its interpretation are unproblematic (cf. “Logical Positivism, Language, and the Reconstruction of Metaphysics”). What are problematic are the claims of the other fellow; it is he or she who has the problems, and they cease to be problems once they are explicated. In this sense, it is always the other chap who has problems, not Bergmann. Bergmann’s task, as he proposes it, is to take those problems that the philosophers of the past have created and bequeathed to their successors, and so explicate them that they admit of commonsensical solutions. Insofar as Bergmann proposes solutions to

584 traditional problems, they are not problematic as they were in the tradition as Bergmann has received it. Once again, this is a positivist stance, but note that it is also philosophical – and, indeed, metaphysical – recall the title of Bergmann’s first book, and one of his early papers (“A Positivistic Metaphysical of Consciousness,” Mind, n.s. 54 [1945], pp. 193-226), where its appearance led to his being excommunicated by the orthodox positivists! It is philosophical insofar as, in contrast to what Carnap & Co. do, takes the metaphysical tradition seriously, as presenting problems that are the core of the philosophical enterprise. In this sense, Bergmann is a positivist who is also a philosopher, rather than a positivist whose interest is the essentially non-philosophical area of Grundlagenforschung with respect to science (cf. “Physics and Ontology,” in Bergmann’s Logic and Reality, pp. 108-123). In a straightforward sense, then, the philosophical problem of universals is solved by pointing to the commonsense fact that for some shirts, two shirts do have the same colour. At least, it is this fact which one argues is crucial to the solution. But insofar as this is the only fact to which one appeals, the specifically philosophical disappears for Bergmann. In this sense, for Bergmann there are, as the positivists said, no philosophical problems. However, Bergmann is not Carnap. If Bergmann is a positivist, then he also is or came to be a philosopher. To explain: Bergmann differs from Carnap in not dismissing as a non-question the question of which alternative solution to a problem, e.g. that of universals, one ought to accept. For Carnap, what mattered was decision, not debate. (Cf. R. Carnap, “Empiricism, Semantics and Ontology,” Revue Internationale de Philosophie, 2 (1950), pp. 208-228; reprinted in his Meaning and Necessity, second edition [Chicago: U niversity of Chicago Press, 1956], pp. 205221.) As he would put it, the issue is practical, without cognitive content. Different solutions to traditional problems amount to different recommendations for an ideal language. The issue between them is to be resolved – dissolved rather than solved – by a simple choice of a convenient language. The issue is to be resolved by a decision, by stipulation and fiat, rather than by argument. Bergmann, in contrast, (cf. his “Comments on Professor Hempel’s ‘The Concept of Cognitive Significance’,” in his Metaphysics of Logical Positivism, p. 267), the questions are to be settled by rational argument, by dialectics, even if he is also convinced – as his adoption of a phenomenological method or Principle of Acquaintance shows him to be – that the solutions are to be found by appeal to what is unproblematic and uncontroversial. Hochberg has warned us against supposing that “...since one can buy two shirts of the same colour, some form of Platonism must be adopted and all forms of nominalism must be rejected, on the basis of that fact alone” (H. Hochberg, “Russell’s Reduction of Mathematics to Logic,” in E. D. Klemke, ed., Essays on Bertrand Russell [Urbana, IL: University of Illinois Press, 1970], p. 403). This warning may well be just: Bergmann may well be wrong. Only practice of the method will reveal that; as Bergmann always said, the proof of any pudding is in the eating. Yet the warning is misleading at least in its implied description of the phenomenological method. To adopt this method, and appeal to two shirts having the same colour to solve the problem of universals, is not to appeal to that fact alone, even if that fact is the only fact to which

585 one appeals. For, one must argue for the relevance of that fact to the traditional problems. That is, one must indicate that fact and also argue dialectically. Hochberg completely misses the latter. Insofar as Bergmann is, and Hochberg is not, prepared to appeal to such unproblematic facts in the solution of philosophical problems, the former remains closer than the latter to the positivism of the early Vienna Circle. Another aspect of this closeness to the early logical positivists was Bergmann’s continuing concern with the problem of analyticity. 20.Cf. Bergmann, “Ontological Alternatives,” p. 132f. 21.Cf. Bergmann, “Strawson’s Ontology,” p. 181. See also E. B. Allaire, “Existence, Independence, and Universals,” Philosophical Review, 69 (1960), pp. 485-496. 22.W. V. O. Quine, “On What There Is,” in his From a Logical Point of View, 2nd edition (Cambridge, MA: Harvard University Press, 1963), p. 10. 23.Cf. R. Grossmann, “Conceptualism,” in E. B. Allaire et al., Essays in Ontology (The Hague: Nijhoff, 1963)., pp. 30-39; Also Allaire, “Existence, Independence and Universals”; and H. Hochberg, “Nominalism, General Terms, and Predication,” in his Logic, Ontology, and Language (Munich: Philosophia Verlag, 1984), pp. 133-149. 24.Cf. H. Hochberg, “Nominalism, Platonism and being True of,” in his Logic, Ontology and Language, pp. 150-156; and “Mapping, Meaning and Metaphysics,” also in his Logic, Ontology, and Language, 157-184/ 25.Cf. W. Sellars, “Naming and Saying,” in his Science, Perception and Reality (London: Routledge and Kegan Paul, 1963), pp. 245-246. Sellars’ argument in this essay uses as its foil Bergmann’s ontology, and, indeed, is directed as negatively against this ontology as much as it is directed positively at a defence of tropes. Sellars considered Bergmann’s realist ontology of universals and particulars to be his main opposition. But if we are correct, then there are universals, and so Sellars’ trope ontology is wrong, and Bergmann’s realism is correct, at least in this regard. 26.Cf. Bergmann, “Strawson’s Ontology,” pp. 185-86; cf. R. Ackermann, “Perspicuous Languages,” in M. Gram and E. Klemke, ed., The Ontological Turn (Iowa City: University of Iowa Press, 1974). 27.Bergmann, “Synthetic A Priori,” in his Logic and Reality, p.288ff. 28.Ibid. 29. Bergmann, “Ineffability, Ontology and Method,” p. 47. 30.Ibid.

586

31.Rosenberg, “The Given and How to Take It,” p. 308. Emphasis added. 32.Ibid. 33.Ibid., pp. 307-308. 34.Ibid., p. 309. 35.Bergmann, “Sketch of an Ontological Inventory.” 36.H. Hochberg and G. Bergmann, “Concepts,” in Bergmann’s Meaning and Existence, pp. 106-114. 37.Bergmann, “Sketch of an Ontological Inventory.” 38.Cf. Sellars, “Naming and Saying.” 39.Bergmann, “Sketch of an Ontological Inventory.” 40.Ibid. 41.Cf. F. Wilson, “The World and Reality in the Tractatus”, in this volume, above. 42.Bergmann, “Sketch of an Ontological Inventory.” 43.G. Bergmann, “Intentionality,” in his Meaning and Existence, pp. 3-38. 44.Cf. G. Frege, “On Sense and Reference,’ (in P. Geach and M. Black, trans. and eds., The Philosophical Writings of Gottlob Frege [Oxford: Oxford University Press, 1960]), pp.59-61. 45.Compare the interesting discussion in R. Van Iten, “Berkeley’s Alleged Solipsism,” Revue Internationale de Philosophie, 16 (1962). 46.Frege, “On Sense and Reference,” p. 60. 47.See F. Wilson, “Acquaintance, Ontology and Knowledge,” in this volume, above. 48.Cf. G. Bergmann, Realism (Madison, WI: University of Wisconsin Press, 1967), p. 133; G. E. Moore, Some Main Problems of Philosophy (London: Methuen, 1953), p. 263. 49.They are often said to be “Platonic” or are described as inhabiting “Plato’s heaven.” But that is misleading. Plato’s reasons for introducing a Third World of Forms are rather different from Frege’s reasons for introducing a Third World of Senses. For Plato the reasons have to do with the problem of sameness (a problem that for Plato is created by his acceptance of nominalism), while for Frege the reasons have to do with

587 the publicity of thought (which, interestingly enough, is a problem created for Frege by his nominalistic assumption that ideas in the mind are in all aspects particular). 50.Cf. W. Sellars, “Notes on Intentionality,” p. 307. 51.Bergmann, “Intentionality,” p. 10ff. 52.G. E. Moore, Some Main Problems of Philosophy, p. 276ff, pp. 308-309, p. 259. 53.Bergmann, “Intentionality,” p. 7; cf. Moore’s remark that the sentence that represents the fact that is the content of a belief is also a name of the belief (Some Main Problems of Philosophy, p. 256). 54.Sellars, “Notes on Intentionality,” p. 310. See also F. Wilson, “Marras on Sellars on Thought and Language,” in the present volume, above. 55.See Sellars, “Inference and Meaning,” Mind, n.s. 62 (1953), pp. 313-338. 56.See F. Wilson, “Implicit Definition Once Again,” this volume, below. See also F. Wilson, “Barker on Geometry as A Priori,” Philosophical Studies, 20 (1969); pp. 4953; and R. Grossmann, Ontological Reduction (Blooomington, IN: Indiana University Press, 1973). 57.Bergmann, “Intentionality,” p. 15ff. 58.Sellars, “Note on Intentionality,” p. 310. 59.Bergmann, “Intentionality,” p. 13. 60.Cf. G. Humphrey, Thinking (New York: Wiley,1963). 61.Bergmann, “Intentionality,” pp. 12-17. 62.For this pattern (which is clearly Kuhnian), see F. Wilson, The Logic and Methodology of Science and Pseudoscience (Toronto: Canadian Scholars Press, 2001). See also F. Wilson, Empiricism and Darwin’s Science (Dordrecht, the Netherlands: Kluwer, 1988). 63.Cf. Bergmann, “Acts”; see also Moore, Some Main Problems of Philosophy, pp. 258-259. 64.Cf. Bergmann, “Logical Atomism, Elementarism, and the Analysis of Value” (in his Metaphysics of Logical Positivism, pp. 243-254). 65.Cf. Bergmann, “Intentionality”. Also his “Analyticity” (in his Meaning and Existence, pp. 73-90). 66.Cf. F. Wilson, “Logical Necessity in Carnap’s Later Philosophy,” in A. Hausman

588 and F. Wilson, Carnap and Goodman: Two Formalists (The Hague: Nijhoff, 1967), pp. 97-225. 67.Bergmann, “Sketch of an Ontological Inventory”, and “Notes on Ontology.” 68.Earlier Bergmann’s L contained defined terms. These represented what he called derived characters. The latter were needed, he believed, to solve certain problems concerning the perception of physical objects; see his “Realistic Postscript,” p. 325, and also “Diversity”, Proceedings and Addresses of the American Philosophical Association, 41 (1968), p. 29. He eventually (“Notes on the Ontology of Minds”) solved these problems without introducing the – clearly problematic – derived characters. He does this by noting that two definitionally equivalent sentences can express the same thought, meanr its intention, and the phenomenological point that consciousness of the intention of (most) thoughts is – I speak succinctly – mediated by the awareness of one of the sentences that meansr the intention. The solution is ingenious, as phenomenologically adequate as these things can be, and, it seems to me, wholly plausible. 69.Cf. Bergmann, “Notes on the Ontology of Minds.” See also Moore, Some Main Problems of Philosophy, p. 256. 70.Bergmann, “Intentionality”, p. 18; “Notes on the Ontology of Minds.” 71.See Note 68, above. 72.D. M. Armstrong, Universals and Scientific Realism, 2 vols. (London: Cambridge University Press, 1978), Chapter 17. 73.If one is not rational, then there will be other rules, e.g. “It is obligatory that one be a believer in what our Holy Book asserts.” But we need not deal with these, at least not in this context. But see F. Wilson, “Dummett’s History: Critical Review of Michael Dummett’s Origins of Analytical Philosophy,” in the present volume, below. 74.The term ‘non-empty’ is redundant; the notion of an empty universe makes no philosophical sense (though the notion of an empty set is legitimate). Cf. H. Hochberg, “A Note on the ‘Empty Universe’,” Mind, n.s. 66 (1957), pp. 544-546. 75.See Wilson, “The Notion of Logical Necessity in Carnap’s Later Philosophy” for a discussion of how the usual set-theoretical techniques can provide a syntactical explication of the notion of “true in all possible worlds.” 76.Cf. Bergmann, “Analyticity.” 77.See R. Carnap, Meaning and Necessity, second edition, p. 174. 78.Ibid., p. 186.

589

79.The point is made by Bergmann in his essay on “The Philosophical Significance of Modal Logic,” Mind, n.s. 69 (1960), pp. 469-485. 80.Nor does J. Hintikka really grasp this point; see his “The Modes of Modality,” Acta Philosophica Fennica, 16 (1963). As Hintikka points out, an adequate set-theoretical semantics for modal logic has been devised, but the issue is not just this, but rather whether this semantics can contribute to the explication of the philosophically problematic notion of “necessary” in which the necessarily true is true in all possible worlds. 81.Bergmann, “Notes on Ontology,” fn. 6. 82.Laws like this are, of course, rough – they are regularities but only gappy regularities; conformity to these regularities is maintained, so far as it is maintained, by rules, that is, by rulish thoughts – these operate causally in a complicated way – but the details don’t really matter: the basic idea is clear enough. For some elaboration of these ideas, see W. Sellars, “Reflections on Language Games,” in his Science, Perception and Reality, pp. 321-358. It is perhaps worth adding, however, that conformity to these patterns and the corresponding rules is a matter of rationality, or, in other words, of being a reasonable person. 83.Bergmann, Realism, p. 310. One might note that for Descartes also, a priori truths similarly command or demand our assent; they so demand it that we cannot resist. Similar things can be said about Humean “impressions.” For discussion, see F. Wilson, The External World and Our Knowledge of it: Hume’s Critical Realism: An Exposition and a Defence (Toronto: University of Toronto Press, forthcoming). 84.Cf. G. Bergmann, “Stenius on the Tractatus” (in his Logic and Reality), pp. 245-46; and “Notes on Ontology.” 85.Bergmann, “Stenius on the Tractatus,” p. 248. 86.Bergmann, “Ineffability, Ontology and Method.” 87.Cf. Sellars, “Naming and Saying.” 88 This was never recognized by the author of the Tractatus, who insisted that statements about particularity tried to make explicit was could only be shown and therefore had to be banished to the mystical . 89.Bergmann, “Sketch of an Ontological Inventory.”

590

90.Ibid. Cf. P. Butchvarov, “The Limits of Ontological Analysis,” (in Gram and Klemke, The Ontological Turn), p. 14, p. 17ff. 91.Bergmann, “Intentionality,” p. 32. 92.Cf. Bergmann, “Notes on Ontology.” 93.Cf. Rosenberg, “The Given and How to Take It,” p. 310. 94.Cf. Sellars, “Notes on Intentionality,” pp. 314-15. 95.Bergmann, “Notes on the Ontology of Minds.” 96.Ibid. 97.Strictly speaking, it is a definite description of the sentence. 98.Bergmann, “Intentionality,” p. 25. See also G. E. Moore, Some Main Problems of Philosophy, p. 256. 99.Bergmann, “Diversity,”, p. 28. This is not a multiplication of entities or wooley metaphysics that is somehow incompatible with positivism (cf. fn. 19, above). Rather, it simply makes explicit something that was always implicit in the positivist tradition. When the positivists constructed their ideal empiricist language L, they interpreted its zero-level constants (names) into individual things and its predicates into observable features of things. E.g. ‘gr’ was interpreted into green, ‘R’ into red (that is, the properties green and red), and so on. The point it that different predicates were assigned to different observable features, that is, observably different features. If the positivist programme of constructing an ideal empiricist’s language is to make sense, then the positivist must be presented with facts to the effect that this property is different from that property. Bergmann is simply drawing our attention to these facts. For the positivists, there was no need to emphasize these facts. Their importance derives from their relevance to the solution of various philosophical problems. But the latter were dismissed by such positivists as Carnap, who were always overly anxious to get away from philosophy to the tasks of trying to help the scientist improve his or her research practice (cf. Wilson, “Logical Necessity in Carnap’s Later Philosophy”, ch. 5) and of mere Grundlagenforschung. Because Bergmann returned to the roots of philosophy, even of positivist philosophy, he must stress the relevance of such simple facts, as that of difference being presented, which the positivists accepted but passed over in silence. 100.Bergmann, “Sameness, Meaning and Identity” (in his Meaning and Existence, pp. 132-138), and “Notes on Ontology.” 101.I.e. in currently fashionable jargon, they are rigid designators. There is no good

591 reason (other than a Freudian one) for adopting this jargon. 102.Bergmann, “Sketch of an Ontological Inventory” and “Notes on Ontology.” 103.Bergmann, “Sketch of an Ontological Inventory.” 104.Ibid. 105.Bergmann, “Sketch of an Ontological Inventory.” Cf. Butchvarov, “The Limits of Ontological Analysis,” p. 11. 106 See F. Wilson, “Universals, Particulars, and Tropes: The Role of a Principle of Acquaintance in Ontology,” in this volume, above. 107.Bergmann, “Diversity,” p. 26. 108.Bergmann, “Sketch of an Ontological Inventory.” 109.J. Peterson, Realism and Logical Atomism (Tuscaloosa: University of Alabama Press, 1966), Chapters IV and V. 110.Bergmann, “Sketch of an Ontological Inventory.” 111.Bergmann, “Diversity,” p. 26. 112.Bergmann, “Sketch of an Ontological Inventory.” 113.Peterson, Realism and Logical Atomism, pp. 58-59. 114.Certain qualifications must be made, I think, about universality, but they are not relevant for present purposes. See F. Wilson, “Grossmann on the Categorial Structure of the World,” this volume, above. 115.Bergmann, “Ontological Alternatives,” p. 137; “Intentionality”; and also his essay on “Semantics” (in his Metaphysics of Logical Positivism, pp. 17-29). 116.Bergmann, “Diversity,” pp. 26-27. 117.Since the original publication of this essay, I have had the benefit of reading H. Hochberg’s impressive study of Bergmann’s thought, The Positivist and the Ontologist: Bergmann, Carnap and Logical Realism (Amsterdam and Atlanta: Rodopi, 2001). There is much to be commented upon, too much, unfortunately, to comment on here.

Sixteen The Aboutness of Thought* It was Brentano who introduced the term ‘intentionality’ into recent philosophy to mark the characteristic aboutness that distinguishes our mental acts from other entities in the world. This use was not wholly original with Brentano, however. He had derived the term from mediaeval epistemology and philosophy of mind, which took over what are essentially Aristotelian ideas. According to the standard Aristotelian account of knowing, the essence or form of the substance known was literally in the mind of the knower: like knows like, and the mind knows its object through becoming identical with that object.1 The form or essence of the thing known is in the mind as properties are in the mind; in Cartesian terms: this is the form as idea, this is its material reality. But it is also the essence of the thing known as that essence in the substance known; in Cartesian terms, this is the objective reality of the idea. By virtue of the idea in the mind having this objective reality, the idea is about or represents or, in Brentano’s terminology, “intends” the object which is before the mind.2 There were problems, however. First, how could the idea come to be in the mind? Aristotle gave the answer that the mind went through a process of abstraction the product of which was the presence of the form or essence as a property of the mind. But Descartes argued that the form of wax involved an infinite number of variations in shape as the wax melted and flowed from one shape to another, whereas what is presented in sense experience is only finite. We could not, therefore, obtain the idea of the form of wax by abstraction from sensible impressions. He concluded that our ideas of forms or essences were all innate. But this was then subject to Locke’s sceptical critique of innate ideas. But second, how could a form or essence be in the mind, as a property of the mind, when it is precisely that feature of the substance of which it is the form that accounts for the properties of that substance: if the * Originally appeared as “The Aboutness of Thought,” in T. M. Lennon, ed., Cartesian Views: Papers Presented to Richard A. Watson (Leiden: Brill, 2003), pp. 151-164.

594 form of a oak tree causes that substance to grow like an oak then why does the same form in the mind not cause the mind to become an oak tree? Ideas as forms in the mind did solve one problem. When we perceive an object, we are conscious in experience of a sense impression. But that is a sense impression of only part of the object perceived: not all of the object as a thing located in space and enduring through time is given in sense experience, there is the future, the past, the other side, the inside, and what it would be like if certain other events were to occur. Present to consciousness is the sense impression but also the judgment that locates what is sensed, the impression, as of an object in space which causes the presence of the consciousness of a sensible object. The perceptual judgment consists of an essence or Form as a property of the mind; it as it were connects the mind not only to the sense impression but also to the larger object of which the sensible object is a part and which is the cause of our awareness of that sensible impression. It intends the whole of the perceptual object, not just the part that is given in sensory awareness. As H. H. Price once put it, there is on the one hand “primary recognition” of the sensible object as qualified in certain ways and as related to other sensible objects in certain ways, and, on the other hand, “secondary recognition” which is a recognition of an object that has qualities not given in sense, not present to consciousness.3 That there is something present in consciousness that as it were connects the mind to the perceptual object is clear. The tree that I perceive is itself not wholly present to my current conscious state; as a tree it has a past and a future which by the nature of the case are not part of my conscious states that is here present now. Indeed, if I am misperceiving, then there is no tree there – think of Macbeth’s dagger. But, even though the perceptual object is not wholly there or even if it is not there at all, I am nonetheless wholly certain that what is perceived is a tree: I am certain that what is perceived is a tree and not, for example, a fence post. What enables me to say that “That is a tree” and to say with certainty that “I perceive that that is a tree” is the presence in consciousness of a property, a property which intends the tree as an object that causes my mental state to be one which is properly expressed by saying “That is a tree” and properly described with certainty by saying “I am perceiving that that is a tree.”4 Moreover, I tend to respond behaviourally to a tree and not to a shrub or a fence post, and even if the tree is not there, even if I am misperceiving, I am still disposed to respond as if there were a tree present: it is a dagger that concerns Macbeth. There is a cause present in consciousness that is

595 connected to that disposition to behave; this present cause is the perceiving which has the property of being a perceiving of a tree.5 Finally, as Price emphasizes, the capacity for secondary recognition is acquired; it is something learned. And like induction, it involves a judgement that goes beyond what is sensibly present. But unlike induction, it does not involve, so far as consciousness is concerned, an inference. In that sense, it is an immediate judgment of no internal complexity. To be sure, the object of the judgement is complex; but the judgement itself is an immediate unity and whole. It is this judgment as a given unity or whole that disposes me to behave in ways appropriate to the object perceived, the object of the judgement. This judgement, the cause present in consciousness that disposes one to behave in certain ways rather than others, disposes one in particular to express one’s judgment by uttering or asserting a sentence describing the complex state of affairs perceived. The Aristotelians captured these points. On the one hand, there is a property present in the mind that, on the other hand, intends the structure of the perceptual object. Moreover, the Form or essence in the mind, like the Form or essence in the thing perceived, is a simple: to be sure, the characteristics of the substance that we observe in sense experience are complex, but the Form or essence which accounts for their presence is simple and unanalyzable.6 Nonetheless, there remain the problems with the notion of a form or essence and of such a form or essence coming to exist as a property of the mind. These problems exclude the Aristotelian solution, but there remains the issue of how it comes to be that the property present in the mind that represents the structure of the perceptual object actually does represent that structure of parts that are not presented. How does one analyze intentionality? There has been some argument about this. The continental philosophers so called have said much about intentionality, but in fact have done little to address the philosophical issues surrounding this connection. One finds greater sensitivity to these issues among analytic philosophers so called, or at least among some of them. Russell, for one, attempted to deal with the issues. He proposed to analyze intentionality in terms of a descriptive relation, in a way similar to the way in which he had analyzed “x kicks y”.7 This seems to make great deal of sense, initially: after all, grammatically the phrases ‘intends’ or ‘is about’ function as relational predicates, connecting the mental act, say of perceiving, to the state of affairs that it intends or is about – thus, “I perceive that this is a tree” grammatically relates the perceiving on the one

596 hand, the conscious state, with, on the other hand, the state of affairs, represented by ‘This is a tree’, which that perceiving intends or is about. There is a problem. If ‘x kicks y’ is true then both x and y must exist: one cannot be a kicker unless there is a kickee. That means that if intentionality is to be analyzed as a relation, then both relata, the perceiving, in our example, and the state of affairs intended, must both exist. Russell argued that this is no problem for sensible acquaintance, since that is infallible. Or at least, Russell argues that sensible acquaintance, though not perceiving, is quite certain, and clearly supposes that such certainty can be accounted for provided that one makes acquaintance a relation like kicking that guarantees the existence of its object. But there is a problem. Though sensible acquaintance might well be certain and a candidate for infallible judgment, the same is not true of either perceiving or believing. (The latter is the example that Russell discusses, but he makes it clear that what he says applies to the former also.) These sorts of mental act are not infallible; to the contrary, they are quite often false. But if false, then their objects do not exist. In that case the simple relational account that Russell gave to sensible acquaintance cannot apply to believing or perceiving. Russell argues his case with respect to the example of Othello believing – falsely – that Desdemona loves Cassio. Since it is false that Desdemona loves Cassio one cannot make believing a relation between Othello and this state of affairs. Russell suggests that though that state of affairs does not exist, its constituents, namely, Desdemona, Cassio and loves, all exist. Believing then is, he proposes, a multi-term relation which relates Othello to Desdemona, Cassio and loves, and that this belief is true (false) just in case that the state of affairs the Desdemona loves Cassio exists (does not exist). This avoids the problem of false beliefs and perceivings only to raise others. In the first place, it does not distinguish between Othello believing that Desdemona loves Cassio and Othello believing that Cassio loves Desdemona. Secondly, as Wittgenstein argued,8 it does not make it impossible to have nonsensical thoughts: from a knowledge of the fact that Othello is related to Desdemona, Cassio and loves, how can one from that information alone guarantee that he is not thinking the nonsense thought that Cassio Desdemonas loves? Thirdly, Russell’s analysis treats two different forms of the same thing – sensible acquaintance, on the one hand, and, on the other, believing and perceiving are all forms of consciousness – in two different ways: the former is a simple two-term relation, the latter is

597 multi-term and in fact could become very complex indeed if there are more than a few entities in the state of affairs believed or perceived. Fourthly, this analysis makes the belief true or false depending upon whether the state of affairs that Desdemona loves Cassio exists or does not exist. Implicitly, therefore, Russell is still smuggling in that state of affairs as the object of belief, that which the belief is really about, re-introducing, albeit covertly, all the problems of how the existing state of believing can be related to a state of affairs that does not exist. Wittgenstein therefore attempted another analysis. He proposed “... that ‘A believes p’, ‘A has the thought p’, and ‘A says p’ are of the form ‘”p” says p; and this does not involve a correlation of a fact with an object, but rather the correlation of facts by means of the correlation of objects.” (Tractatus, ¶ 5.542) This makes thought a special case of language: the meaning of thoughts is to be analyzed on the model of the meaning of language. As Wittgenstein would have it, a thought is an image or set of images that functions as a sentence functions, playing the role in inner language that the corresponding spoken or written sentence plays in overt speech. Any sentence becomes a meaningful proposition, the Tractatus argues, when it pictures a state of affairs. The state of affairs to which it is thus coordinated is its meaning or sense (Sinn). The sentence becomes meaningful by virtue of sharing a logical form with the state of affairs pictured. This shared logical form establishes an isomorphism between the picturing sentence of thought and the pictured state of affairs. But the isomorphism by itself does not establish that the sentence or thought is about the state of affairs pictured; for, given that a relation of isomorphism is symmetric, that would imply that the thought or sentence is itself pictured by the state of affairs, that the state of affairs as much means the sentence as the sentence means the state of affairs. What makes the sentence of thought about the state of affairs is the fact that the names that, in concatenation, make up the sentence designate the objects in the state of affairs rather than conversely. By virtues of these relations an object becomes the meaning in the sense of referent (Bedeutung) of a word. In Wittgenstein’s metaphor, these relations of designation or reference are as it were “antennae” that reach out from the words in the thought or sentence to the objects of the state of affairs pictured: “The pictorial relationship consists of the correlation of the picture’s elements with things. / These correlations are, as it were, the feelers of the picture’s elements, with which the picture touches reality.” (Tractatus , ¶ 2.1514, ¶ 2.1515) This does avoid the problems that are raised by Russell’s account of

598 belief. But it has its own difficulties. The meaning of a term in its most general sense is given by the role that it plays in language. This role consists of regularities or patterns that describe the patterns of use or dispositions to use of the word. These regularities include the syntactical regularities (word-word connections) and semantical regularities (worldword connections). It also includes pragmatic aspects of language (wordworld connections), but Wittgenstein ignores these in the Tractatus; as the “Slab!” example at the beginning of the Philosophical Investigations9 makes clear, he was later to discover these aspects of language and of linguistic meaning. Now, in any use of a word, many of the dispositions that form part of its meaning are not exercised. We can know that a disposition is present only if we infer it or we experience its exercise. In the case of thoughts we do not infer – our knowledge of what we are thinking about is in this sense immediate –, and many of the dispositions determining the meaning of the words and sentence are not exercised at that moment. But we do know what we mean: our sense of knowing what our thought is about is immediate and patent. Since meaning is as it were wholly present, it follows that it cannot be simply a sentence or words as images that carry the meaning; there must be some non-image present in consciousness by virtue of which we know what our thought is about. Wittgenstein saw this point clearly in the Philosophical Investigations: Experiencing a meaning and experiencing a mental image. “In both cases,” we should like to say, “we are experiencing something, only something different. A different content is proffered – is present – in consciousness.” – What is the content of the experience of imagining? The answer is a picture or a description. And what is the content of the experience of meaning? I don’t know what I am supposed to say to this. ...10

What that meaning is is that it is not an image: it is an imageless thought, where the content is expressed by a sentence but not described by sentence. Wittgenstein is here discovering, as it were, what had earlier been discovered by the Würzburg psychologists of the school of O. Külpe.11 It was G. Bergmann who took up both the point of view of the Tractatus and that of the Würzburg psychologists. He argued that this problem confronting the analysis of the Tractatus could be solved if one took that which carried the meaning to be an imageless simple character, exemplified by mental acts, including mental acts of thinking, believing, perceiving, approving, wishing, and so on. This simple character he referred to as a proposition or propositional character. The relation between

599 the thought or belief and that state of affairs which is its object he took to be a logical relation, by which he then meant a linguistic or syntactical relation. He takes the sentence that describes the state of affairs which the thought is about and refers to it as the “text” of the thought. He then forms what he refers to as a predicate by putting the text in quotes; this predicate designates, he has it, a simple character. This character is a character or property of thoughts, mental acts, including beliefs and perceivings – Bergmann refers to these as the different species of mental act; but all species have in common that the acts they characterize are also characterized by a propositional character, where the latter is that character by virtue of which the thought intends the state of affairs described by its text. It is that character present in consciousness of one who knows the meaning of the sentence that is the text, and absent from the mind of one who hears it uttered but does not know its meaning. By virtue of this character – this propositional character – being present in consciousness, we know what state of affairs it is that the mental act is about or intends. Bergmann represents this connection by the sentence + p ,M p where “M” means “means”. This is supposed to be true by virtue of its syntax – it is true if and only if the ‘p’ after the ‘M’ is the same as the ‘p’ in parentheses –, and it is therefore supposed to be analytically true.12 But on this view the connection between the propositional character and the state of affairs that it means or presents is merely linguistic. This would make the connection one that holds by virtue of the conventions that govern the use of language. But the connection is that which holds between the propositional character and the state of affairs which the mental act, by virtue of that character, intends. Conventions, however, are contingent; they could be other than they are. On this view, just as ‘bachelor’ could mean ‘unmarried budgerigar’, so the propositional character designated by ‘+ this is red ,’ could be used to mean what is now meant by ‘+ this is square ,’. That is, it could mean (“M”) what is now represented by the sentence ‘this is square’. Of course, though it would be the same simple character, it would in such circumstances be designated by ‘+ this is square ,’ rather than the way it is now designated, but that is no problem. The problem is that it would seem that we could change the intentional object of our thoughts simply by changing the conventions of our language. And that does not seem plausible: one can easily see that one can change the conventions for the use of ‘bachelor’ or for ‘red’ but it is quite another thing to suppose that simply by deciding to use words differently

600 we can change the objects that our thoughts present to us. Bergmann later changed his account of “M”. On his later view, it is not merely linguistic, but represents an intrinsic feature of propositional characters, and as such has ontological status.13 Searle has later taken up Bergmann’s point and offered a similar account of mental acts. He offers the same linguistic representation of mental acts. As he puts it, “A belief is intrinsically a representation in this sense: it simply consists in an Intentional content and a psychological mode.”14 Here the mode of the psychological state – whether it is a believing, or a willing or whatever – is what Bergmann calls the species. He considers John’s belief that King Arthur slew Sir Lancelot, and argues that the statement that John believes that King Arthur slew Sir Lancelot is one whose “...truth requires only that John has a belief and that the words following ‘believes’ [what Bergmann calls the text] accurately express the representative content of his belief [what Bergmann calls the propositional character]....in reporting his belief I present its content without committing myself to its truth conditions.”15 As for Bergmann on his later view, so for Searle, this intentionality of mental states is an intrinsic feature: “Beliefs, fears, hopes, and desires ... are intrinsically Intentional.”16 Searle differs from Bergmann only in making the character denoted by +p , or by the sentence in the that-clause, a characteristic of brain states rather than the specifically mental, and non-physical, particulars that Bergmann regards as individuating mental acts. Mental states, Searle says, are “realized in the structure of the brain.”17 There is a problem, however, one which Bergmann recognizes but which Searle simply ignores. It is this. Once one gives “M” ontological status as an intrinsic feature of propositional characters, then one must ipso facto give ontological status to the propositional character, which is no problem, and to the state of affairs meant, which is a problem since when the belief is not true that state of affairs does not exist. Bergmann accepts the point and is prepared to grant ontological status to states of affairs that are merely possible but not actual.18 This solves the problem, but only at the cost of introducing as an ontological category, as existents, things that do not exist. But as one of the central principles of Spock’s metaphysics maintains, “Nothing unreal exists”; or as Russell insisted, an ontologist must have a robust sense of reality, at least a sense of reality sufficiently robust that the unreal is not made somehow real. Bergmann, in order to solve one problem, gives up this robust sense of reality; as for Searle, we

601 do not know how robust is his sense of reality because he ignores the issue. There is another problem which neither Bergmann nor Searle addresses. That is the fact that in granting status to “M” one is creating an absolutely unique ontological category. That is in effect to give up any attempt to explain the intentionality of thought. Richard Watson has put the point nicely, when he states that What is required to support a way of ideas is an ontological model that shows how ideas provide knowledge of things. To be explanatory, the model cannot consist merely of a set of entities whose ad hoc nature it is to do what is required, entities that are related by relations that are named but not analyzed.19

What Watson goes on to show is that most recent attempts to provide an analysis of “M”, one that in his sense is explanatory, do so in terms of a relation of isomorphism. He traces this notion from D to C – Descartes to Churchland. The difficulty is that this proposal runs into the difficulty that Wittgenstein’s proposal that all thought is to be analyzed in the way that one in general analyzes “‘p’ says p”. We thus seem to be caught in a dilemma. Either we provide an analysis of “M” in terms of isomorphism, in which case the result is inadequate to the phenomenology of mind, or we take “M” to be a primitive term, in which case we are stuck with something that remains an ontological mystery together with unacceptable unreal but existent states of affairs. I do think that Watson is on the correct track, however. There is a point to making isomorphism central to the analysis of “M”, intentionality, that is, using the conventions of language including what Wittgenstein referred to as “feelers” or “antennae”, designation or referring relations, to explain aboutness. After all, it does work reasonably well for the aboutness of language. The problem is that the thoughts language as “text” express cannot, on phenomenological grounds, be construed as images or tokens of words: those thoughts must be construed as Bergmann construes them, as simple unified wholes. Now, what made us conclude that the characters that carry the meaning in consciousness are simple unified wholes was the fact that when they are in consciousness we know, immediately and without inference, the texts the utterance of which would express those thoughts. But what is it to “know” in this context? It means to be present in consciousness. Also it does not mean to “judge” that the propositional character has the text it has. Nor is the description of this character something for which we search and then discover: it is just there, so to speak, as is the “knowledge” that

602 we have of its nature. All that it seems to mean, besides the presence of the character in consciousness, is that we are disposed, directly and without inference, to express the thought by uttering its text as a meaningful expression representing a state of affairs. This is a causal relation: the propositional character functions as the ground in consciousness of the disposition or rather set of dispositions that constitute the meaning of the text. And to say that Beverley knows directly when he or she believes or perceives that p that that thought is the thought that p is simply to say that that is the thought the presence of which disposes him to utter the text ‘p’, to make the assertion that p.20 And the thought that p is about p precisely because it is the thought that disposes one, directly and without inference, to make the assertion that p.21 If this is correct, then there is nothing mysterious about intentionality, about the aboutness of thought. We do not need to make a special ontological category for it, transforming it into a special relation of a sort of which there are no other examples.22 There is no need to introduce as special objects of false thoughts states of affairs that do not exist but have some sort of wispy reality. We have explained the aboutness of thought in terms of the aboutness of language, as Watson suggested we must do if we are to provide an ontological analysis that explains how such aboutness works, but have done so while retaining the thought as a simple character grounding our knowledge of the state of affairs the thought it about. The objection to thus attempting to understanding the intentionality of thought in terms of the aboutness of language is that the intentionality of language presupposes the intentionality of thought. Thus, Searle, for one, speaks of the “pervasive and fundamental confusion” that “we can analyze the character of Intentionality solely by analyzing the logical peculiarities of reports of Intentional states.”23 As Searle sees it, words are not intrinsically intentional, unlike thoughts. Their meaning, the states of affairs that they represent, or, as Searle speaks, their conditions of satisfaction, is something imposed upon them. A string of sounds or marks as it occurs in an utterance act acquires conditions of satisfaction through an intentional act of the speaker: “the utterance act is performed with the intention that the utterance itself has conditions of satisfaction.”24 Thus, the correlations which connect words to objects concatenated in states of affairs are dependent upon the intentions of the speaker to endow those words with those referents, where these intentions are the intentions in the sense of volitions that cause the speaker to make the utterance and are

603 themselves intentional states, and where the latter is Brentano’s sense of ‘intentional’. But Wittgenstein argued effectively against this viewpoint. As he put it, considering the correlation of names to things named, “...we may indeed fancy naming to be some remarkable act of mind, as it were a baptism of an object. And we can also say the words ‘this’ to the object, as it were address the object as ‘this’ – a queer use of this word, which doubtless only occurs in doing philosophy.” (Philosophical Investigations, ¶ 38) In this way “Naming appears as a queer connexion of a word with an object.” (Ibid.; his emphasis) The name-named correlation is a world-word connection. Searle’s suggestion is that an act of naming creates and establishes the correlation. Wittgenstein’s point is that baptism is itself a rite or ritual, thoroughly convention-laden, and so is ordinary naming. It cannot therefore be the source of the conventions that define language. Wittgenstein’s argument here is parallel to Hume’s argument that promising or entering into a contract cannot be the source of the property and exchange conventions that define civil society because promising can create social conventions only by virtue of the previously established conventions which define that practice.25 Addis has argued in a way that is similar to Searle: ...there cannot be conventional representation unless there is also, at some level, unconventional or natural representation; for deciding or coming to agree that one entity – the word ‘red’, for example – shall represent another – the property of being-red – presupposes that one is already able to represent each, not of course by words (which form of representation would be circular) but, ultimately, in some non-linguistic way that is non-conventional.26

Now, it is true that one cannot agree with others to establish the correlation recorded in the norm (n) ‘red’ means red unless each of the group is able to attend to both the marks represented by “‘red’” and the colour red, and such attention is a case of mental acts being about or intending those things. The establishing by agreement of the rule (n) to establish a correlation of marks (or sound) and the colour red thus cannot be the sort of process that is used to establish all correlations of marks and objects. In other words, agreeing to establish the practice governed by the norm (n) itself presupposes conventions: here is simply another case of the practice of naming, and this practice, just like baptising, involves prior conventions, and therefore cannot be the basis for

604 establishing all conventions of naming. The original conventions must be established by some other means. For Addis, such conventions presuppose non-conventional aboutness, established by an ontologically grounded “M”. But Addis’ conclusion does not follow. All that follows is that the original conventions cannot be established by some conventional practise. Contrary to Addis, such a process does not need to involve nonconventional aboutness. All that one needs is that there be a regularity established between the mark or sound and the colour. Such a correlation could be established by simple learning or by reinforcement. Such learning presupposes that it is possible to discriminate both the mark or sound, on the one hand, and the colour, on the other. But the presence of the capacity to discriminate does not require that one have intentional conscious states. Addis and Searle are wrong, then, in their claim that conventional aboutness presupposes non-conventional aboutness.27 It would seem, then, that after all intentionality can be analyzed as Watson has proposed in terms of isomorphism, and specifically, as Wittgenstein suggested in the Tractatus, in terms of an isomorphism constituted by the conventionally established correlations by which language can picture or represent states of affairs in the world.

605 Endnotes to Study Sixteen

1.Cf. the epistemological principles J, K, and L laid out by R. A. Watson, The Breakdown of Cartesian Metaphysics (Atlantic Highlands, NJ: Humanities Press, 1987), pp. 51-52. 2.Cf. Watson’s principle M, p. 52. 3.H. H. Price, Thinking and Experience (Cambridge, Mass.: Harvard University Press, 1953), pp. 44-52. 4.Cf. L. Addis, Natural Signs: A Theory of Intentionality (Philadelphia: Temple University Press, 1989), p. 51ff. 5.Ibid., p. 48ff. 6.Cf. F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies (Toronto: University of Toronto Press, 1999), especially Study One. 7.B. Russell, Problems of Philosophy (Oxford: Oxford University Press, 1912), Ch. Seven. For other aspects of Russell’s thinking on these points, see F. Wilson, “Burgersdijck, Coleridge, Bradley, Russell, Bergmann, Hochberg: Six Philosophers on the Ontology of Relations,” in the present volume, above. 8.L. Wittgenstein, Tractatus Logico-Philosophicus, trans. D. Pears and B. F. McGuiness (London: Routledge, 1961): “The correct explanation of the form of the proposition, “A makes the judgement p’, must show that it is impossible for a judgement to be a piece of nonsense. (Russell’s theory does not satisfy this requirement.)” (Tractatus, ¶ 5.5422) 9.L. Wittgenstein, Philosophical Investigations, trans. G. E. M. Anscombe (Englewood Cliffs, NJ: Prentice Hall, 1958). 10.L. Wittgenstein, Philosophical Investigations, pp. 175-76. 11.Cf. G. Humphrey, Thinking (London: Methuen, 1951), Ch. 1, for the story of the Würzburg school and the entrance of imageless thoughts into introspective psychology. The work of the Würzburgers is located in the history of psychology in F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill (Toronto: University of Toronto Press, 1990), Ch. Eight. J. N. Findlay has correctly pointed out that one of Wittgenstein’s main discoveries that led from the account of mind in the Tractatus to the philosophy of the

606 Philosophical Investigations was the discovery of imageless thoughts; see his “Wittgenstein’s Philosophical Investigations,” Revue Internationale de Philosophie, 7 (1953), pp. 201-216. 12.Cf. G. Bergmann, “Intentionality,” in his Meaning and Existence (Madison, WI: University of Wisconsin Press, 1959). See also “Body, Mind and Acts,” and “Logical Positivism, Language and the Reconstruction of Metaphysics,” both in his The Metaphysics of Logical Positivism (New York: Longmans, 1954), the former at pp. 132-152, the latter at pp. 30-77. For discussion of Bergmann’s views and his philosophical development, see H. Hochberg, The Positivist and the Ontologist: Bergmann, Carnap and Logical Realism (Amsterdam and Atlanta: Rodopi, 2001). 13.G. Bergmann, “Acts.”, in his Logic and Reality (Madison, WI: University of Wisconsin Press, 1964), pp. 3-34. 14.J. Searle, Intentionality (Cambridge: Cambridge University Press, 1983), p. 22. For discussion of Searle, see H. Hochberg, The Positivist and the Ontologist, p. 32ff.. 15.Searle, Intentionality, p. 23. 16.Ibid., p. 27. 17.Ibid., p. 265; his emphasis. 18.Cf. G. Bergmann, “Notes on the Ontology of Minds,” in P. French et al., The Foundations of Analytic Philosophy: Midwest Studies in Philosophy, 6 (1981), (Minneapolis, MN: University of Minnesota Press, 1981), pp. 189-213. 19.R. A. Watson, Representational Ideas: From Plato to Patricia Churchland (Dordrecht, The Netherlands: Kluwer, 1995), p. 17. 20.This clearly presupposes some sort of psychophysical parallelism, and therefore the possibility of a purely behaviouristic reconstruction of human behaviour, including verbal behaviour. But mind remains mind, and is there as part of the world as we experience it. For a discussion of parallelism, see F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill (Toronto: University of Toronto Press, 1990), Chapter Eight. 21.For further discussion, see F. Wilson, “Effability, Ontology and Method,” in the present volume, above.

607

22.Mind of course is itself unique, but on the present analysis the category of thought belongs to the ontological kind of property; thoughts are unique in so far as they are all thoughts, just as colours are unique in so far as they are all colours. 23.J. Searle, Intentionality, p. 180. 24.Ibid., p. 167. 25.For Hume’s philosophy on this point, see F. Wilson, “Árdal’s Contribution to Philosophy,” in Páll Árdal, Passions, Promises and Punishment (University of Iceland Press: Reykjavik, 1998), pp. 7-40. 26.L. Addis, Natural Signs, p. 57. 27.For more on this point in a slightly different context, see F. Wilson, “Grossmann on the Categorial Structure of the World,” in the present volume, above.

Seventeen Language and (Other ? ) Abstract Objects* Linguistics as a science has often tried, sometimes more, sometimes less successfully, to separate itself from psychology. Already in 1916 Ferdinand de Saussure, the person who can perhaps be called the founder of modern linguistics, tended towards a rejection of methodological individualism and developed in his influential study, Cours de linguistique générale,1 the view that language has a kind of “sociological reality” that is there in the world somehow quite apart from its use by individual persons: language supposedly maintained a kinds of abstract existence all its own, independent of its speakers.2 More recently, Esa Itkonen3 has argued from a rather naïve Popperian account of science to the conclusion that linguistics is not a natural science, but rather a non-empirical science like logic or mathematics, based on a non-empirical source of knowledge, to wit, something called intuition. His argument fails to convince: an argument that linguistics is not a natural science based on Popperian views of how to demarcate science, even if they are less naïve than Itkonen’s, will not do.4 J. J. Katz has also developed an interesting case for linguistics being a nonempirical science.5 But where Itkonen wishes to link linguistics with the hermeneutical tradition, Katz, patterning his case on Frege and the early Husserl, develops what he reckons to be a Platonistic account of science. Bloomfield, in his Language,6 accepted that linguistics was part of (a behaviouristically construed) psychology, but proposed a division of scientific labour in which linguistics would concern itself with but a single part of the socio-psychological language process, namely, language as a product, or output of language processes. Bloomfield and his students carried out studies that were largely phonological and morphological, with little or no work on syntactic or semantical systems, and certainly not on anything that touched the pragmatics of language. The focus came to be on the construction of discovery procedures, that is, mechanical devices or methods that would work on a set of linguistic data to produce an explicit and accu* Originally appeared as a Critical Review of J. Katz, Language and Other Abstract Objects, Canadian Journal of Philosophy, 14 (1984), pp. 663-673.

610 rate grammar for that language. Zelig Harris continued that work, attempting to develop such procedures to formalize such traditional or imprecise categories as Noun, Verb, etc. Chomsky’s transformational grammars were natural outgrowths of this structural linguistics. The new twist contributed by Chomsky was the idea of a generative grammar as part of an attempt to achieve the goal of an economical characterization of the set of sentences of a language – in any reasonable case, an infinite set. This amounted to the determination to construe natural languages as special cases of the sort of formal language dealt with by the branch of mathematics known as automaton theory; instead of merely listing various recurring syntactic features, one attempted to define them recursively – which would, of course, solve the problem of picking out the infinite set of sentences of any language. The determination so to construe natural languages amounted to the hope that it would prove fruitful in the structuralist task of giving a concise and precise specification of the set of sentences of a language. Perhaps it has been fruitful. In any case, these developments were accompanied by a terminological shift. The phonological and syntactic categories were, for Bloomfield and Harris, categories developed by linguists for describing language, and were neither intended nor expected to have any reality to the speaker. They were therefore said to be “convenient fictions.” Chomsky argued otherwise: when people respond to a string of marks, they respond to it differently depending on whether it is grammatical (well-formed) or not. So Chomsky renamed the formulations that captured these structures “theories.” And where Bloomfield had tried to establish linguistics as a semi-independent discipline within a psychology understood behaviouristically, Chomsky has moved or has been moved to refer to it as a “branch of cognitive psychology.” However, for Chomsky, how language performs its role in human life is held to depend, so far as linguistic theory is concerned, exactly what he, as a linguist, is concerned with: linguistic structure. This structure defines the competence of the ideal speaker, and the linguist, according to Chomsky, is concerned to describe this linguistic competence, not actual performance, which, alas! may at times be incompetent. In fact, what the “psychologizing” terminology implies is that linguistics can challenge psychology, the empirical science, while the restriction of linguistics to structure=competence implies that psychology, the empirical science, that deals with what people actually do, that is, with their performance, cannot challenge linguistics. Thus, suppose that one, as experimental cognitive psychologist, argues on the basis of experimental data that language speaks cannot understand certain sentences (e.g. sen-

611 tences that contain more the n embeddings), and therefore concludes that such sentences are not among the sentences constituting the language spoken by the experimental subjects. But, according to the latest fashionable recursive grammar, such sentences (e.g. sentences with any finite number of embeddings) are reckoned grammatical. This grammar defines the competence of all speakers of the language and the performance of the ideal speaker. For the linguist, what the experimental psychologist can provide is a description of the conditions under which the performance of real speakers deviates from that of the ideal speaker. The linguist concludes that the speakers all have the competence of the ideal speaker, it is their performance that falls short: the subjects really do understand the problematic sentences – that is simply a matter of their linguistic competence – while their actual behaviour – their performance – is a matter of external factors (e.g., inadequate memory mechanisms) interfering with the competence. Thus, the linguist is permitted to challenge the claims concerning the structure of language that are made by the (experimental) psychologist, while, aat thesame time, the latter is systematically denied the possibility of challenging the claims concerning the structure of language that are made by the linguist. Linguistics may be defined as a part of cognitive psychology, but cognitive psychology is not allowed to affect it. Apart from the research programme based on the idea of a generative grammar, and the terminological shifts, what Chomsky has achieved for linguistics is to emancipate its practitioners from the constraints of empirical and experimental psychology, and to allow their analytic imaginations complete freedom in the formulation of grammars, or, rather, “theories,” constrained only by their “intuitions” about which forms are correct and which are not. Nonetheless, theories explain: so part of what is involved in the terminological shift is the claim that the new grammars “explain” language forms where the older structuralists merely attempted to describe them. But, since language forms – and any “linguistic intuitions” about which forms are correct – do not just happen but are produced, often deliberately, in contexts, and consequent upon social learning processes, any explanation of these forms – and “intuitions” – must ultimately make reference to somethings inside the skin of the human organism; and this in turn implies that grammars, formal constraints, etc., can in principle be explanatory only to the extent that they can be shown to be “psychologically real.” Thus, while in practice the linguist’s analytic freedom is unconstrained in the Chomskyan shift, it is in theory at least still subject to the constraint of being psychologically real. It is Katz’s object to complete task of rendering

612 linguistics autonomous by freeing it from even this last theoretical constraint on grammatical constructions. On Katz’s view, linguistics ought to be construed as solely about the abstract objects called sentences (sentence types); and when it is so construed, it is (he claims) an a priori science like logic and mathematics.7 Bloomfield and Harris are said to be nominalists, and Chomsky a conceptualist, where Katz is a Platonist. This Platonism does not consist solely in its claim that linguistics is about abstract objects. For that distinguishes Katz’s position neither from that of Chomsky nor from that of Bloomfield and Harris. To be sure, the latter are said to be nominalists because of their professed concern for utterances. But as linguists their concern was with utterance types, i.e. sentences. And Chomsky, too, is concerned with sentences. So everyone holds that linguistics deals with abstract objects. The point is that Bloomfield, Harris, and Chomsky all see linguistics as abstracting from the more embracing science of psychology the point of which to explain why this particular or individual person, say Beverley, uttered such and such a sentence at time t. Bloomfield and Harris abstracted from this broader context, to concentrate on the product of the process, that is, the sentence. As they spoke, they were content to describe the product, rather than attempt to explain its production. But that was done simply as part of the division of labour of science: ultimately, the patterns they discerned in language would take their place within the larger context of the patterns discovered by psychology in general. Psychology therefore, in the long run at least, places constraints on what the linguist can count as a grammar. Chomsky introduces generative grammars, which, as part of this terminological shift, are said to be theories. Katz accepts this terminology, and therefore reads Chomsky as aiming at explanation. But Chomsky, too, holds that the patterns that he discovers in language will take their place in the broader explanatory context, and that psychology, in principle at least, places constraints on grammars. Chomsky holds that grammars explain, so he is not a nominalist, but he holds that there are psychological constraints on admissible grammars. So Chomsky is said to be a nominalist. Katz denies that psychological constraints ought to play a role in linguistics. Grammars are still construed as explanatory, but rather than being of explanatory relevance to the utterance of sentences are said instead to explain the properties of sentences in the way in which the algebraic law for the commutativity of addition “explains” why 1 + 2 = 2 + 1. Grammar is thus solely about the abstract objects called sentences. This is Katz’s Platonism.

613 As was said, Katz infers from this claim about how linguistics explains, that linguistics is an a priori science like logic or arithmetic. This conclusion does not follow. Abstract objects are not in space and time.8 That makes colours also abstract objects, to give another example of objects that are not as such or in themselves in space and time. Consider orange (O), yellow (Y) and red(R). Orange is between yellow and red on the spectrum: (a)

B(Y, O, R)

This statement is about abstract objects. Yet, on any reasonable construal, it is an atomic statement, and any atomic statement is contingent. In contrast, (b)

Fa v ~Fa

and (b’)

B(Y, O, R) v ~B(Y, O, R)

are both necessary. A fact like (a) about abstract objects is atemporal, and therefore true at all times; but it does not follow from this that it is necessary, true in all possible worlds, like (b) or (b’). Katz has simply not thought through clearly enough the ontological notion of an atemproral aspatial object. Nor does he have a clearly explicated notion of logical necessity. Husserl, by the way, whom Katz cites as an ally, also thought he could explicate the notion of necessity in terms of an ontology of abstract objects (his ontology is much more articulate than that of Katz); but, as is well known, it simply won’t do.9 As a consequence of his not having explored thoroughly the relevant ontological ideas, Katz recognizes neither that (a), though about abstract objects, is not a necessary truth, nor that (c)

‘[(p v q) & r]’ is a geometrical pattern exemplifying a relational structure satisfying the (recursive) definition of well-formed formula of propositional logic

while also about an abstract object, to wit, a sentence, that is, a sentence type, is also not a necessary truth. Nor, finally, are we forced to hold as

614 necessary the statements of linguistics to the effect that certain sentences in a natural language exemplify a relational structure satisfying a certain recursively defined grammar. Thus, from Katz’s claim that linguistics is a science solely of abstract objects called sentences, it does not follow that grammar constitutes an a priori theory in the sense of a set of truths that are necessary in the way in which logic and mathematics are necessary. We ca perhaps see the inadequacy of Katz’s inference by recalling that a musical tune is a set of relations among notes, or, more accurately, a set of relations among relations among notes. Now, musical notes are abstract objects. But it does not follow that music is an a priori discipline, nor that a symphonic score is a necessary truth. (At least, no one now seems to follow Boethius in such a radical Platonistic, or Pythagorean, analysis of the being of music.) Thus, the most that Katz can conclude is that linguistics is an a posteriori science aiming to construct grammars analyzing the structures of sentences in natural languages. If he is correct, however, then no psychological constraints need be placed on the grammars developed. On his view, then, linguistics is a radically autonomous discipline. His argument10 for denying the relevance of psychological constraints is that there may well be grammars adequate to a natural language which are not psychologically real. Thus, there may be a grammar G’ that generates all the same sentences as the grammar G of, say, English, where the physiological information processing mechanisms of humans are such that G is realizable in them but G’ is not. G’ might be realizable in some alien race of non-humans. The aliens and humans could communicate; in that sense, they speak the same language. So G and G’ define the same language. Hence, G’ is adequate so far as linguistics is concerned, and since psychological constraints would not rule it out, the latter are not admissible constraints on grammars. Now, the situation described by Katz is no doubt possible. But the conclusion that psychological constraints on grammars are unreasonable does not follow. Only some relations among notes define a tune. But there are many other relations among notes that are musically irrelevant. The relations that are musically relevant are those that we respond to as music. Similarly, many relations among sentences and the parts of sentences are not linguistically relevant. (E.g. in the formula quotes above in (c), it is the order established by the parentheses, and not their specific shapes, such as the size or font in which they are printed, that is grammatically relevant.) Those

615 relations are linguistically relevant that we respond to as such. We suppose G picks out the relational structures to which human English speakers respond. Either G’ picks out those same structures in another way, or it does not. If it does, then the aliens do speak English. If it does not, then it is not, I would argue, a grammar of English, even if it delimits the same class of sentences. After all, it is not just the sounds or marks to which we respond, it is those marks or sounds in so far as they are ordered or structured in a certain way. Grammatical categories are not just “fictions”; that words fit into just certain categories (e.g. noun, adjective and verb) and qua in those categories are related in a certain way (e.g. adjective is followed by noun which is followed by verb, and not noun-adjective-verb), is relevant to whether we respond to a string of words as a sentence or not. Grammar, in other words, is real: after all, that is what Chomsky had argued against Bloomfield and Harris. Katz supposes that his alien grammar G’ picks out the same strings of words as G. G picks them out according to the structures to which we are responding as competent speakers of English. Either G’ picks out the same strings qua structured as in English, or it does not. If it does not, then it is not a grammar of English. So there is a psychological constraint on whether G’ counts as a grammar of English. Katz has thus not made out his claim that linguistics should be autonomous, free to formulate its grammars independently of all psychological constraints. Katz argues that it is one of the virtues of his position that it can resolve certain problems in the philosophy of language. For example, there is an intuitive difference between (d)

Beverley is a bachelor ˆ Beverley is unmarried

and (e)

Beverley is a bachelor ˆ Beverley is unmarried or Beverley is a bartender

The validity of the latter depends on the logical law ‘p ˆ (p v q)’. In contrast, the validity of the former depends upon the meanings of the terms involved. Katz rejects,11 wisely,12 the implicit definition approach to explaining the difference,13 developing the case I had earlier made against the notion,14 which in turn grew out of certain ideas of Bergmann15 and Good-

616 man,16 and, ultimately, Frege.17 Katz argues instead that the intuitive difference between (d) and (e) can be captured only by a grammar that so construes the concepts that the meaning of ‘unmarried’ is contained in the meaning of ‘bachelor,’ i.e. construes the rules to be such that ‘bachelor’ is explicitly defined to be short of ‘unmarried male’, to use the more traditional term which, given the terminological shift in linguistics, it is now fashionable to eschew. Katz is, I am convinced, correct in his insistence upon the relevance of the idea of explicit definition to account for the validity of arguments like (d). Nor is he wrong to construe such rules as among those that define the syntactical structure of the language. But he is wrong to suppose that appeal to such structures can constitute any answer to Quine’s argument concerning necessary truth in his famous paper, “Two Dogmas of Empiricism.”18 Quine did not attack the claim that explicit definition constituted a clear case of necessary truth. That is, it is clear provided that one has a clear case of explicit definition. What he argued was that there are no settled linguistic regularities in natural language that could be systematically construed as explicit definitions. Thus, save for a few exceptional cases like (d), the notion of analyticity, or explicit definition, contrary to what Carnap, for one argued, is not a concept that could be useful to a science of linguistics, that is, useful in expressing the regularities that science aims to discover. In this, Quine is no doubt correct. For example, as Kuhn has pointed out,19 in the everyday practice of science the formula I = VR functions sometimes as a law, and sometimes as a definition of resistance I. Hanson20 has made a similar point about F = ma.. There is no settled disposition to define I in terms of a Wheatstone bridge or in terms of I = VR; if the former, then the formula expresses law or regularity, if the latter, then the formula expresses a definition – if the one, it expresses a synthetic generality, if the other then it expresses an analytic statement. There is no fact of the matter, that is, no linguistic fact of the matter, that determines once for all that this is how I is defined in English or in the language of science. (In fact Kuhn argues that it is an important part of the research strategy of science that there ought not to be an explicit definition settled once for all.) In Katz’s terms, there is a grammatical rule G1 that defines the structure in which resistance is defined by a Wheatstone bridge, and there is a second grammatical rule G2 that defines the structure in which resistance is defined by I = VR. Sometimes it is the language defined by G1 that is the language we use, sometimes it is the structure defined by G2; sometimes it is G1 that is realized in our internal processing mechanism, sometimes it is

617 G2. Most of the time it is settled in neither way, just when it doesn’t matter. That’s Quine’s point. Talk of structures and/or realizations of rules does not meet Quine’s point that there is no settled grammar for the concept of “resistance” or of “force”, or, in fact, most of the concepts of natural language, and that part of natural language which is the language of science. So Katz has not answered Quine. Quine’s argument does not, of course, vitiate the relevance of the concept of explicit definition for an adequate explication of the notion of necessary truth and valid inference. To establish that a concept is not significant for describing linguistic regularities is not to establish that it is not relevant to logic. Nor does Quine’s argument establish the impossibility of the positivist norm for science, that concepts ought to be defined on an observational basis. All that that norm requires is that, when concepts are actually used in inferences, the logical house must be put in order, and they must be explicitly defined. But for them to be defined on those special occasions does not require them to be always defined, or to always have the same definition. Quine’s argument, that there is no settled use, nor, therefore, any settled explicit definition, does not establish that scientists do not set their concepts logically straight when they make inferences, much less that they cannot. And, to repeat, it does not invalidate the positivist norm. Assuming that norm is efficacious in securing conformity to what it enjoins, then, at the relevant times, scientists and other clear thinkers, do realize definite definitional rules for the terms they use. What Katz is suggesting, then, is how the linguist should construe the grammars of (portions of) languages at those (few) times when they are put in logical order. Katz does not refute Quine. But he has succeeded in showing how the positivist norm, or, at least, the consequences it has at relevant times for grammatical structures, can be incorporated into recent accounts of grammar. Bloomfield wanted linguistics to be an empirical science. Katz holds that it cannot be empirical since the entities with which it deals are abstract, not in space and time, and what is not in space and time is not given in the sensory experience that the empiricist holds is the source of data for science.21 This claim clearly depends on what has been called the Principle of Localization, which states that, if an entity is not localized in space and time then it is not given in sense experience. A sentence and the relational structure that constitutes its grammaticality, and so, by the Principle of Localization, it is not given in sensory experience. At the same time, it is true, all agree, that we can recognize sentences as grammatical and ungrammatical, i.e. we do know or grasp these abstract entities. If it cannot be done in

618 sense experience, then it must be by some sort of non-sensory intuition, some sort of knowledge not recognized by the positivist. Bloomfield’s idea of an empirical science of linguistics therefore cannot be realized, it is an impossible goal. Katz accepts that knowledge in linguistics is based on some non-sensory intuition, but he rejects a passive, or, as he says, a perceptual model of such intuition, arguing instead for an activity model; the intuitions we have are a consequence of active information processing based on certain innate capacities of the sort described by Chomsky.22 It is undoubtedly true that the relational structures constituting grammaticality can be known only as a consequence of information processing. But it does not follow from this, as J. S. Mill pointed out long ago,23 that, where the knowledge of an entity or entities is the product of such processing, that entity or entities cannot be given in experience: in fact, there is no reason whatsoever for denying that what is given in experience might well be so given as a consequence of information processing. And, as Herbert Spencer pointed out,24 even if aspects of the processing are a matter of innate capacities, it is still not incompatible with the entity or entities known being given in sense experience. So Katz’s – and Chomsky’s – argument about how we know grammatical structure can be accepted while also accepting the empiricist’s claim that such structures are given in sensory experience. Katz’s case that they are not, and his case that they are given in a way that is incompatible with empiricism, rests upon his acceptance of the Principle of Localization. It is true that Wundt argued for the non-primitive nature of rela25 tions. The early Bloomfield more or less accepted the Wundtian position, and when he later turned towards behaviourism, he tended in his practice to ignore grammatical relations and to stick instead to the simpler elements of phonology. But Spencer and James argued for the phenomenological givenness and introspective irreducibility of (some) relations, and the Gestalters made it final: relations are given in experience and exist as they are given (and Russell established not only their logical acceptability but also their necessity, both in logic and in ontology). Not surprisingly, one of the examples the Gestalters used to make their case were musical tunes: a melody is a structure, not merely the sum of its notes. James insisted on it, Spencer made it clear, the Gestalters established it: the Principle of Localization is not merely a dogma, but a false one at that.26 It follows that Katz’s case fails, and we have no reason to accept his claim against Bloomfield that, by virtue of its concern for grammatical structures, linguistics cannot be an empirical science.

619 Katz’s book is full of ingenious arguments, but its main thesis, that linguistics is an autonomous and non-empirical science, is obviously implausible. It is one that only a philosopher would either find appealing or defend. Not surprisingly, then, Katz, as I have suggested, fails to sustain that thesis.

620 Appendix

We’re going to comment on some aspects of Chomsky’s thought. But before getting on to some specific issues, a general comment in order. It is generally assumed by people who do linguistics or, more generally, cognitive science – it is clearly assumed by Katz – that Chomsky has shown the untenability of something they call “behaviourism.” This Chomsky was supposed to have done in his review of Skinner’s Verbal Behavior.27 Thus, Fodor and Katz say baldly that “Comsky’s paper shows that verbal behavior cannot be accounted for by Skinner’s form of functional analysis.”28 “Chomsky shows”, they say – no qualification, it is taken as a given. But it shouldn’t be, as K. MacCorquodale has shown in his review of the review, “On Chomsky’s Review of Skinner’s Verbal Behavior.”29 The latter makes it perfectly clear that Chomsky barely touches Skinner’s programme. Nonetheless, philosophers like Fodor and Katz are so under the Chomsky myth that they act as if such critiques of Chomsky do not exist. In fact, Chomsky has done little to show that experimental studies of behaviour are useless, and, indeed, the journals remain full of such studies. Chomsky’s review, more belligerent in tone than successful in argument, seems to have persuaded those doing certain sorts of linguistics and certain sorts of cognitive science, that this body of literature can safely be ignored. MacCorquodale shows that it cannot. But his critique is rarely mentioned by those for whom Chomsky has become dogma. That is hardly the stuff of good scholarship. I don’t propose here to go over everything MacCorquodale does. But still, there are things to be said. In particular, many of those who don’t pursue these issues beyond Chomsky’s review, take for granted the apparent success of Chomsky’s claims to have refuted behaviorism and, more importantly, empiricism. But if empiricism can be rejected then there is nothing wrong with Cartesianism, rationalism and even Platonism: Katz’s Platonism comes easily to one who accepts that empiricism has been shown to be untenable. But this rejection of empiricism, of a philosophy that takes a Principle of Acquaintance to be central to its method, should be addressed.

621 I The linguist, since Chomsky, has drawn a distinction between “competence” and “performance.” Competence describes what might be called normal linguistic behaviour; it describes normal linguistic performance. But there is also “incompetent” linguistic behaviour. Both exist, both are effects, both need explaining. It is the task of the scientist to search for such explanations. Chomsky, focussing on competence, may be doing linguistics, but he should recognize that a fully developed science of human being will cover not only competence but also “slips”: linguistics that aims to describe no more than linguistic competence can be no more than a mere part of a more fully developed science of human behaviour. Now, Chomsky suggests that Skinner uses the notion of ‘probability’ in an “obscure” way.30 It may be obscure in details, but it is no so obscure at to be nonsense. Consider an example for Hume. He notes that It would be very happy for men in the conduct of their lives and actions, were the same objects always conjoined together, and we had nothing to fear but the mistakes of our own judgment, without having any reason to apprehend the uncertainty of nature. But as it is frequently found that one observation is contrary to another, and that causes and effects follow not in the same order, of which we have had experience, we are obliged to vary our reasoning on account of this uncertainty, and take into consideration the contrariety of events. The first question that occurs on this head is concerning the nature and causes of the contrariety.

We don’t always say “Lo, a red thing” when we see a red thing, though often we do (or something like it that includes the word ‘red’) and often we don’t pass the salt when someone asks “Please pass the salt” (though quite often we do pass the salt). These are “contrarieties”, as Hume calls then, not in nature but in human behaviour. Hume indicates two ways to respond to such contrarieties, one reasonable and one unreasonable. The vulgar, who take things according to their first appearance, attribute the uncertainty of events to such an uncertainty in the causes, as makes them often fail of their usual influence, though they meet with no obstacle nor impediment in their operation. But philosophers observing that almost in every part of nature there is contained a vast variety of springs and principles, which are hid, by reason of their minuteness or remoteness, find that it is at least possible the contra-

622 riety of events may not proceed from any contingency in the cause, but from the secret operation of contrary causes. This possibility is converted into certainty by further observation, when they remark, that upon an exact scrutiny, a contrariety of effects always betrays a contrariety of causes, and proceeds from their mutual hinderance and opposition. A peasant can give no better reason for the stopping of any clock or watch than to say, that commonly it does not go right: but an artizan easily perceives that the same force in the spring or pendulum has always the same influence on the wheels; but fails of its usual effect, perhaps by reason of a grain of dust, which puts a stop to the whole movement. From the observation of several parallel instances, philosophers form a maxim, that the connexion betwixt all causes and effects is equally necessary, and that its seeming uncertainty in some instances proceeds from the secret opposition of contrary causes.31

When you find a contrariety in nature, one – the scientist –, assuming something like a Principle of Determinism, takes for granted that the difference has a cause, and searches for this factor or variable that makes the difference.32 One has (a) F is frequently followed by G and the Principle of Determinism leads us to formulate the hypothesis that there is a factor H*, which makes the difference, but which, however, we do not yet know specifically what it is. We form, in other words, the hypothesis that (b) All and only F and H* are G The research task is to find specifically what this factor H* is. The vulgar don’t care about the irregularity in nature, they just let it pass – “that’s the way the world is”. But the philosopher, the scientist, who is motivated by curiosity or the love of truth, and knowing that something like the Principle of Determinism does obtain, formulates the hypothesis (b) and sets out to find the relevant factor H*. Now, if we were to know specifically what H* is – say that specifically it is H –, then we would be in a position to predict and explain G by using the law (c) All and only F and H are G but not knowing that H* is H we do not have (c) available: all we have is (b), and not knowing what H* is we cannot identify it antecedently to the occurrence of G, and so we cannot predict when G will occur. Of course, once G has occurred then we can infer the presence of this unknown cause from the occurrence of the effect. In the presence of F, we can use (b) to infer from G the presence of H*, and then turn it about and explain G by citing the presence of F and H*. We do this all the time. We

623 know that germs cause the flu, but often do not know specifically which species of germ it is. Not knowing which germs cause this sort of flu, we cannot predict when it will occur, but if the flu does occur in a person we can infer the presence of “the flu bug”, and then explain the presence of the flu by citing the flu bug: “Beverley has the flu because he (or she) picked up the flu bug somewhere.”33 Still, though we don’t know specifically what H* is and therefore cannot predict G, it would be useful if we were actually able to predict G. We can go some distance towards this, however, before we actually locate H*. We can discover the probability of the effect given the presence of the partial cause which we do already know. With our statistics, we can discover a frequency p such that (d) An F is followed by a G p % of the time Thus, we know the frequency with which smokers die of lung cancer. Not every smoker gets lung cancer; if the smoker gets lung cancer then there must also be present some sort or other of physiological factors that predispose this person to get cancer. If a smoker gets lung cancer then we can infer the presence of these conditions, and explain the cancer by citing the smoking and the presence of these predisposing conditions. But not knowing what those conditions specifically are, we cannot predict whether this or that smoker will get cancer. But we do have (d), and can give the probability that a smoker will get cancer. Then, knowing that probability the person will know what to bet on his or her getting cancer if he or she smokes. One can then estimate whether it is worth one’s while to smoke. In effect, all we can do is use (d) to tell us how to bet whether an F will be a G. In practice, we cannot do much better than the vulgar: (d) is simply a more precise form of (a). However, in our thinking we are much better than the vulgar. For we, or scientists moved by curiosity or love of truth, form the hypothesis (b) and set out on our research programme to locate specifically what the factor H* is: we set out to replace the gappy knowledge (b) by the less gappy knowledge (c).34 What Skinner proposes is a set of functional laws that connect responses R and stimulus F, laws of the form R = f (S) or, more simply, laws of the form All and only f (S) are R which should remind us of (b). But he recognizes that only sometimes f(S) are R, i.e. that we are in fact only in the position of the vulgar and know only that

624 (a’) f(S) are frequently R The schema R = f(S) should be R = f(S, P, H, I) where P = physiology, H = history and I = individual differences. That is, Skinner recognizes that what have in fact as our analogue of (b) is (b’) All and only f(S, P, H, I) are R But we don’t know many of the relevant variables, and at best we have as our analogue of (c) is the relationship (c’) All and only f(S, P*, H*, I*) are R Chomsky accuses Skinner of ignoring the relevance of physiological factors P. Thus, he asserts that “One would naturally expect that prediction of the behavior of a complex organism (or machine) would require, in addition to information about the external stimulus, knowledge about the internal structure of the organism...”.35 But as MacCorquodale points out,36 Skinner in fact again and again draws attention to the necessity of introducing – in due course – physiological variable, ones which relate very often factors in the central nervous system: there are factors, Skinner argues, of the sort P. It is just that for now we don’t know what those factors are – we have P* rather than P. What we should do, instead of trying to replace P* by P, is see whether we can discover the factors in the history of the person, or more generally, of the subject – the details in other words of H* so that we can replace this by H: this is what the search for knowledge of which and how schedules of reinforcement are relevant. This, Skinner is arguing, is likely for now to be the most fruitful research programme. Chomsky is simply arguing that his research programme into facts P* will be more fruitful than that of Skinner into H*. Skinner may be wrong, but that is no reason to get verbally abusive with him. The test will not be argument, but seeing which leads more quickly to filling in the gaps in our knowledge of the laws of psychology. But of course there are still the factors P and I, so that even if Skinner is correct and his programme will prove fruitful and we obtain knowledge of how reinforcement goes, knowledge in other words of H, which can replace H*, we still have at best f(S, P*, H, I*) is frequently followed by R This to put it how the vulgar would put it. Skinner would do it in terms of probabilities: (d’) An f(S, P*, H, I*) is R p% of the time Chomsky complains that Skinner’s notion of a reinforcer unsatisfactory

625 (Skinner states that it increases the strength of an operant which precedes it), arguing, or, rather, asserting that is “perfectly useless ... in the discussion of real-life behaviour, unless we can somehow characterize the stimuli that are reinforcing ...”.37 He is complaining that we can locate the presence of a reinforcer only ex post facto from the presence of the effect R. But as we have just seen with the example of lung cancer, we are often in this position, inferring the presence and the nature, what we know of it, of the cause from the nature of the effect. And we do what we also often do, namely, search for statistical laws like (d) or (d’) when we can’t do better. We rely, in other words, on probabilities when we are looking for ways to predict behaviour when we do not know all the details of the causes. Contrary to what Chomsky asserts, there is nothing “obscure” about using probabilities to do this job, of enabling us to predict as best we can when our knowledge of the relevant variables remains gappy: to the contrary, it is just what we do all the time. Now, Chomsky has his distinction between competence and performance. The language user has his or her competence, and sometimes his or her linguistic behaviour exhibits this competence – his or her verbal behaviour performance is grammatical. But sometimes it is not: sometimes he or she utters strings of words that ungrammatical. Chomsky ignores this contrariety in this person’s verbal behaviour: he responds to it by ignoring it or what causes it, just as the vulgar, in Hume’s discussion, ignore the contrariety in the working of the watch. But as Hume also points out, this is not the response that the philosopher, the scientist, has or, at least, ought to have: the scientist should search for causes, and Chomsky should undertake a search for the causes not only of verbal behaviour which exemplifies the competence of the speaker but also the causes of verbal behaviour, performance, which is contrary to that competence. The scientist should be concerned about the whole variety of human verbal behaviour, both the good and the bad. Chomsky takes up only the good, and ignores the bad. Skinner clearly is concerned with both. In this respect Skinner’s attitude is more that of the scientist than that of Chomsky. In fact, it is one of the virtues of Skinner’s approach that we are forced to deal with verbal behaviour in its whole range, the grammatical but also the ungrammatical.38 Chomsky ignores this strength of Skinner’s project. That may be the job of the linguist, to establish the rules that define the grammatical and distinguish it from the ungrammatical. Let that be the job of the linguist. But do not let that provide an excuse for denigrating and deploring the work of those who, like Skinner, propose to deal, as the

626 scientist, in contrast to the vulgar, should deal, with the broad range of human linguistic behaviour. II Charles Darwin published his On the Origin of Species in 1859, but he felt the argument was only sketched, and went on in further books to flesh out the details, and make the case stronger for the origin of species by natural selection. The Variation of Animals and Plants under Domestication came out in 1868 as the next step in the development of the argument.39 Darwin takes for granted that natural selection is the origin of species: the study begins by a brief exposition of the theory – its structure and the evidence for it. He points out how it is normally the case that populations grow in numbers beyond what the environment can support. There is therefore a struggle for existence. Now there are always heritable variations present, and those variations spread among the population that increase the chances of survival and reproduction. These favourable or fitmaking characteristics gradually accumulate until what was just a variety becomes a new species, better adapted to the environment and which therefore replaces the older species which becomes limited in its range or even extinct. All one needs for speciation is an abundance of variations which can be inherited and a long amount of time – which the geological record provides. This is the origin of species.40 Darwin allows that his account of the origin of species may be considered a mere hypothesis, but, he argues, that hypothesis may be transformed into a probable theory provided that it brings together in one comprehensive scheme, a comprehensive pattern of empirical laws. “In scientific investigations it is permitted to invent any hypothesis, and if it explains various large and independent classes of facts it rises to the rank of a well-grounded theory.”41 His hypothesis does indeed explain large classes of facts; it therefore “ought to be received,”42 that is, received as a wellgrounded theory.43 Darwin contrasts how plausible is this theory and how implausible is its competitor – the notion that each species, all of them, present and past, came to be by a divine creator through originating acts of intelligent design. There are many facts that “have as yet received no explanation on the theory of independent Creation...”:

627 ,,, they cannot be grouped together under one point of view, but each has to be considered as an ultimate fact.

Darwin’s theory is one of simplicity when compared to this competitor that requires “innumerable miraculous creations having been necessary at innumerable periods...”.44 Nor, or course, has anyone ever seen such a species-originating creation occur. In contrast, we do seen natural selection at work all the time, in the work of humans selecting animals and plants for domestication, and in selecting those variations that will, through breeding, survive.45 Darwin did not lack critics.46 The central charge was that Darwin had ignored the need to go beyond what was termed “materialism” and introduce final causes.47 Moral implications too were drawn: the materialism which ignored intelligent design amounted to atheism,48 and, since (it was said) moral order was grounded ontologically in final causes which not only described how things are but determined through divine ordination what objectively they ought to be, this materialism implied a (“merely”) utilitarian morality.49 Acceptance of the theory is therefore morally pernicious.50 The deficiencies of Darwin’s account of the origin of species were many. Some charged that Darwin had not developed his theory on the basis of “induction” – it is “not based on a series of acknowledged facts pointing to a general conclusion.”51 This criticism supposes that Darwin ought to have proceeded on the basis of induction by simple enumeration – as if the reliance on a consilience of inductions, a method to which (as we have noticed) Darwin directs our attention,52 was not a safe method of inductive inference53 – where in fact the simplistic inductive method advocated by Sedgwick is the method of inductive inference that is not safe.54 But among the more serious was the charge that variations within varieties were too few to account for the diversity we observe among the species, and in any case are generally not inherited. Thus, “it has been urged against the theory that existing species have arisen through variation of pre-existing ones and the destruction of intermediate varieties, that it is a hasty inference from a few facts in the life of a few variable plants, and is therefore unworthy of confidence...”.55 Sedgwick makes the same point in a contemptuous way by charging that Darwin holds that “a white bear ... might be turned into a whale,” and that if true “we could hatch rats out of geese...”.56 More soberly, ... the varieties, built upon by Mr. Darwin, are varieties of domestication and

628 human design. Such varieties could have no existence in the old world. Something may be done by cross-breeding; but mules are generally sterile, or the progeny (in some rare instances) passes into one of the original crossed forms.57

This, contrary to Darwin, is evidence that “The Author of Nature will not permit His work to be spoiled by the wanton curiosity of Man.”58 Variation under domestication was questioned, and the heritability of those variations was also challenged. Since the diversity of species required even greater diversity of variation, and required variations to be inheritable, this charge was one that had to be met. This was the task Darwin undertook in the Variations. So he documented in great detail the variations that occur for domesticated species. And he is able finally to draw the conclusion that “the fluctuating, and, as far as we can judge, never-ending variability of our domesticated productions – the plasticity of their whole organisation – ” is a fact59: no hypotheses here. So the critics are wrong about the number of variations, they are not few, but rather incredibly many. From this “we may infer that all natural species, if exposed to analogous conditions, would, on an average, vary to the same degree.”60 Moreover, while it “has often been asserted that important parts never vary under domestication,” in fact “this is a complete error.”61 Moreover, variations are often inherited: “New characters may appear and old ones disappear at any stage of development, being inherited at a corresponding stage.”62 Darwin concludes that the arguments offered by his critics on these points are mistaken: the facts are otherwise. There was another criticism, however, that Darwin felt he should address. Sedgwick, for example, suggested that “final cause would stand good as the directing cause under which successive generations acted and gradually improved.”63 Darwin had not provided or located a mechanism which would allow for the inheritance of variations and would account for the development of an offspring in the form of the parents but including the variations. To help on this point Darwin brings his discussion to a close with a chapter devoted to the “Provisional Hypothesis of Pangenesis.”64 This sketches an account of mechanisms of heredity by means of which characteristics are passed on in reproduction. This must account for the transmission of information from parents to offspring, and how this means of transmission results in the development of the offspring from conception to adult. It is evident that final causes are introduced by their proponents precisely to account for this development of the individual so that it has the appropriate form. Darwin must show how this is possible without the supposition of final causes. He must show something about the mechanisms

629 that could, were they to actually exist, explain processes of inheritance and development; he must show, for example, “how it is possible for a character possessed by some remote ancestor suddenly to re-appear in the offspring...”.65 Now, hitherto in the book Darwin has firmly stuck to establishing the facts in these cases, facts about the immense variability of forms, and facts about heredity, that, for example, characteristics in grandparents may not re-appear in the offspring but then may re-appear in the offspring of the offspring. Some of the facts are plain facts, some are generalizations, but where they are of the latter sort, they are fully supported by strong inductive evidence. When it comes to the theory of pangenesis, however, there is a great contrast, as he himself makes clear: here he is careful to describe it not as “fact” but rather as a “provisional hypothesis or speculation.”66 It involves hypothesized small particles, “gemmules,” given off and received by every cell, bringing information from the environment to the various cells, and transmitting information for parent to offspring, and effecting the development of the offspring into the adults of the next generation. It is a complex hypothesis, but not without its virtues – it certainly does succeed in bringing together a wide variety of facts. Others it misses. The existence of the gemmules is wholly guesswork. The details need not concern us. As it turned out in the history of science, the theory was almost wholly to be discarded. Again, the details need not concern us. For us, the point is that this theory has a different status from the facts and patterns covered by the earlier parts of the book. Those are facts and clearly established patterns. Pangenesis is almost wholly hypothetical. Its presence is not to provide evidence in (further) support of the theory of the origin of species by natural selection. It is there two show “how possibly” these facts can be explained in terms of natural (“materialistic”) mechanisms – to establish that there is no need after all for final causes and special non-material forces.67 Darwin’s theory of pangenesis is what has been called a “how possibly” explanation. Our attention was drawn to this sort of explanation by W. Dray,68 and the occurrence of this sort of explanation in biology was emphasized by T. A. Goudge,69 who referred to them as “narrative explanations.”70 Such “explanations” are not endorsed as worthy of acceptance and for prediction in the way in which ordinary explanations are, as in the case of Newton’s law for the solar system or, for that matter, the theory of the origin of species by natural selection. Rather, they are intended as a response to a question to the effect that, how is such a process possible? gen-

630 erally intended to raise sceptical doubts preliminary to the argument that, such a process is not possible, which is then used to justify the claim that ordinary science, the science of planets and stones, breaks down and we need to introduce non-material forces and laws. In order to perform this task, of showing “how possibly”, it is not necessary that the hypothesis have well-defined terms. They may even be vague, though they must bear some analogies sufficient to recognize how they fit into more rigorous and already accepted theory. Nor need the hypotheses be in any serious way confirmed, though they must at relevant places share some generic logical form with the laws of accepted theories. Now, Skinner’s account of verbal behaviour is far from being a wellconfirmed theory. But it is analogous in form at appropriate points with laws that are well-known and well-confirmed from the situations devised by experimental psychologists. Chomsky criticizes Skinner’s account of verbal behaviour for this lack of empirical support and for the relative vagueness of central terms such as ‘stimulus’ and ‘response’ in the context of language use. He [Skinner] utilizes experimental results as evidence for the scientific character of his system, and analogic guesses (formulated in terms of a metaphoric extension of the technical vocabulary of the laboratory) as evidence for its scope. This creates the illusion of a rigorous scientific theory with a very broad scope, although in fact the terms used in the description of real-life and laboratory may be mere homonyms, with at most a vague similarity of meaning.71

What surprises one is why Chomsky finds this way of formulating hypotheses about verbal behaviour puzzling: of course the crucial terms are used in metaphoric and analogic ways, and even more surprising is why he (Chomsky) uses the terms in a pejorative tone. What Skinner is doing is offering a “how possibly” explanation: how possibly might one be able to describe human verbal behaviour in the ways one can describe non-verbal behaviour in laboratory situations and for nonhuman species? He is attempting to show how there can be laws for verbal behaviour similar to the laws for behaviour of pigeons – not to show these laws with full scientific rigour, as, naturally enough, he cannot do, but to show how possibly the theory might go when it is fully developed. He is attempting to do the same sort of thing as Darwin was attempting to do with his hypothesis of pangenesis. And the target is much the same. Darwin is attempting to show that one need to introduce final causes into biol-

631 ogy: his hypothesis, vague though it is in detail, aims to show that one can in principle account for the facts in a way wholly acceptable to science. Skinner’s target are those who would argue – as Sedgwick argued similarly against Darwin – that there are aspects of human behaviour that elude the net of science, that there are acts of will, of reason, and of moral judgment, all evidenced in our use of language, that set humans off from pigeons, as metaphysically and ontologically different in kind from other, “lower” species. Skinner aims to make implausible the claim that people are different in kind from stones and pigeons by showing how possibly a theory for verbal behaviour might go that was analogous in kind as the theories we have for pigeons.72 It does seem possible that regularities governing verbal behaviour are different in radical ways from what goes on in the rest of the world and in the laboratory. But it seems odd to suggest, as Chomsky seems to do, that there is one set of laws for what goes on in the psychologist’s laboratory and what goes on in, as Chomsky puts it, “real life” – whatever that may be, as if laboratories were not really real. It is rather like those who claimed that the support Darwin claimed for his theory of natural selection came from cases of artificial selection – as if these cases were not themselves part of nature, and subject to the same sorts of laws. We just don’t know whether patterns of verbal behaviour are of the general sort proposed by Skinner – whether or not patterns of verbal behaviour can be understood. But Skinner is not to be faulted for failing to “prove” his hypotheses. He was not attempting to prove anything. He was merely attempting to show how possibly these things actually go, with aim of showing, contrary to the hopes of those who still yearn for final causes, that a science of human being, parallel to a science of stones and a science of pigeons is in fact possible.73 III Chomsky argues that Skinner’s learning theory is inadequate to account for the known facts of language and of language learning in particular. Chomsky’s basic notion is that any language has an infinite number of sentences, each of which is grammatical, and each of which fits into a basic recursive structure.74 Thus, in a simple model, we might have the following rules (call them “R”):

632 ! ! ! ! ! ! !

1. any term of the shape ‘xi’ is a subject term 2. any term of the shape ‘Fj’ is a predicate term 3. any string of the form ‘Fjxi’ is a sentence 4. if ‘p’ is any sentence then so is ‘~p’ 5. if ‘p’ and ‘q’ are any sentences then so is ‘(p & q)’ 6. if ‘p’ and ‘q’ are any sentences then so is ‘(p v q)’ 7. nothing else is a sentence

Thus, the following are sentences according to the rules (R) F1x2 , F1x3 , F2x2 (wf) ~F1x2 , ~~F1x3 (F1x2 & ~F1x3) ((F1x2 v F1x3) & ~~~F2x2) but the following are not F1F2 , ~x1F2 ~(F1x2 & F1x3 v F2x2) In learning a language, a child has to master a set of rules like (R), and in mastering those rules the learner learns to reckon, in both speaking and hearing, that any one of the infinite set of strings like the strings (wf) are grammatical while any other strings are ungrammatical. Now, says Chomsky, imagine a language learner. This learner confronts the world and must infer the rules of grammar from a small set of grammatical strings like (wf). This requires what Chomsky refers to as a process of “theory construction.” The child who learns a language has in some sense constructed a grammar for himself on the basis of his [or her] observation of sentences and nonsentences (i.e. corrections by the verbal community)....[T]his grammar is of an extremely complex and abstract character....[T]he young child has succeeded in carrying out what from the formal point of view, at least, seems to be a remarkable type of theory construction.75

The theory is the set of grammatical rules (R), which the learner takes as describing the structures of (well formed) sentences (strings) he or she has heard but also predicts the from that will be had by any sentence that he or she will hear in the future, and which also determines (in part) the structure of the strings that the learner will produce as he or she communicates with others. The learner learns: what strategy can we attribute to the learner in this process of theory construction? Chomsky emphasizes that the set of strings (wf) could be generated

633 by any of an extremely large number, perhaps an infinite number, of sets of rules like (R). If the learner proceeds by induction by simple enumeration, then he or she will get nowhere: induction by simple enumeration is not a safe rule of inductive inference. So the learner cannot proceed by induction by simple enumeration. Chomsky proposes instead76 that the learner in fact confronts the world with the knowledge that there is a set of specific rules for the language that he or she is hearing and that these rules will all conform to the same generic form. Only a (relatively) small number of sets of rules have this generic form, so the learner can proceed by elimination to discover which among this smallish set of the specific set of rules that constitute the grammar of the language. Seen in this way it is evident that Chomsky holds that one learns a language through induction by elimination. But for induction by elimination to work one needs a Principle of Determinism and a Principle of Limited Variety, the Principle that there is a set of rules and the Principle that these rules that we know to exist must be of a certain generic form, however different they are specifically. According to Chomsky, the learner comes to the learning situation already equipped with the knowledge of these principles, for otherwise the learner could not rapidly conclude that he or she has eliminated all possibilities for the rules defining the language one is hearing save one specific set.77 These Principles of Determinism and Limited Variety constitute a universal grammar. Since the learner comes to the process of language learning already equipped with knowledge of these Principles, this knowledge, or, what is the same, the universal grammar, is innate. Now, to hold that there is innate knowledge is to be a rationalist, and to hold otherwise is to be an empiricist. What Chomsky claims to have shown, then, is the poverty of empiricism. “No one has succeeded in showing why the highly specific empiricist assumptions about how knowledge is acquired should be taken seriously.”78 And so empiricism is to be rejected. Which opens the way for a return to Platonism. So it is claimed. Several remarks need to be made. One. Even if we allow Chomsky’s argument, it hardly opens the way to giving up the empiricist Principle of Acquaintance. For, the knowledge that is claimed to be innate is empirical knowledge of matters of fact. The innate theoretical principles must conform to the empiricist requirement of PA, otherwise they will not be about the world of sense experience in which language is learned. Two. The opposite of empiricism is rationalism, so the suggestion is

634 that the Chomskyan view of linguistics is a return to Cartesianism. Like Cartesianism, there is supposed to be a set of innate principles, so good linguistics is a Cartesian linguistics.79 But thinking historically, Cartesian innate ideas are in fact forms that are themselves entities that transcend the world of sensible experience, and which give us knowledge of entities, like God, outside our world of ordinary experience. Innate or not, the universal grammar which Chomsky suspects is innate to us all is a set of empirical principles that constitute knowledge of this, our world of ordinary experience, and not of a world which transcends this ordinary experience. Three. Chomsky agrees with K. S. Lashley in holding that syntactic structure is “a generalized pattern imposed on the specific acts as they occur.”80 “Imposed”: what could this mean? When a string is heard, then, if it is grammatical, then it exemplifies the appropriate relational structure, a structure the form of which it shares with other grammatical strings – just as the strings (wf) all have a structure which they share with other strings of the form defined by (R). This form which it shares is objectively there in the sentences the learner hears. It is certainly not something subjective and somehow “imposed” upon the strings of sounds that we hear. The task of the learner is to learn what this shared form is to which it is expected that he or she respond. The gestalters showed us that when one responds to a musical tune, one is responding not just to a series of notes but to a relational structure exemplified by the set of notes; and one learns that even if the key is changed so that the notes heard are all different, one is still being presented with the same tune, that is, the same relational structure. Chomsky points to the string Sheep provide wool and argues that this “has no (physical) frame at all, but no other arrangement of these words is an English sentence.” He argues that Furiously sleep ideas green colorless Friendly young dogs seem harmless “have the same frame, but only one is a sentence in English.” He concludes that “It is evident that more is involved in sentence structure than insertion of lexical items in grammatical frames; no approach to language [like that of Skinner, or like that of classical associationists like James Mill] that fails to take these deeper processes into account can possibly achieve much success in accounting to actual linguistic behavior.”81 It is true that the physical frame, “this followed by that” given by the relation of concatenation alone will not serve to separate those forms which are grammatical

635 from those that are not. But there are grammatical classifications (“noun”, “verb”) that are not in ordinary language given by form alone, that is, shape and place in an ordered sequence (as grammatical categories are in our set of rules (R)). These have to be learned. More importantly, there are relations among relations and relations among relations among relations, and so on, as is shown by sentences in our simple language involving ‘&’ and ‘v’. These patterns among patterns are objectively there in the language that is learned. The first and third of Chomsky’s examples have relational structures that share this generic form, while the relational structure of the second does not have this generic form. As (R) makes clear, we have to take into account not just concatenation but also concatenations of concatenations, and concatenations of concatenations of concatenations, and so on. The language learner learns to process information from the environment that enables him or her to identify and respond to these complex relational structures – just as we learn to process information obtained auditorily to identify not just the notes we hear but the tune which they exemplify. The point to be made is that Chomsky’s argument hardly shows that language learning must involve the genetic pre-determination of a universal grammar. If we can learn to respond to the complex patterns of a Mozart symphony without presupposing an innate “universal tune”, then there is no reason to suppose that we cannot learn to respond to the complex structures of grammar without invoking an innate universal grammar. The child does have the capacity to recognize certain complex abstractions with regard to relational structures and the capacity to generalize these to a diversity of new instances. It does not follow that processes of stimulus generalization are unable to account for these.82 It may be hard to conceive, and it may be surprising, but it was Skinner’s objective to develop in outline and hesitantly a theory that could show how possibly this is so. Chomsky has not shown that Skinner’s how possibly explanation fails. IV Chomsky has not shown that linguistics requires one to abandon the empiricist’s PA. At best, he has established the inadequacy of the learning theory of the early empiricists, namely, associationism, or, what he thinks amounts to much the same, Skinner’s reinforcement theory of learning. But, we have just seen, he hasn’t even done that.

636 What of behaviourism? Has he at least not disposed of that? Thus, it has been argued by George Graham, in a well-known encyclopaedia article, that Chomsky established in his review of Skinner, that ...behaviorist models of language learning cannot explain various facts about language acquisition, such as the rapid acquisition of language by young children, which is sometimes referred to as the phenomenon of “lexical explosion.” A child’s linguistic abilities appear to be radically underdetermined by the evidence of verbal behavior offered to the child in the short period in which he or she expresses those abilities. By the age of four or five (normal) children have an almost limitless capacity to understand and produce sentences which they have never heard before.83

Furthermore, Chomsky also argued that it seems just not to be true that language learning depends on the application of reinforcement. A child does not, as an English speaker in the presence of a house, utter “house” repeatedly in the presence of reinforcing elders. Language as such seems to be learned without, in a sense, being taught, and behaviorism doesn’t offer an account of how this could be so.84

The comment about learning the meaning of the word ‘house’ and the capacity to recognize the characteristic of objects to which it refers is simply a travesty of what Skinner proposed. But leave that be as it may. The important point is Graham’s identification of behaviourism with a particular theory of learning. It may be that Chomsky has shown Skinner’s views on how language is learned are implausible (though MacCorquodale does argue in response that he has not shown that). It follows that reinforcement theories of learning do not hold across species, and break down when it comes to human behaviour, linguistic behaviour any way. But why should one hold that behaviourism is to be identified with reinforcement theories of learning? Behaviourism, taken in its most reasonable sense, is the proposition that one can, in principle at least, explain and predict human behaviour using only environmental, physiological and behavioural variables. It is the thesis that laws explaining human behaviour are of the form R = f(S, P, H, I) where R = behavioural response, S = stimulus (often environmental, sometimes internal, i.e., physiological), P = physiology, H = history and I = individual differences. This was the position, the behaviourism, defended by

637 John B. Watson.85 It was – and is – the position that in order to explain and predict human behaviour one does not need to introduce variables refer to mental entities or states of consciousness. The older psychology had made it the business of psychology to analyze mental states. This was analysis in a technical sense of ‘analyze’,86 but whatever the details it was held that one proceeded in psychology by looking inward and attending to conscious states. It was conscious states which were observed, and it was these that formed the subject-matter of psychology. Watson said No to this psychology, it is not consciousness which is, or ought to be, the subject-matter of psychology but rather behaviour, and in fact, he also held, in order to explain and to predict behaviour one need not refer to consciousness at all. This was his methodological behaviourism, and today psychology almost always is done as methodological behaviourism. This is so, whether or not those doing that psychology know it or not. For example, Chomsky in his practice is a methodological behaviourist. He is concerned with linguistic behaviour and he holds that among the relevant variables are environmental stimuli but above all factors built into the central nervous system genetically as the universal grammar. Far from having refuted behaviourism, Chomsky is himself a behaviourist, that is, a methodological behaviourist. It is important to distinguish methodological behaviourism and particular theories of learning. It is in fact a confusion of these two, as in the passage from the Graham Article on “Behaviorism” quoted above, that (mis)leads people into thinking that because Chomsky has shown (it is supposed) that language is in a sense not learned but innate therefore he has refuted behaviourism. But the refutation of a particular learning theory does not invalidate in any way the (methodological) behaviourism that Watson defended and which became, and still is, the standing practice of the science of psychology. It is of course true that, besides methodological behaviourism, Watson also defended a particular theory of learning, namely, classical conditioning. Skinner also defends (methodological) behaviourism, but proposes a theory of learning, namely, a reinforcement theory, which differs in important respects from that of Watson. To refute (even if that has been done) one or the other or both these theories of learning does not show that behaviourism is false. Even if one accepts Chomsky’s arguments that Skinner’s position is inadequate as an account of how language is learned, one need not, and should not also conclude that behaviourism in psychology is somehow dead, or, worse, out of date. There is also the argument that

638

The deepest and most complex reason for behaviorism’s demise is its commitment to the thesis that behavior can be explained without reference to nonbehavioral mental (cognitive, representational, or interpretative) activity. Behavior can be explained just by reference to its “functional” (Skinner’s term) relation to or co-variation with the environment and to the animal’s history of environmental interaction. Neurophysiological and neurobiological conditions, for Skinner, sustain or implement these functional relations. They do not serve as ultimate or independent sources of behavior.87

Notice how the “mental” states that are relevant are “neurophysiological” and “neurological” conditions. Skinner begins his account of learning by considering “operants,” which are items of behaviour with no obvious cause in the environment. He does not deny that they have cause; it is just that the causes do not include environmental stimuli, those causes are therefore likely to be internal – “neurophysiological” or “neurological.” But for what he (Skinner) is about, those causes are not important: they cannot (as yet) be observed, so Skinner proposes that science do what it can do and proceed in psychology to search for laws – inevitably laws that are gappy – which do not include any physiological variables. Watson also held that physiological variables are relevant. But, he also argued, these factors are for the most part on the periphery of the nervous system – thinking or the reasoning, the inner use of language, is a matter of motion in the larynx, and emotional behaviour can be located as in the gonads. The point is that one may disagree with Watson on this point – as indeed one should, since thought or reasoning and our moral sense are almost certainly to be located in the neurological states of the brain – one can disagree with Watson’s peripheralism without giving up his behaviourism, just as one can disagree with Skinner on the relative unimportance (for now) of neurophysiological variables without giving up his (methodological behaviourism). Behaviourism is not refuted just because one thinks it important (as perhaps it is) to refer to neurophysiological variables. It is easy historically to see what pushed Watson to hold to his peripheralism. He was a social reformer, as were many of his time, and reform would be easier if one did not have to as it were penetrate into the central nervous systems of human beings in order to have them learn the skills needed for a better social order. But wishing it were so will not make it so. And Chomsky is surely correct to emphasize against Watson (and Skinner) the importance of the central nervous system. But that does not invalidate methodological behaviourism.

639 There is another feature of Watson’s behaviourism that needs to be noted. Watson was so intent on securing methodological behaviourism that he also went on to deny the very existence of conscious states: he was not only a methodological behaviourist but was also a metaphysical behaviourist. This is one of those positions that C. D. Broad once characterized as “silly”. These are positions that one talks oneself into even though it is perfectly obvious that they are false. Watson is in fact clearly himself not really a metaphysical behaviourist. Thus, he denies that there are states of conscious states of emotion and then proposes to identify them with certain bodily states: but he cannot have it both ways, if they don’t exist then there is no need to identify them with anything, and if they are things you are required to identify with other things then they do after all exist. The issue is not whether conscious states exist. Of course they do, they are given to us in our inner awareness, we experience them all the time. Nor is the issue whether they causally affect our behaviour. Of course they do: I will my arm to go up – that is a conscious state – and my arm goes up – that is a behavioural event. The issue is, how can one recognize the obvious, that mental states do exist and that they do cause behaviour, on the one hand, and also accept methodological behaviourism on the other. The answer is that one does this by recognizing that mental states always have parallel bodily states, parallel neurophysiological states in the brain.88 We have (crudely) (c1) Whenever there is a stimulus S then there is a brain state B. (c2) Wherever there is a brain state B then there is behavioural state R (c3) Whenever there is a brain state B then, and only then, there is a mental state M So suppose Jo experiences stimulus S. This causes Jo to have brain state B. This in turn causes Jo to respond behaviourally with R. We can say that being S caused being R: R = f(S). But the brain state B also played its role. Moreover, if Jo is in S, then he or she also has the state of consciousness M, and M in turn causes Jo to be R. So mind does play its role in causing behaviour. But, while one can explain Jo being R by citing Jo’s conscious state M, one can also explain Jo being R by reference to no variables other than environmental S and physiological B: the causal process which results in behaviour R is closed to outside influence, it completely explains R without reference to mind or consciousness. Commonsensically, mind does cause and explain behaviour, but because mind M has a physical embodi-

640 ment in B – M if and only if B – we can also recognize the truth of methodological behaviourism, that behaviour can be explained without reference to consciousness. This is secured through the fact that there is a parallelism between mental states and bodily states: each mental state has its unique embodiment in a neurological state of the brain.89 Consciousness is there, in the world, and is causally relevant, but for all that one can affirm with the methodological behaviourist the possibility of an objective science of human being. Thus, as our critic of behaviourism, Graham, wishes to assert, “neurophysiological and neurobiological conditions” do enter into the explanation of behaviour, where these are embodiment of “non-behavioral mental (cognitive, representational, or interpretative) activity.” When Graham asserts that “behaviorism’s demise is its commitment to the thesis that behavior can be explained without reference to non-behavioral mental (cognitive, representational, or interpretative) activity,” he is neglecting the fact behaviourism does not, or at least, methodological behaviourism does not, deny a role in the explanation of behaviour of physiological states – states which constitute the mental in the sense of being the embodiment of our conscious states. It is therefore too soon to speak, with Graham of “the deepest and most complex reason for behaviorism’s demise [was] its commitment to the thesis that behavior can be explained without reference to non-behavioral mental (cognitive, representational, or interpretative) activity.” Methodological behaviourism is not committed to that thesis, nor, therefore, could that be a reason for its demise: in fact, there has been no demise – methodological behaviourism is alive and well, whether linguists and cognitive psychologists know it or not, and whether or not they recognize in their own practice that they are methodological behaviourists. V Graham, our critic of behaviourism, hints at another point when he asserts that for the behaviourist “neurophysiological and neurobiological conditions ... do not serve as ultimate or independent sources of behavior.” The suggestion is that mental states or their physiological embodiment can be an independent source of behaviour, and that behaviourism denies this. Many make a reference to Kant at this point,90 suggesting that he (Kant) was on the right track when he spoke of the spontaneity of the mind, its freedom from outside causation, and its contribution through this free ac-

641 tivity of a structure to the world in conformity to its own innate principles. Dealing with Kant would take us to areas that cannot be dealt with in any sort of compact form: suffice it to say that the Kantian mind is metaphysically a substance in the traditional sense, existing through its own activity, and is therefore excluded by the empiricist’s PA from any place in ontology or in any genuine science of human being. But Kant aside, there remains the point that behaviour does occur as it were spontaneously, unprompted it seems by any environmental stimulus. But let us see. Suppose that spontaneously Jo displayed behaviour R. So sometimes the outside environment occurs with behaviour R and sometimes it does not. Does saying that being R, when it occurs, is a matter of spontaneity, provide an explanation? Surely not. That is the response of the vulgar to whom Hume referred when he discussed the wound watch that sometimes worked and sometimes did not. A matter of chance, say the vulgar, and let it go at that. The behaviour R is a matter of spontaneity, says our critic of behaviourism, and lets it go at that. But in neither case, as Hume observes, would such a response be appropriate to the philosopher, that is, the scientist: for the latter, there is some state of affairs, in both cases, watch and behaviour, that makes for the difference, and the task ahead is the research needed to uncover that state of affairs. Our critic of behaviourism, when he or she insists upon spontaneity is simply abandoning the role of the scientist. Of course, our critic might say that naturally Jo’s behaviour R has a cause, what is spontaneous is not the behaviour R but the consciousness which causes it, or, what amounts to the same, the embodiment of that consciousness in the neurological state of the brain. But this does not eliminate the problem. Suppose Jo being in behavioural state R is a consequence of being in brain state B, and that what is said to be spontaneous is the brain state B, or, perhaps, its parallel mental state M. But now we have, sometimes the person is in brain state B and sometimes not. Once again we see that the scientist does not leave it at that a, but goes out to search for the state of affairs that makes the difference between being in B or not being in B. It is the behaviourist who insists upon this further research; our critic of behaviourism, focussing on spontaneity, locates him- or herself among the vulgar.

642 VI In classical psychology, one set oneself up as the observer of conscious states. One attended to the state, gave a phenomenological description of that state, then attended to the parts of that state and gave a second description of the state in a more restricted vocabulary. There were, then, two responses, R and R’, the phenomenological description and the analytical description. The idea was to find a restricted vocabulary such that the phenomenological description R could be inferred from the analytical description R’.91 From the point of view of the behaviourist, that is, the methodological behaviourist, the verbal behaviour R is not the report of an observation. It is simply a piece of behaviour, and the person who utters R is simply the subject of an experiment. From this point of view, what the classical psychologist is doing is searching for R-R laws. Now, there is nothing wrong with R-R laws. Such laws can be very useful, as are, for example, those regularities which correlate two different tests for intelligence. But from the point of view a more complete science, they leave much to be desired. One wants not just R and not just R’ but also the cause of R and the cause of R’. The classical psychologists were doing satisfactory science, but the behaviourists had a better vision of what a science of human being might be when they insisted that one search for laws of the form R = f (S) They were in fact insisting that science should not restrict itself to R but in every case should look for the S that is the cause of R. Consider a Chomskyan student of linguistics. What he or she does is present the subject with a string of words, and ask whether they are a wellformed or grammatical string or not. The subject responds. Oftentimes the linguist lets him- or herself be not only the observer but also the subject. The linguist will give his or her response based on the “intuition” he or she has of whether the string is grammatical or not. That response of the linguist as subject will become part of the data to be used by the linguist as observer. The aim is to sift out those relational structures that evoke the response “grammatical.” Then one constructs a set of “rules” that pick out those structures that are grammatical. One then uses this set of “rules” to predict about certain forms to be heard in the future whether they evoke the response “grammatical.” If that response is evoked then the set of “rules” as a “theory” has done what a theory ought to do, namely, successfully predict. If the prediction is incorrect, then the “theory” is rejected as false,

643 and the set of “rules” modified to try to bring them in line with the data which now include the anomaly. Now note that the linguist is searching for a set of “rules” that will enable him or her to predict further responses. The linguist is searching for R-R laws. In this respect, the linguist has the same limited view of what science is about as the classical psychologists. In contrast, the behaviourist, when he or she insists that the psychologist search for S-R and not just R-R laws is insisting that if psychology is to become a science of human being in the sense in which there is a science of stones then it is laws of this sort that must be the object of research. Thus, the behaviourist has the better view of the sort of science that a science of human being might be.

644 Endnotes to Study Seventeen

1.Ferdinande de Saussure, Cours de linguistique générale (Paris: Payot, 1916); trans. W. Baskin, Course in General Linguistics (New York: Philosophical Library, 1959). 2.For an analysis of such claims about there being a social reality supervening on individual persons and a defence of methodological individualism, see L. Addis, The Logic of Society (Minneapolis, MN: University of Minnesota Press, 1975); and May Brodbeck, “Methodological Individualisms: Definition and Reduction,” Philosophy of Science, 25 (1958) and various essays in her Readings in the Philosophy of Social Sciences New York: Macmillan, 1968). Methodological individualism had already been defended by John Stuart Mill; see F. Wilson, “John Stuart Mill on Psychology and the Moral Sciences,” in J. M. A. Skorupski, ed., Cambridge Companion to John Stuart Mill (Cambridge: Cambridge University Press, 1997), pp. 203-254. Mill, in his defence of methodological individualism, had in mind the views of Coleridge in particular. 3.E. Itkonen, Grammatical Theory and Metascience (Amsterdam: Benjamins, 1978). 4.Popperian views on the use of “falsifiability” to demarcate science are criticized in F. Wilson, The Logic and Methodology of Science and Pseudoscience (Toronto: Canadian Scholars Press, 2000). 5.J. J. Katz, Language and Other Abstract Objects (Totawa, NJ: Rowan and Littlefield, 1981). 6.L. Bloomfield, Language (New York: Rinehart and Winston,1933). 7.Katz, Language and Other Abstract Objects, p. 186. 8.Katz, p. 186. 9.Cf. G. Bergmann, “The Ontology of Edmund Husserl,” in his Logic and Reality (Madison, WI: University of Wisconsin Press, 1964), pp. 193-224. 10.Katz, Language and Other Abstract Objects, pp. 83-92. 11.Katz, p. 111. 12.See the essay in this volume, below, “Implicit Definition Once Again.” 13.Specifically, he rejects Carnap’s version of the implicit definition approach, based on what Carnap calls “Meaning Postulates”; see Carnap’s paper of that title, Philosophical Studies, 3 (1952), pp.65-75, reprinted in his Meaning and Necessity, Second Edition (Chicago: University of Chicago Press, 1956), pp.222-229.

645

Carnap had already made the (unwise) move towards the idea that meaning can be conferred by “implicit definition” when, in 1936, he allowed the meanings of some terms in science to be introduced by so-called “reduction sentences”; see his “Testability and Meaning,” Philosophy of Science, 3 (1936), pp. 419-471, and 4 (1937), pp. 140. 14.See F. Wilson, “The Notion of Logical Necessity in Carnap’s Later Philosophy,” in A. Hausman and F. Wilson, Carnap and Goodman: Two Formalists (The Hague: Nijhoff, 1967), pp. 97-225, in particular Ch. Three. 15.G. Bergmann, “Comments on Prof. Hempel’s ‘The Concept of Cognitive Significance’,” in his Metaphysics of Logical Positivism, first edition (New York: Longmans, Green, 1954), pp. 255-267. 16.N. Goodman, Fact, Fiction and Forecast, third edition (Indianapolis IN: Hackett, 1979). See also R. M. Martin, The Notion of Analytic Truth (Philadelphia, Penn.: University of Pennsylvania Press, 1959); Martin comments that “Meaning postulates are distinguished, apparently, only by the fact of appearing on a page under the heading ‘Meaning Postulates’,” (p. 90). 17.G. Frege, “On the Foundations of Geometry,” translated from the German in the Philosophical Review, 69 (1960), pp. 3-17. 18.W. V. O. Quine, “Two Dogmas of Empiricism,” in his From a Logical Point if View (Cambridge, MA: Harvard University Press, 1953), pp. 20-46. For discussion, see G. Bergmann, “Two Cornerstones of Empiricism,” in his Metaphysics of Logical Positivism, pp. 78-105. 19.T. Kuhn, “Second Thoughts on Paradigms,” in F. Suppe, The Structure of Scientific Theories, Second Edition (Urbana, IL: University of Illinois Press, 1977) 20.N. R. Hanson, “The Law of Inertia: A Philosopher’s Touchstone,” Philosophy of Science, 30 (1963) 21.Katz, Language and Other Abstract Objects, p. 193. 22.Katz, p. 204. 23.J. S. Mill, An Examination of Sir William Hamilton’s Philosophy (London: Macmillan, 1865; reprinted as Vol. ix of Mill’s Collected Works [Toronto: University of Toronto Press, 1979]; references are to the latter), p. 138, p. 178. 24.Herbert Spencer, Principles of Psychology, 2 vols. (New York: Appleton and Co., 1902), vol. ii, p. 195, p. 163. 25.Cf. G. Bergmann, “The Problem of Relations in Classical Psychology,” in his

646 Metaphysics of Logical Positivism, pp. 277-299. 26.Cf. R Grossmann, “Sensory Intuition and the Dogma of Localization,” in E. B. Allaire et al., Essays in Ontology (The Hague: Nijhoff, 1963), pp. 50-63; and also his “Conceptualism,” ibid., pp.30-39. See also F. Wilson, “Weinberg’s Refutation of Nominalism,” Dialogue, 8 (l969), pp. 460-474. 27.B. F. Skinner, Verbal Behavior (New York: Appleton-Century-Crofts, 1957); N. Chomsky, Critical Review of B. F. Skinner’s Verbal Behavior,” Language 35 (1959), pp. 26-58; reprinted in The Structure of Language: Readings in the Philosophy of Language (Englewood Cliffs NJ: Prentice-Hall, 1964), pp. 547-578. Page references to Chomsky’s review are to the latter. 28. J. Fodor and J. J. Katz, Introduction to Section VI, “Psychological Implications”, in their set of readings in the philosophy of language, The Structure of Language, p. 546. 29.K. MacCorquodale, “On Chomsky’s Reveiw of Skinner’s Verbal Behavior,” Journal of Experimental Analysis of Behavior, 13 (1970), pp. 83-99. 30.Chomsky, Review of Skinner, p. 555f. 31.D. Hume, Treatise of Hume Nature, ed. L. A. Selby-Bigge, rev. P. H. Nidditch (Oxford: Oxford University Press, 1978), Book I, Part iii, sec. 12, pp. 132-133. 32.One is working here with Hume’s “rules by which to judge of causes and effects” (see Treatise, Book I, Part iii, sec. 15, p. 173ff), but that is another story (for which see F. Wilson, Hume’s Defence of Causal Inference [Toronto: University of Toronto Press, 1997], and also F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies [Toronto: University of Toronto Press, 1999], esp. Study Five [“‘Rules by which to Judge of Causes’ before Hume”]). 33.Inferences of this sort are discussed in detail in F. Wilson, Explanation, Causation and Deduction (Dordrecht, Holland: D. Reidel, 1985). See also F. Wilson, The Logic and Methodology of Science and Pseudoscience (Toronto: Canadian Scholars Press, 2000). 34.Concerning “gappy” knowledge, as Mackie calls it, or “imperfect knowledge”, as Bergmann calls it, see J. Mackie, “Causes and Conditions,” in E. Sosa, ed., Causation and Conditionals (London: Oxford University Press, 1959), pp. 15-38; and G. Bergmann, Philosophy of Science (Madison, WI: University of Wisconsin Press, 1956), ch 2. Also F. Wilson, Explanation, Causation and Deduction, and The Logic and Methodology of Science and Pseudoscience. 35.Chomsky, Review of Skinner, p. 458. 36.MacCorquodale, Review of Chomsky, p. 88ff.

647

37.Chomsky, Review of Skinner, p. 557. 38.There is a similar virtue in the work of Freud. Before him, slips of the tongue were just that, “slips”, and were treated as the vulgar treated the watch that sometimes didn’t work in Hume’s story. But Freud saw them as facts that needed explanation: they too were part of the broad range of human behaviour and demanded the attention of the scientist just as much as regular, “non-slippy,” behaviour should be investigated and its causes discovered. No one before Freud saw such slips as things worthy of the attention of the scientist. When he investigated them he was being more true to the demands of science that those who ignored them, paying attention only to “competent” behaviour and ignoring the other kinds of “performance,” 39.Charles Darwin, The Variation of Animals and Plants under Domestication, original ed., 1868, rev. ed., 1875, new ed., with foreword by Harriet Ritvo, 2 vols. (Baltimore: Johns Hopkins University Press, 1998). Page references are to the new edition. 40.Darwin, Variation, “Introduction.” 41.Darwin, Variations, vol. 1, p. 9. 42.Ibid. 43.Darwin is correct in his estimation of the strength of support for his theory; see F. Wilson, Empiricism and Darwin’s Science (Dordrecht, Holland: D. Reidel, 1991). Also F. Wilson, The Logic and Methodology of Science and Pseudoscience. 44.Darwin, Variation, vol. 1, p. 12. 45.Darwin, Variation, vol. 1, p. 13.. 46.We are indebted to David Hull for collecting in one volume much of this literature: see D. Hull, Darwin and His Critics (Cambridge, MA: Harvard University Press, 1973)). See in particular the criticisms of Adam Sedgwick, “Objections to Mr. Darwin’s Theory of the Origin of Species,” in Hull, pp. 159-166; originally published in The Spectator, March 24, 1860. 47.Darwin’s theory ought to be rejected “because it utterly repudiates final causes...”. (Sedgwick, p. 166) 48.It implied a “cold atheistic materialism,” it was said. (Sedgwick, p. 161) 49.Darwin’s theory “strips man of all his moral attributes.” (Sedgwick, p. 161) And stripped of these “he becomes entirely bestial.” (Sedgwick, p. 165) Since the theory ignores final causes, acceptance of it “thereby indicates a demoralized understanding on the part of its advocates.” (Sedgwick, p. 166)

648

50.It is “intensely mischievous,” it was said. (Sedgwick, p. 161) 51.Sedgwick, p. 160. 52.See Wilson, Empiricism and Darwin’s Science, for an account of the logic of consilience in Darwin’s argument. 53.As W. Whewell put it, “That rules spring from remote and unconnected quarters should thus leap to the same point, can only arise from that being the point where truth resides”, and “the cases in which inductions from classes of facts altogether different have thus jumped together, belong only to the best established theories ...”. (The Philosophy of the Inductive Sciences, second edition, 2 vols. [London: John W. Parker, 1847], vol. 2, p. 65). John Stuart Mill makes the same point; see his System of Logic, 2 vols., in his Collected Works, vols. VII and VIII, ed. J. Robson (Toronto: University of Toronto Press, 1973), vol. 1, p. 571. 54.For the unsafe nature of induction by simple enumeration, see Mill, System of Logic, vol. 1, p. 311ff. But see also the qualification on this point, vol. 1, p. 569ff. For discussion of this point and of Mill on consilience, see F. Wilson, “Mill’s Logic”, in British Logic in the Nineteenth Century, volume four of the Handbook of the History of Logic, edited by Dov M. Gabbay and John Woods. (Amsterdam: NorthHolland, 2007). 55.Richard Owen, “Darwin on the Origin of Species,” in Hull, Darwin and His Critics, pp. 175-213, at p. 182. (Originally appeared in the Edinburgh Review, April 1860.) Owen is quoting from J. D. Hooker, On the Flora of Australia (London: Lovell Reeve, 1859), p. xxv. 56.Sedgwick in Hull, p. 166, p. 163. 57 Sedgwick in Hull, p. 160. 58.Sedgwick in Hull, p. 160. 59.Darwin, Variations, vol. 2, p. 400. 60.Ibid., pp. 101-102. 61.Ibid., p. 403.

62.Ibid., p. 404. 63.A. Sedgwick, letter to Charles Darwin, Dec. 1859, in Hull, Darwin and His Critics,

649 pp. 157-158, at p. 158. 64.Darwin, Variations, vol. 2, Chapter XXVII. 65.Darwin, Variations, vol. 2, p. 348. 66.Ibid. 67.P. Kyle Stanford, in his Exceeding Our Grasp (Oxford: Oxford University Press, 2006), defends the thesis that induction by elimination can never succeed in even coming close to locating a true scientific theory because there are always alternatives which we have not conceived and therefore which we could not have eliminated – truth therefore always exceeds our grasp. The argument is based on the idea that we know (on the basis of simple induction from the past on scientific theorizing we have known) that the Principle of Limited Variety is never true. The thesis is dubious. It does not recognize the capacity of abstract generic theories, such as the axioms of Newtonian mechanics, to serve as confirmed Principles of Determinism and of Limited Variety in limiting the range of alternatives that must be explored: these Principles, as confirmed in experience, serve to eliminate other alternatives, even those that are unconceived. But Stanford fuzzifies the issues by, for example, suggesting that Aristotelian mechanics and Newtonian mechanics are logically much of a piece as scientific theories (see, e.g., p. 22), when in fact the latter is an empirical science where the former introduces substances and forms as non-empirical entities. (For this issue see F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies, Study One.) Stanford, since he includes non-scientific theories as among the alternatives, naturally has little difficulty in making out a sort of case for his claim. Stanford also discusses Darwin’s theory of pangenesis (Exceeding Our Grasp, Ch. 3). He ignores the fact that for Darwin this theory does not have the same sort of status as a confirmed and therefore genuinely explanatory theory such as the theory that the origin of species is through the work of natural selection. He (Stanford) ignores the fact that pangenesis is not conceived as anything more than a how possibly explanation, and is therefore reckoned as having many unconceived alternatives. That is always so for how possibly explanations, but does not hold for highly confirmed theory such as that of natural selection. No wonder Stanford is able to easily make a case for unconceived alternatives. But this does make plausible his thesis that wellfounded scientific theories, those that are well-confirmed by experiment and by the consilience of inductions, always have unconceived alternatives that have not been eliminated. 68.W. Dray, Laws and Explanation in History (Westport, Conn.: Greenwood Press, 1957), p. 157. 69.T. A. Goudge, The Ascent of Life (Toronto: University of Toronto Press, 1961).

650

70.Goudge, Ascent of Life, p. 74. Goudge, unlike Dray, did not (illegitimately) draw the conclusion that explanations of this sort called into question the standard “covering law” or “deductive-nomological” analysis of explanation, that is, the view that explanation consists of the subsumption of individual facts under matter-or-fact laws or regularities. For discussion of Goudge on these points, see F. Wilson, “Goudge’s Contribution to the Philosophy of Science,” in L. W. Sumner, J. G. Slater and F. Wilson, eds., Pragmatism and Purpose: Essays in Honour of T. A. Goudge (Toronto: University of Toronto Press, 1980). 71.Chomsky, Review of Skinner, pp. 551-552. 72.MacCorquodale, Review of Chomsky, makes much the same point. But he goes into greater detail to evaluate (negatively) the worth of Chomsky’s comments on such Skinnerian concepts as stimulus, reinforcement, and others. 73.Chomsky seems to be among those who wish for a return, if not to final causes, then at least to forces and powers that are beyond the capacity of natural selection to explain and that elude the net of physical law. Thus, he writes that “It is an interesting question whether the functioning and evolution of human mentality can be accommodated within the framework of physical explanation, as presently conceived, or whether there are new principles, now unknown, that must be invoked, perhaps principles that emerge only at higher levels of organization than can now be submitted to physical investigation.” (N. Chomsky, Language and Mind, enlarged edition [New York: Harcourt Brace Jovanovich, 1972]), p. 98. Or again, “With no knowledge of the laws that determine the organization and structure of complex biological systems, it is ... senseless to ask what the ‘probability’ is for the human to have reached its present state...”. (Ibid., p. 97) Or still again, “In fact, the processes by which the human mind achieved its present stage of complexity and its particular form of innate organization are a total mystery....It is perfectly safe to attribute the development to ‘natural selection’, so long as we realize that there is no substance to this assertion, that it amounts to nothing more than a belief that there is some naturalistic explanation for these phenomena.” (Ibid.) But just as plausibly, Chomsky is saying, human mentality and the capacity for language, including the innate universal grammar, might be emergent phenomena that have no naturalistic explanation. So, too, one supposes, as part of that human mentality, is our moral sense. Adam Sedgwick would have approved. It was Darwin’s aim in his Descent of Man (New York: D. Appleton, 1871), to show how possibly human mentality, and in particular human reason and the human moral sense, could in fact be accounted for by biological forces of natural selection. This was part of Darwin’s continuing argument begun in The Origin of Species in 1859. It was Skinner’s aim, too, to show how possibly human mentality could be explained naturalistically. But Chomsky does not treat their how possibly explanations fairly as sketchy hypotheses but simply dismisses them with ridicule and invective,

651 Skinner’s argument explicitly, Darwin’s implicitly. However, it is evident that for Chomsky such dismissal has the virtue of permitting him to hope, as Sedgwick hoped, that what is characteristic of our humanity is beyond, or outside of, but intervening in, the merely natural. 74.See Chomsky, Language and Mind, Ch. 2, pp. 24-64. 75.Chomsky, Review of Skinner, p. 577. 76.Ibid. 77.Chomsky, Review of Skinner, p. 576ff; Language and Mind, p. 65f. 78.Chomsky, Language and Mind, p. 64. 79.Cf. N. Chomsky, Cartesian Linguistics (New York: Harper and Row, 1966). 80.Chomsky, Review of Skinner, p. 575; quoting K. S. Lashley, “The Problem of Serial Order in Behavior,” in L. A. Lawless, ed., Hixon Symposium on Cerebral Mechanisms in Behavior (New York: John Wiley & Sons, 1951), reprinted in F. A. Beach, D. O. Hebb, C. T. Morgan, and H. W. Nissen, eds., The Neurospychology of Lashley (New York: McGraw-Hill, 1960), referring to the latter at p. 512. 81.Chomsky, Review of Skinner, p. 574. 82.MacCorquodale also makes this point. 83.George Graham, Article, “Behaviorism,” Stanford Encyclopedia of Philosophy (Winter 2005 Edition), ed. Edward N. Zalta, URL = . Graham is highly critical of behaviourism. Interestingly, MacCorquodale’s Review of Chomsky’s Review, does not appear in his bibliography: he (Graham) seems not to have read this essay, and certainly his essay might have been the better had MacCorquodale’s arguments been recognized. Graham might also have benefited from G. Bergmann, “The Contribution of John B. Watson,” Psychological Review, 63 (1956), 265-276, which also does not appear in Graham’s bibliography. Also relevant would be L. Addis, “Ryle’s Ontology of Mind,” in L. Addis and D. Lewis, Moore and Ryle: Two Ontologists (The Hague: Nijhoff, 1965); The Logic of Society (Minneapolis, MN: University of Minnesota Press, 1973); and “Behaviorism and the Philosophy of the Act,” NoØs, 16 (1982), pp. 399-420. One might also mention F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill (Toronto: University of Toronto Press, 1990), Ch. 8. 84.Graham, Article, “Behaviorism.”

652

85 The psychologist, not the philosopher. 86.For details, see F. Wilson, “Some Controversies about Method in NineteenthCentury Psychology,” Studies in History and Philosophy of Biological and Biomedical Science, 30C (1999), pp. 91-142. 87.Graham, Article, “Behaviorism.” 88.For greater detail, see Addis, The Logic of Society, and Wilson, Psychological Analysis and the Philosophy of John Stuart Mill, Ch. 8. 89.It might also be, and indeed no doubt generally is the case that we have not so much M if and only if B that is, a one-one correlation, but rather something like M if and only if (B1 or B2 or B3) that is, a one-many parallelism. 90.Including Chomsky; see his Language and Mind, p. 96. 91.See Wilson, “Some Controversies about Method in Nineteenth-Century Psychology.”

Eighteen Implicit Definition Once Again* I If any set of scientific laws is to be expressed in language, logic is presupposed. Consider a language which contains the connectives, the quantifiers, etc., interpreted in the customary fashion. Assume that besides these logical terms, there are also descriptive terms, subject terms, predicate terms, both those of one place and those of several places. Assume also the usual formation rules. The class of all sentences of this language divides into two mutually exclusive subclasses, those customarily called “analytic” and those customarily called “synthetic.” An analytic sentence is one that contains at most vacuous occurrences of descriptive terms. Or, as one also says, an analytic sentence is one either true or false by virtue of its logical form alone. Or, still differently, a true (false) analytic sentence is a sentence that is true (false) for all possible interpretations of the descriptive signs that occur in it. The above are but formulae that must be unpacked. But the results of the unpacking are familiar. So I shall simply use the formulae when referring to that in virtue of which certain sentences are said to be analytic, others synthetic. There are, therefore, two questions to be asked about every sentence. First, is it analytic or synthetic? Second, is it true or false? The answer to the first does not answer the second. The answer to the first determines how one is to go about discovering the answer to the second. Yet they remain two questions and not one. True analytic sentences are often referred to as tautologies. Among them, we can note for future reference, is (E) (x)[fx = gx] e f = g By this formula, any two coextensive predicates are identical. Now consider a domain of science, say a part of chemistry, for which a certain set of laws, 7, is known. Relative to this domain, certain descriptive terms will be primitive; others are explicitly defined on the basis of * Revised version of the original, which appeared as “Implicit Definition Once Again,” The Journal of Philosophy, 62, (l965), pp. 364-374.

654 these primitive ones. Eliminate from 7 all defined terms; then replace the primitive ones by mere marks of the appropriate types. Call the resulting set G. The sentences of G are of the same logical form as those of 7, except that they are uninterpreted. The “descriptive signs” occurring in them are mere marks. It may now be possible to single out a finite subset, !, of G, such that any sentence of G can be deduced from one or more sentences of !, i.e., such that the sentences of ! are a set of premises from which all members of G can be deduced by the usual logical operations. The members of ! are of course called the axioms. An uniterpretated set in which such a finite subset ! has been singled out, I shall call an axiomatic system (G*). An axiomatic system is thus uninterpreted. That is not to say that it cannot be interpreted. For example, the axiomatic system G* that we have just obtained from part of chemistry becomes, upon the obvious interpretation, a true chemical theory. So interpreted, it is an axiomatic development of that theory. But an axiomatic system can be given more than one interpretation. For example, it would be easy to find a false chemical interpretation of G*. G* may also be given a true arithmetical interpretation.1 That makes it an axiomatic development of part of arithmetic. Thus, depending upon the meanings assigned its primitive signs, the resulting sentences of an axiomatic development may be either synthetic (e.g., chemical) or analytic (e.g., arithmetical), and they may be either true or false. If certain meanings are assigned to its primitive descriptive signs, the axiomatic system G* can thus become a set of tautologies, 1. But from the fact that in the sentences of 1 the descriptive terms occur at most vacuously, it does not follow that the original sentences of G are true by virtue of their logical form alone, i.e. true for all possible interpretations of the descriptive signs occurring in them. All that one can say is that certain features of logical form are shared by the sentences of G and of 1.2 Thus, the sentences of an axiomatic system may be synthetic even though, when certain meanings are assigned to the descriptive signs they contain, they become analytically true (tautologies). Let me indicate how I am using my words. As I speak, an axiomatic system (where the descriptive signs are mere marks) is uninterpreted. An axiomatic development is interpreted. An uniterpreted set of sentences is transformed into an interpreted set when a meaning is assigned to each descriptive sign that occurs in the uninterpreted set. Some claim that the uninterpreted axioms of an axiomatic system “implicitly define” the primitive descriptive signs the primitive descriptive

655 terms they contain. Hilbert, for instance, commenting on his axiomatization of geometry, wherein ‘point’, ‘between,’ etc., all occur as primitive descriptive signs, claimed that “The axioms of [order] define the idea expressed by the word ‘between’.”3 In order to evaluate this claim, we should first, by way of contrast, examine the nature of definition, i.e. “explicit definition.” Let ‘Gi’ (I = 1, 2, ... , n) be the primitive descriptive signs of an axiomatic system; and let ‘!(Gi)’ be the conjunction of its axioms. Suppose that, in order to facilitate deduction, defined signs are being introduced. The only condition a definition must fulfill is that the sign that it defines be in all contexts eliminable in favour of those on the basis of which it is defined. Two points are relevant to the issue of “implicit definition.” 1. Suppose that we wished to define ‘g1’ on the basis of (some of) the ‘Gi’. Suppose further that the ‘Gi’ as well as the defined sign ‘g1’ are one term predicates of the first type, and ignore technicalities connected with the replacement of ‘g1’ in those expressions where it occurs in the subject place. That permits us to think of the definition as laid down by (A) ‘g1x’ is short for ‘N(Gix)’ where ‘N’ represents the way in which ‘g1x’ is built up by the connectives and quantifiers from the ‘Gix’. The rule, or rather statement (A) is a statement about the system. Upon its authority, however, (B) (x)[g1x = N(G1x)] which is a sentence in the system, may be treated as if it were a tautology. Thus, for example, one may add it to any set of premises without thereby increasing the descriptive content. However, in and of itself (B) is not a tautology, since it is not true by virtue of its logical form alone. For, as long as one considers (B) by itself and ignores (A), it simply is not true that the descriptive signs occur in it vacuously. Only if one appeals to (A) may one say that (B) is short for (C) (x)[N(Gix) = N(Gix)] which is a tautology. Thus, while computationally (B) may be treated as if it were a tautology, it is false that by itself, i.e. independently of (A), it literally is one. 2. Definition is purely a matter of syntax: it places no limitations upon the interpretations that may be given to the system. Once one has assigned a meaning to each of the ‘Gi’ of the system, then, by virtue of (A), ‘g1x’ will also be assigned a meaning. Furthermore, once the ‘G1x’ are assigned meanings, the equivalence (B) will automatically become a true sentence (though not: automatically analytic); for (A) assures that both the

656 left and right sides have the same designation. Before examining the notion of “implicit definition,” let me make a series of three points which will be useful later. (i) The meaning of an undefined sign is determined by a (semantical) rule of interpretation. A rule of interpretation specifies the extension of a term. The meaning of a defined sign is determined by the definitional rule and the meanings of the terms occurring in the definiens. The terms in the definiens may themselves be defined, in which case they can be replaced by the expressions occurring them. Thus, the definiens can be rewritten so as to contain only undefined terms. The meanings of the terms occurring in the fully expanded definiens are determined by the rules of interpretation. It follows that the meanings of all terms are determined by the linguistic rules of definition and interpretation and nothing else. At least, that is one meaning of ‘meaning’. And it is the one which for our purposes will do. (ii) Linguistic rules determine the meaning of a term. Suppose for the sake of simplicity that the ‘Gix’ are undefined; and that they are given rules of interpretation. (B) is therefore true by virtue of linguistic rules, namely (A) and the rules of interpretation of the ‘Gix’. These rules alone are sufficient to guarantee the truth of (B). Since (B) is true, then, by (E), g1 = N(Gix) Thus, the identity of the definiendum and the definiens is guaranteed by the linguistic rules alone. This is but another way of saying that the definiendum and the definiens have the same meaning. Generally, we may say that two terms have the same meaning just in case that they are coextensive by virtue of linguistic rules of interpretation and definition alone. (iii) By assigning a meaning to each of the primitive signs of an axiomatic system, it is transformed into an axiomatic development. A primitive sign may be assigned a meaning in one two ways: (1) by providing a rule of interpretation for it; or (2) by replacing it with a term (say ‘G’) which is itself defined on the basis of others. In case (2) the definition of ‘F’ will in no sense be about the axiomatic development produced by substituting ‘G’ for a primitive sign of an axiomatic system. The point I wish to make is simple. The primitive terms of an axiomatic development may be either undefined or defined. Hence, the sorts of rules one must examine in order to determine the meaning of a primitive of an axiomatic development will depend on whether the term is undefined or defined. We are ready to examine the claim that the axioms of a system “implicitly define” its primitive signs. The use of the phrase ‘implicit definition’ could only be based on the belief that there exists some fairly close

657 analogy between a sign’s being (“explicitly”) defined and a sign’s occurring as primitive in an axiomatic system. The point I shall make is that the two kinds of “definition” so-called differ in three respects which are so crucial that the use of the same word cannot but be misleading. First: The equivalence (B) is not literally a tautology. Yet it may be treated as if it were a tautology. We are not free to say even this much concerning the axioms of a system. For neither do the primitive descriptive signs in ‘!(Gi)’ occur vacuously; nor, ex hypothesi, since they are primitive signs of an axiomatic system, is there any provision such as rule (A) for their elimination from the system. Second: If ‘!(Gi) is a definition, then it ought to share with (B) the property of becoming automatically true upon interpretation. One may, of course, assign a meaning to each ‘Gix’ so that ‘!(Gi)’ becomes true. Equally obviously, though, one can also so assign meanings that the conjunction of the axioms becomes false. Thus, unlike (B), if the “implicit definition” of ‘!(Gi)’ is to become true, only certain meanings may be assigned to the primitive descriptive signs. This is no doubt what Frege had in mind when he insisted that “...one cannot expect basic propositions and theorems to determine the reference of a word or symbol.”4 In short, “implicit definition,” unlike “explicit definition,” is not just a matter of syntax. Third: One should also note that on the notion of “implicit definition,” the “meaning” of any one of the postulates, or of any terms occurring in them, is “determined” not by the postulate itself but, rather, by the whole set of postulates. Drop one postulate, e.g. in geometry, Euclid’s fifth axiom (the axiom concerning parallels), and the meanings of all the remaining postulates and terms is totally changed. So the axioms of order, in Hilbert’s axiomatization of geometry have a totally different meaning depending on whether one adds or drops Euclid’s fifth axiom. Frege put the point this way: “Only by stating all the axioms, which, according to Hilbert, belong, for example, to the definition of a point, does the word ‘point’ receive a sense. Accordingly, only through the totality of axioms in which the word ‘point’ occurs, does each of them receive its full sense. It is impossible to separate the axioms to regard some of them as holding and others as not holding, because in doing so we would also alter the sense of the ones we wanted to count as holding good.”5 If we wish to hold a subset of the axioms as true while dropping some others, when the others are dropped, the meanings of the remaining ones has changed – we are not after all retaining the chosen subset as true, for the subset we have after dropping the axioms is not the same subset with which we started; the

658 meanings have changed. On the doctrine of “implicit definition,” meaning becomes holistic, with intolerable, or at least very odd, consequences which are entirely absent in the case of “explicit definition.” Is there any point to the notion of “implicit definition”? A small one: We are usually interested only in true interpretations of an axiomatic system. Only certain assignments of meanings to the primitive descriptive signs will result in the axioms becoming true. Which assignments will produce this result depends in part on the logical form of the axioms. So the latter imposes certain limits on the kinds of assignments of meanings yielding true sentences. This much may be said for the notion of “implicit definition.” Having said it, however, one must note what was just pointed out, namely, that “implicit definition”, unlike “explicit definition” is not just a matter of syntax, and is, therefore, holistic in its implications. In view of the indicated differences, I conclude that to speak of the axioms of an axiomatic system as “implicitly defining” the primitive descriptive signs they contain is simply misleading: there is no analogy to the case of (“explicit”) definition.6

II Quine, in one of his essays, comes to the defence of “implicit definition.”7 At least, that is what his title announces. What he actually does is something quite different. He presents, instead, an argument to the effect that if arithmetical truths are analytic then the axioms (“implicit definitions”) of any theory are analytic. Actually, his aim is not so much to defend “implicit definition” as to criticize the analytic-synthetic distinction. This comes out if we present Quine’s argument somewhat less succinctly than we have just done, and, indeed, less succinctly than Quine himself does. Quine starts from the supposed distinction between arithmetical truths, on the one hand, and synthetic truths, such as those of an axiomatic development of (part of) chemical theory, on the other hand. Then he questions this distinction by arguing that for every such synthetic axiomatic development there is an axiomatic development of the same theory, whose truths follow deductively from those of arithmetic. And, he concludes ironically, “So much the worse, surely, for the notion of analyticity.”8 If the argument were a good argument, the irony would be justified. I shall show that the argument does not stand up under examination.

659 Quine takes an axiomatic system ‘!(Gi)’ and gives it three interpretations. First, he interprets the ‘Gi’ into a set of chemical predicates ‘Ci’, such that ‘!(Ci)’ is true. Second, he interprets the ‘Gi’ into a set of arithmetical predicates ‘Ki’ such that ‘!(Ki)’ is true.9 The third interpretation is into a set of predicates defined on the basis of the ‘Ci’ and ‘Ki’. This interpretation requires that the zero-type variables of the first two axiomatic developments range over the same set of individuals. That is, the admissible substitution instances for the zero-type variables, or, as one says, the universe of discourse U of the two theories, must contain both physical objects and natural numbers. There is of course no such U for what we normally understand when one speaks of chemical and arithmetical predicates. Yet, if a language is taken to have such a universe of discourse, then “by a slight and innocuous reinterpretation”10 involving what Quine calls “hidden inflation,”11 the range of values of both kinds of predicates (i.e. chemical and arithmetical) may be appropriately extended to the whole of U as follows. Take an arbitrary element, say a, from the universe U’ of physical objects, and extend the definition of the chemical predicates to numbers by stipulating for all numbers that the chemical predicates are true for them if and only if they are true for a. Extend the arithmetical predicates in the same manner to the physical objects of U. Assume, first, if only for the sake of the argument, that it makes good philosophical sense to treat both physical objects and natural numbers as individuals in the same universe of discourse. Assume, second, again for the sake of the argument, that the inflated predicates have the “same meaning” as the noninflated ones. I believe in fact that neither of these two assumptions makes good philosophical sense. As it happens, however, the point I wish to make and which by itself shows the flaw in Quine’s reasoning, does not depend on this belief’s being justified. So I simplify matters by making these two assumptions.12 We are ready to look at Quine’s third interpretation. He defines the predicates ‘Fi’ which in this interpretation he puts in place of the ‘Gi’, as follows: (D) ‘Fix’ is short for ‘{[!(Ci) & Cix] v [~ !(Ci) & Kix]}’ (I suppose, again for the sake of simplicity, that all predicates involved are one-term and of the first type.13) ‘!(Fi)’ is logically deducible from the arithmetical truth ‘!(Ki)’ alone. This is easily seen. p e {[(p & q) v (~p & r)] = q}

660 is a sentential tautology. Thus, for each i, !(Ci) entails {[!(Ci) & Cix] v [~!(Ci) & K(Ci)]} = Cix Hence, by (D), it also entails that ‘Fix’ is coextensive with ‘Cix’, for all i. It follows that ‘!(Ci )’ entails ‘!(Fi) = !(Ci)’ and therefore ‘!(Fi)’ alone. Similarly, by the sentential tautology ~p e {[(p & q) v (~p & r)] = r} and by (D), ‘~!(Ci)’ entails that ‘Fix’ is coextensive with ‘Kix’, for all i. It follows that ‘~!(Ci)’ entails ‘!(Fi) = !(Ki)’. But if ‘P’ entails ‘Q = R’ then ‘P & R’ entails ‘Q’. Hence, ‘~!(Ci) & !(Ki)’ entails ‘!(Fi)’. ‘!(Fi)’ is thus entailed by both ‘!(Ci) and ‘~!(Ci) & !(Ki)’. Hence it is also entailed by the disjunction !(Ci) v [~!(Ci) & !(Ki)] Because [p v (~p & q)] = (p v q) is a sentential tautology, this disjunction is in turn equivalent to !(Ci) v !(Ki) ‘!(Fi)’ is thus entailed by the disjunction ‘!(Ci) v !(Ki)’, one of whose disjuncts is the arithmetical truth ‘!(Ki)’. So it is entailed by the arithmetical truth alone. As we have seen, ‘!(Ci)’ entails, for each i, that ‘Fix’ is coextensive with ‘Cix’. Furthermore, as a matter of chemical fact, ‘!(Ci)’ is true. Hence, as a matter of chemical fact, for each i, ‘Fix’ is coextensive with ‘Cix’. This is the basis of Quine’s claim that ‘!(Fi)’ is a development of chemistry: “[the] chemical interpretations were ... preserved (extensionally anyway).” And thus, “the erstwhile chemical axioms ... become, under this definitional hocus pocus, arithmetically true.”14 In other words, we have an axiomatic development of chemistry which is arithmetically true. So Quine rejects the analytic-synthetic distinction. The rejection is hasty. The argument contains a flaw. To be sure, ‘!(Fi)’ is arithmetically true: it follows deductively from the arithmetical truth ‘!(Ki)’. The flaw is not to be found here. Again, the definition (D) of ‘Fix’ is perfectly permissible: it does not violate the one condition, eliminability, which definitions must fulfill. So neither does the flaw lie here. We must look elsewhere. Now, there is a common way of picturing definitions. This picture is misleading. For, many definitions, including that of ‘Fix’, do not conform

661 to this picture. So, to exhibit the nature of (D), I shall contrast it with this common way of thinking of definitions. Suppose ‘g1x’ in (A) to be defined on the basis of (some of) ‘G1x’, ‘G2x’, ... , ‘Gnx’, which latter are undefined. The customary picture takes for granted that, in order to determine whether ‘g1x’ is true of some object, say a, i.e. in order to determine the truth of ‘g1a’, one has to do two things and two things only. First one determines the truth values of (the appropriate members of) ‘G1a’, ‘G2a’, ... ,‘Gna’; then one notes the logical structure of the definiens. For example, suppose that in (A) the definiens ‘N(Gix)’ is G1x v G2x In order to determine the truth value of ‘g1a’ one determines first the truth values of ‘G1a’ and ‘G2a’; then one notes that ‘g1a’ is short for the disjunction of these two sentences.15 Now consider ‘Fix’. In order to determine the truth value of ‘Fia’, the two steps of the common picture do not suffice. It is not enough to determine the truth values of ‘Cia’ and ‘Kia’ and then note the logical structure of the definiens. Note how the definiens contains a sentence in which the predicates ‘Cix’ occur but in which ‘x’ does not occur free. This closed sentence is ‘!(Ci)’. Thus, in order to determine the truth value of ‘Fib’, in addition to the two steps just noted, one must also determine the truth value of this closed sentence. This further step is not thought of in our customary picture. Yet this additional step or, better, need is crucial to Quine’s argument. By hypothesis, as a matter of chemical fact, ‘!(Ci)’ happens to be true. Upon this hypothesis, ‘Fia’ is true if and only if ‘Cia’ is true. But now take as our hypothesis the contradictory of the one Quine makes: suppose that as a matter of chemical fact ‘!(Ci)’ happens to be false (always assuming with Quine that ‘!(Ki) is an arithmetical truth). Upon this new hypothesis, ‘Fia’ is true if and only if ‘Kia’ is true. Thus, that each ‘Fix’ has the “chemical interpretation” it has, i.e. that it has the extension it has, crucially depends upon ‘!(Ci)’ happening to be true – as a matter of chemical fact. The flaw in Quine’s argument is his claim that ‘!(Fi)’ is a sentence of chemistry. I reject this claim on the ground that ‘Fix’ and ‘Cix’ do not have the same meaning. The meaning if ‘Fix’ is determined by (D) and by the interpretation rules for the undefined signs that occur in its definiens. But the identity of ‘Fix’ and ‘Cix’ is not guaranteed by the linguistic rules of definition and interpretation alone. To be sure, ‘Fix’ is coextensive with ‘Cix’. Thus, by (E) they are identical. The point is that their identity is not

662 guaranteed by linguistic rules alone. They are identical (coextensive) only if ‘!(Ci)’ is true. If ‘!(Ci)’ were false, ‘Fix’ would not (we saw) be coextensive with ‘Cix’ but with ‘Kix’. And o f course, the truth value of ‘!(Ci)’ is not just a matter of linguistic rules, but rather of chemical fact. Hence, ‘Fix’ does not have the same meaning as ‘Cix’. Thus, ‘!(Fi)’ is not a sentence of chemistry. Quine’s error is like that of one who argues that (S1) ‘The President of France in 1965 is tall’ may be “adopted” in place of (S2) ‘De Gaulle is tall’, and that, therefore, S1 and S2 have the same meaning. Yet S1 and S2 are not equivalent by virtue of linguistic rules alone. Thus they do not have the same meaning. One can, is some sense of ‘adopt’, adopt S1 instead of S2 if the definite description succeeds. And, of course, its success is a matter of fact.16 Quine’s error, though similar, is not so perspicuous. The role of ‘!(Ci)’ is obscured by the picture to which we lazily believe all definitions must conform. The definiens of this picture contains no such sentence. That is why in the case of ‘Fi’ one may overlook its crucial role. Still another confusion, between being a primitive term of an axiomatic development and being undefined, may have misled Quine. The belief that only extensions are relevant to meaning may be a reason for his claim that ‘Fi’ and ‘Ci’ have the same meaning. This belief is erroneous. One must distinguish between undefined and defined terms. In the case of undefined terms, one might say, after a fashion, that meaning and extension coincide. For the meaning of an undefined term is determined by its interpretation rule. In the case of defined terms, however, meaning and extension do not coincide even in this restricted sense. For the meaning of a defined term is determined not only by interpretation rules but also by the definitional rule. If this is not so obvious as it should be in all cases, it certainly is brought home in the case of ‘Fi’. For, as we saw upon examining how its definition differs from the customary lazy picture of definitions, it still depends upon the chemical facts, i.e. upon the truth of ‘!(Ci)’, which set of entities constitute its extension. So, in the case of any defined term, and in the case of ‘Fi’ in particular, meaning and extension do not coincide, even after a fashion. At the same time, ‘Fi’ is a primitive term of an axiomatic development. If one confuses such terms with undefined terms, one might come to believe that, only its extension being relevant to its meaning, ‘Fi’ has the same meaning as ‘Ci’. Quine may, of course, claim that ‘Fi’ and ‘Ci’ have the same meaning merely because he so uses the term ‘meaning’ that, upon this sense

663 both defined and undefined terms have the same meaning. On this use, ‘!(Fi)’ and ‘!(Ci)’ would indeed have the same meaning. But one cannot by a use legislate out of existence distinctions that can be made. One can still distinguish defined and undefined terms; and terms like ‘Fi’, whose extension depends on the truth of a closed synthetic sentence that occurs in the definiens, from terms whose extensions do not depend on the matterof-fact truth of such sentences. And one can thereby distinguish ‘!(Fi)’ from ‘!(Ci)’, which is all that the defender of the analytic-synthetic distinction need do. One who defends the analytic-synthetic distinction holds that the class of all sentences divides into two mutually exclusive subclasses. Thus, he or she must believe that the differentiae of these two classes can be specified. Or, to say the same thing differently, he or she must believe that the distinction can be explicated in a manner that throws light on philosophical issues. (The result, it is usually held, will involve those differentiae hinted at by the formulae at the beginning of this essay.) Such an explication will not be adequate unless the differentiae it proposes place the sentences of all scientific theories in one class, those of arithmetic in the other. Quine argued that no explication could be adequate by claiming that for every axiomatic development of a true scientific theory one can, by means of definitions analogous to (D), obtain an axiomatic development of the same theory whose truths follows deductively from those of arithmetic. If this claim were justified, the resulting sentences would be both scientific and arithmetical. If there were such sentences, it would indeed follow that there is no adequate explication of the analytic-synthetic distinction. But there are no such sentences. Making the distinctions which can be made, one can show, as I believe I have shown, that the two axiomatic developments are not of the same theory. Thus the argument has failed.

But what of “implicit definition”? Regardless of in what else he may have failed, has not Quine at least successfully sustained this notion? To do this, he would have to produce some fairly close analogy between a sign’s being (“explicitly”) defined on the one hand and a sign’s occurring as primitive in an axiomatic system on the other. He has not produced such an analogy. Recall definition (A) and the equivalence (B). Computationally, we saw, (B) may be treated as it were a tautology, whatever meaning is assigned to the ‘Gi’ in the definiens. That is, (B) may be treated as if it were a tautology whatever interpretation it is given. All Quine has done is provide a

664 means by which one can, for every (satisfiable) set of axioms, obtain an interpretation that is tautologous. That there is such an interpretation may be of some interest. But its existence has not the slightest tendency to imply that, whatever interpretation an axiom is given, it may be treated as if it were a tautology. Indeed, as we say in Part I of this essay, some axiomatic systems have several interpretations, sone analytic, some synthetic, some true, some false. This alone shows that the existence of the interpretation that Quine has so cleverly constructed does not justify obliterating the distinction between axioms and (“explicit”) definitions. Quine, we saw, has not compromised the analytic-synthetic distinction. Neither, we now see, has he sustained “implicit definition.”

665 Appendix to “Implicit Definition Once Again” I This was my first published paper. It came from part of my doctoral dissertation, written under the supervision of Professor Gustav Bergmann.17 Quine later re-published his paper in his collection The Ways of Paradox and Other Essays.18 He did not take notice of my criticisms in the original edition of this collection in 1966 nor in the revised and enlarged edition of 1976. Neither did he mention it when he later discussed his essay in his autobiography, Time of My Life.19 Perhaps Quine had not noticed my criticisms, but it would have been nice to think that he had at least read them.20 Certainly, they were drawn to his attention by J. J. C. Smart, in his Critical Notice of The Ways of Paradox, 21 where he remarks, In an interesting paper, ‘Implicit Definition Once Again,’ Fred Wilson has argued that the defender of analyticity ought not to be seriously perturbed by Quine’s argument, since at one point it depends on the assumption of the truth of the chemical axioms. The definitions would therefore not be of the sort which the defender of analyticity would permit as vehicles of transformation of analytic truths into further analytic truths. This seems to me to be a fair criticism of Quine’s paper ...22

Smart does go on to qualify his point, noting that it is “a fair criticism of Quine’s paper if this is read as an attack on analyticity ...”23 – which is how it should be read, or, at least, that is how it is presented. After all, Quine does present as his conclusion the claim that anyone who holds that the truths of arithmetic are analytic ought also to reckon as analytic the truths that are stated in any axiomatized theory whatsoever, e.g, as but one example, the theory of chemistry that he mentions – from which one ought to conclude, Quine asserts, that the notion of analyticity turns out to be, as he has long held, simply a useless or, perhaps, an absurd, notion. If this be Quine’s argument (as I think it is) then Smart agrees with the thesis of “Implicit Definition Once Again” that the argument is unsound – no one who defends analyticity would fail to note that Quine’s “definitions” succeed in their task of making chemical truths into arithmetical truths only if one assumes, not only the “definitions,” but also the truth of the chemical axioms. And therefore, Quine’s argument, being unsound does not in any way call into question the notion of analyticity. But Smart allows that Quine’s essay might be making another point;

666 it is interesting, he suggests, as “a sidelight on the notion of implicit definition.”24 After all [Smart argues], definitions which depend on the assumption of factual truths are quite common in science. Thus [for example], the definition of electrical resistance as the ratio of potential difference to current assumes the truth of Ohm’s Law.25

But let use see. Ohm’s Law deals with the relationship between voltage (V) and current (I) in an ideal conductor. This relationship states that: The potential difference (voltage) across an ideal conductor is proportional to the current through it. Thus, we have (*) V%I where V = V1 - V2 is the voltage across the object, and I is the current through the object. Since V varies directly with I, there is a constant of proportionality; this constant of proportionality is called the “resistance”, R. So Ohm’s Law is given by the formula (^) V=RI where V is the potential difference between two points which include a resistance R, and I is the current flowing through the resistance. Resistance is a function of the material and shape of the object. The inverse of resistivity is conductivity. Now, the logic of this is that, since (*) holds, we know that (**) there is a unique constant of proportionality between V and I Since we know (**) to be true, we can decide to ($) call this unique constant of proportionality between V and R the “resistance” R and we can treat ‘R’ in (^) as another well-defined variable. This means that we can state (*) equivalently as R=V/I Smart is quite correct that we can introduce in this way R as another well-defined variable only because we know that Ohm’s Law (*) is true. But introducing it in this way does not amount to defining it in the sense in which we define ‘bachelor’ as ‘unmarried male’ (in spite of the use by mathematicians of such phrases as ‘well-defined’). We can “introduce” ‘R’ as a well-defined variable only because (**) obtains, and (**) obtains only because (*) obtains. We can of course introduce ‘R’ in any way we wish,

667 but we can use it as a well-defined variable only because (**) obtains, that is, only because, as (**) establishes, certain existence and uniqueness conditions are fulfilled. The point is, we are introducing ‘R’ as in effect a definite description and not as a (nominally) defined concept. This is what Smart misses. In simpler terms, as Russell showed, we can “introduce” the definite description “the F”, or, in symbols, “(,x)(Fx)”, by means of the formula (“Russell’s Rule”) (RR) “G (,x)(Fx)” is short for “(›x)[Fx & (y)(Fy e x = y) & Gx]” Here, ‘(,x)(Fx)’ seems to be introduced by definition as a singular term. But it is not singular term, not in the way a logically proper name ‘a’ is a singular term. For a logically proper name, the inference rule of Universal Instantiation (UI) (x)(Gx) so, Ga holds, as does the identity relation (Id) a=a But neither of these hold for ‘(,x)(Fx)’: for this, UI is invalid (x)(Gx) so, G (,x)(Fx) since the premise might be true but the conclusion false (if either there are no F’s [the existence condition is not fulfilled] or there is more than one F [the uniqueness condition is not fulfilled]); and moreover (,x)(Fx) = (,x)(Fx) might be false (if either the existence condition or the uniqueness condition is unfulfilled). So, ‘(,x)(Fx)’ is not a singular term and does not function logically like a logically proper name. The definite description ‘(,x)(Fx)’ can function logically like a logically proper name, in valid UI and Id contexts, only if existence and uniqueness conditions are fulfilled. Similarly, (+) F (,x)(Fx) far from being an analytic truth, as one might initially suppose, is only contingently true – for it expands by Russell’s rule (RR) into (›x)[Fx & (y)Fy e x = y) & Fx] which is logically equivalent to (++) (›x)[Fx & (y)(Fy e x = y)] which means that (+) is true if and only if the existence and uniqueness conditions for (+) are fulfilled, and whether or not both these hold is a contingent matter. Thus, (+) is true, if at all, if and only if (++) is as a matter of fact true.

668 “Resistance” is in effect a definite description. It can be used as a variable like “V” and “I” only if the uniqueness and existence conditions (**) obtain. But if it is introduced as a definite description then it is not introduced by a nominal or explicit definition. Thus, Smart has not after all shown by way of his example that it is “quite common in science” to have “definitions which depend on the assumption of factual truths.” It is simply not common, in fact it does not happen at all. It is true that in science it is quite common to introduce definite descriptions and functions the use of which in the usual ways does “depend on the assumption of factual truths.” Smart is suggesting that Quine has shown an analogy between implicit definition and explicit definition. They can both, after a fashion, be shown to be necessary. To be sure, to show this Quine uses a transformation in the case of implicit definition which works only if certain factual assumptions are true. But often in science, Smart suggests, we introduce by explicit definition various concepts that can be put to their normal uses only because certain factual assumptions are true. He uses an example to establish that this suggestion is true. But the critic (such as Frege) of implicit definition will argue that the concepts to which Smart is alluding by way of his example are not introduced by explicit definition, but rather are introduced as definite descriptions or functions – where the use in normal ways of definite descriptions and functions does depend on the truth of factual assumptions, in contrast to explicitly defined concepts the use of which requires no such justification. Recall our earlier point that while the chemical predicates were coextensive with the inflated predicates, he had not shown that they had the same meaning – since their being co-extensive depends upon a matter-offact truth. We there made the point that Quine’s error is like that of one who argues that (S1) ‘The President of France in 1965 is tall’ may be “adopted” in place of (S2) ‘De Gaulle is tall’‘, and that, therefore, S1 and S2 have the same meaning. Yet S1 and S2 are not equivalent by virtue of linguistic rules alone. Thus they do not have the same meaning. One can, is some sense of ‘adopt’, adopt S1 instead of S2 if the definite description succeeds. And, of course, its success is a matter of fact. We now see that Smart makes a similar error. So Smart has failed to show that Quine has in any way justified or made legitimate the notion of “implicit definition”: it remains as improper a notion as Frege argued it to be.

669 II Quine seems to holds that if, one can give an arithmetical, or, more generally, a set-theoretical interpretation of a set of axioms of a theory, in chemistry, say, then that theory has been reduced to arithmetic or set theory. It is assumed that the arithmetical or set theoretical meaning is to be taken as primary. For such a one, set theory is all one needs in order to do ontology: anything can be modelled, as one says, in set theory, and that model, if it is adequate, can be substituted, without loss, for whatever it is that it is supposed to model. And so, for example, the “inflated” predicates of the chemical axioms of the theory yield a true interpretation of the same axioms as the original; “uninflated” predicates, then the model obtained by the former of these interpretations can be substituted, without loss, for the original chemical model. Or, for ontology, it can similarly be assumed that instead of talking about properties of objects, we can rather talk about sets of objects, thus reducing properties to sets. Or even, at a level that is perhaps more sophisticated but is even further from ontology, it is assumed that we can conceive of properties as functions from possible worlds (themselves conceived, no doubt, as sets of objects) to sets of objects taken to be “in” or, as the discourse goes, “at” those worlds. Thus, the property red might be taken to be, or “represented” as, a function, taking possible worlds as its arguments, and having as its value for each such argument the set of objects which are in that world red. The process could be extended, taking, for example, functions to be sets of order pairs. But of course, nothing is eliminated by such reductions.26 The models are indeed isomorphic, but it does not follow that these isomorphic models have the same meaning, nor that the theories of which both models are interpretations are somehow identical. Or at least, are sufficiently identical that the needs of ontology are satisfied. Much to the contrary, it is simply false that the set-theoretic model can be substituted without significant loss for the things of the original model which it is said to “represent”. And so, we are aware in experience of red as a property of things – it is presented to us – we are not given in experience the set of objects which are red – even though this set and the property red are in a sense coextensive – and certainly we are not given in experience some function from a possible world to a set of objects in that world – and in any case,

670 what in heaven’s name is a “possible world”? Is the last not philosophically problematic? Does not the notion of ‘possibility’ which occurs in that phrase cry out for explication? What has happened to the red, the property red, that is given to us in our experience of the world (that is, the actual world, not some “possible world”)? Does a set of objects make itself manifest in experience in the way in which the property red makes itself manifest? Simply to ask such questions is to realize that all such “reductions” are spurious. As Reinhardt Grossmann once put it, “ordered couples are no more identical with their coordinated classes than people in a theatre are identical with the chairs on which they sit.”27 To suppose otherwise is simply a way to avoid coming to grips in a serious way with the problems of ontology. It is unfortunate that so many conceive that the best way to do philosophy is to make things look like the things that logicians and mathematicians do. It may look rigorous, it may even be rigorous, mathematically at least, but such rigour is achieved only by giving up on the real issues in ontology – and while in its rigour it may look good, is isn’t ontology, and the ontological problems, whatever the appearances, remain unsolved, and, indeed, undiscussed. In the latter parts of the nineteenth century it was thought by some that the best way to treat of metaphysical problems was to present them in terms of evolutionary theory. It might have been good biology, but it was not good metaphysics. Nowadays we present metaphysical problems in terms of set theory. And nowadays it might be good logic or good mathematics, but it still isn’t good metaphysics. David Lewis, though he of course looks more up-to-date, is hardly better than Herbert Spencer.

III Here is another point about the notion of “implicit definition.” It has been claimed recently, by Robert Hanna in particular, that the necessity of logical truth rests on the fact that the principles of elementary logic (“proto- logic”) are innate structures of the human mind.28 The idea is that such a structure determines patterns of thought and inference, that is, a mind with such a structure must think and infer in these ways; since these structures are innate and must therefore be conformed to in all thinking by any human or by any creature that can be deemed rational; and since we must think and infer in these ways, it is only reasonable that we do so –

671 there is no point, as it were, to fighting such inclinations; it is therefore rational to accept as necessary the basic principles of logic.29 This inference to the rationality of logic clearly depends upon the principle that must implies ought30 – the converse of ought implies can.31 The principle that must implies ought is simply the principle that one make a virtue of necessity: if something must be then it is unreasonable to suggest that the opposite is either permitted or is obligatory.32 It is unreasonable because to hold that something is necessary but also that its opposite is either obligatory or permitted is to hold that it is either obligatory or permitted to do or to be something that it is impossible to do or to be: one will be trying to do or to be something that one cannot be or do, or, in other words, in trying to do or to be that something one will inevitably fail. And trying where one knows one will inevitably fail is simply unreasonable.33 So, one ought to accept the principle that must implies ought.34 Then, since the principles of logic are innate, we must think and infer in conformity to them, it follows that it is reasonable so to do. Were it the case that these things could be so simple! Let us suppose for the sake of the argument that All F are G is one of our innate principles. We must therefore think thoughts that conform to it. That depends upon what ‘F’ and ‘G’ refer to. If ‘F’ designates what ‘crow’ designates in English and ‘G’ designates what ‘white’ designates, then our innate principle is simply false. It may be innate, it may be how we must think, but then we are beings who are required to think something false. What we need to do is have assurance that the concepts in our innate principles so refer to or designate things or properties or relations that the principles are true. Innateness provides no such assurance. What goes for our simple generalization that “All F are G” goes also for the principles of logic. It is supposed to be innate to our ways of thinking and inferring that (f) Fa v ~Fa holds no matter what ‘a’ and ‘F’ refer to. That is, we have (f)(x)(fx v ~gx) But, in the first place, if we are to know that this is true then we have to know what the form represents: we have to know how all this hooks up to the world, which means we have to know what is it, in the world that language is about, that the forms ‘f’ and ‘x’ represent. For, if we don’t know that, then all we have are a meaningless string of marks.35 What we have argued, of course,36 is that this logical form is there, objectively, in the

672 world, and, more specifically, that ‘x’ presents the property of particularity and that ‘f’ represents the property of being not a particular, that is, the property as some would say of universality. Making principles of logic innate will do nothing to make them necessary, by itself all that it does is ensure that innately we juggle symbols in a certain way. It does not assure one of truth, so besides asking which principles are innate one must also ask the ontological question, what is the logical form of the world? But this is often never asked – Hanna for one does not ask it, and so, since he focuses solely on the issue of innateness, he does nothing to establish the rationality of logic. He simply fails to raise the deeper question, the ontological question, answering which is necessary if one is to establish the rationality of logic as opposed to the rationality of moving certain strings of symbols about in certain ways. Moreover, in the second place, there is another issue about logical form that he fails to investigate. This turns on the fact that (f) is supposed to be true in all possible cases. That is why it is taken to be a necessary truth, and, indeed, a logically necessary truth. But (f) can be true in all possible cases if and only if every instance of the ‘Fa’ is either true or false and not both. For, if ‘Fa’ is both true and false, then (f) is false, and if ‘Fa’ is neither true nor false then neither is (f); and therefore in either case, (f) will not be true in all possible case, i.e. in either case it will turn out to be not a necessary truth. If (f) is to be a necessary truth and, more generally, if logic and deductive inferences are to be necessary then it must be true of the world that every atomic sentence purporting to picture a fact in the world must be either true or false and not both. But this is clearly a matter of fact. After all, it is easy to imagine a world in which, for example, basic sentences are both true and false. It is a fact that one can be both in a room and not in a room, if one is standing in the doorway, but in the real world we can distinguish between being partly in the room and partly outside, and therefore not partly in the room, so that our sentences picturing the situation will be either true or false and not both. The point is that it is a fact – a fact about the world – that we can always draw distinctions among the ways the world is in such a way that every basic or atomic sentence will be either true or false and not both. It is this basic fact about the world – that it is as it were two-valued with regard to its logic – that we can ascribe to it the logical from that is reflected in ordinary or even “proto” logic.37 John Stuart Mill saw the relevant point. Aristotle had argued in his presentation of syllogistic logic, that syl-

673 logisms or the second and third figures (and also the fourth) could be reduced to, in the sense of validly deduced from syllogisms in the first figure, which therefore came to be called the “perfect figure.” This justified syllogisms of the other figures in terms of those of the first figure. He then went on to argue the syllogisms which occur in the first figure could be justified by what came to be called the dictum de omni et nullo. This is the maxim that “whatever can be affirmed (or denied) of a class can be affirmed (or denied) of everything included in the class.” In the traditional metaphysics this did indeed have a place. According to that tradition, the proposition (a) S is P is about substantial forms. But it is true about individuals that (b) All S are P in the sense that (b’) All individuals which are S are also P Since no substantial forms exist, according to Mill, this latter is of course the only sense that Mill allows. The traditional system has to account for why (a) implies (b) = (b’). The traditional doctrine has it that the forms as active entities cause the individuals of which they are the forms to be in a way that guarantees the truth of (b) = (b’): what is predicable of the universal or form is predicable of the individual substances subordinate to it. Indeed, the metaphysics guarantees not only that (b) = (b’) is true if (a) is true, but ensures that it is necessarily true.38 Thus, on the traditional view, the dictum de omni et nullo expresses a fundamental principle or law about the ontological structure of the universe; as Mill put it, the principle stated that “the entire nature and properties of the substantia secunda formed part of the nature and properties of each of the individual substances called by the same name...”39 But in a world in which the empiricist’s PA has excluded the substantial forms of Aristotle, the dictum amounts to nothing more than the principle that what is true of a class of certain objects, is true of each of those objects. It is a trivial proposition that does little more than pleonastically explain the meaning of the word class. It has retained its place in logic texts, Mill suggested, only as a consequence of doing what others did who went before, even though the metaphysics which alone justified it has disappeared. There really is, however, Mill goes on to argue, a fundamental principle on which deductive logic, as a set of truths or patterns of inference which are necessary, rests:

674 Every proposition which conveys real information asserts a matter of fact, dependent on laws of nature and not on classification. It asserts that a given object does or does not possess a given attribute; or it asserts that two attributes, or sets of attributes, do or do not (constantly or occasionally) coexist. Since such is the purport of all propositions which convey any real knowledge, and since ratiocination is a mode of acquiring real knowledge, any theory of ratiocination which does not recognise this import of propositions, cannot, we may be sure, be the true one.

This yields two principles, one for affirmative syllogisms – “things which coexist with a third thing, coexist with one another” – and one for negative propositions – “a thing which coexists with another thing, with which other a third thing does not coexist, is not coexistent with that third thing.” Nor are these trivial propositions like the dictum: These axioms manifestly relate to facts, and not to conventions; and one or the other of them is the ground of legitimacy of every argument in which facts and not conventions are the matter treated of.40

The point is that every proposition which attributes a property to an individual is either true or false and not both – properties are, in other words, logically and ontologically wholly contained within themselves, wholly self-contained, separable from and not intrinsically tied to other properties. Contrary to the older tradition which held that one property can be necessarily tied to another property, so that the being of the one implicates the being of the other −, contrary to that tradition, everything is what it is and not another thing. This itself is a matter of fact about the world, a fundamental law of nature; if the world were not this way, ratiocination would not be possible: all other laws of nature, including the laws of logic, rest on this fundamental fact. This fundamental fact about the categorial structure of world is what grounds the truth and necessity of logic, and what therefore makes logic rational. It is, perhaps, true that our tendency to think in terms of the principles of deductive logic is innate, as Hanna argues, and that it is therefore reasonable for a person to confirm his or her thought as it must conform to these principles, but that is not wherein lies either their truth or their necessity: what makes these principles true and what grounds their necessity as constituting the logical form of the world, is itself a fact about the world – it is no doubt the case that this is a deep truth, but it is nonetheless a truth about the world, a matter-of-fact truth.41 Hanna quite misses this point that the rationality of logic depends on the logical form of the world. He proceeds as if one did not have to worry

675 about any factual grounding of the principles of logic. By now we can see where this comes from. It comes from the assumption that the principles of logic like (f) derive their meaning from the whole set of principles that define logic. The assumption is that the principles of logic implicitly define the forms they contain and that the principles are therefore true ex vi terminorum. But the doctrine of implicit definition is bad philosophy for logic just as it is bad philosophy for chemistry.42 Smart’s suggestion, following Quine, that there is something to be said for the doctrine of implicit definition has as little to be said for it in logic as it has in chemistry.

IV Similar remarks about implicit definitions hold also for ontology. Thus, for example, E. J. Lowe has raised the issue of method in ontology, how ought one to proceed? He suggests as an answer a method which “divides ontology into two parts.” One of these parts is “wholly a priori” and the other of which “admits empirical elements.” He elaborates: The a priori part is devoted to exploring the realm of metaphysical possibility, seeking to establish what kinds of things could exist, and, more importantly, coexist to make up a single possible world.

As for the empirical part, this ...seeks to establish, on the basis of empirical evidence and informed by the most successful scientific theories, what kinds of things do exist, in this actual world.

Lowe also indicates what he conceives to be the relationship between these two aspects of the task of ontology. ... the two tasks are not independent: in particular, the second task depends upon the first. We are in no position to judge what kinds of things actually do exist, even in the light of the most scientifically well-informed experience, unless we can effectively determine what kinds of things could exist, because empirical evidence can only be evidence for the existence of something whose existence is antecedently possible.43

His own ultimate conclusion is that there are four categories. There are objects (which appear to be substances), which are characterized by modes

676 (that is, by properties, where these properties are understood to be tropes or perfect particulars). Then there kinds (which appear to be substantial forms) which are instantiated by objects. Kinds are characterized by attributes (forms of modes), which in turn are instantiated by modes. Attributes are exemplified by objects characterized by modes which instantiate those attributes.44 Thus, let Socrates and Phaedo be two individual crows. Socrates and Phaedo are objects. They both instantiate the kind crow. Socrates and Phaedo are both black; the black in Socrates and the black in Phaedo are tropes which are properties or modes of these objects; the black in Socrates and the black in Phaedo are both blacks by virtue of instantiating the attribute blackness, the black itself; the fact that crows are black is constituted by the kind crow being characterized by the attribute blackness; and this fact about forms guarantees that amongst objects, all crows are black. Now, there is, in the first place, the problem raised by Bergmann45 and Earman,46 about what guarantee there is that a fact about forms, such as the fact that the kind crow has the attribute blackness, finds itself reflected in a fact about objects and their properties, in our example, the fact the all crow objects are characterized by modes falling under the attribute blackness: why should the fact that the kind crow is characterized by the attribute blackness find itself reflected in the observed world by all individual crows being black, why could it not be the case that Socrates is black but Phaedo is red?47 More deeply, kinds and attributes are universals and therefore, upon Lowe’s scheme, not among the objects in space and time. They are in effect Platonic forms, and are certainly not given in our ordinary experience of the world, either sensible experience or inner awareness. They are therefore to be rejected by one who begins ontology from the world as we experience it, that is, by one who accepts the empiricist’s PA. This leaves a problem of sameness with regard to the properties of things, as these are taken to be tropes and as particular as the objects they characterize. But an appeal to PA yields the conclusion that properties are not to be construed as tropes, but as universals capable of characterizing several objects: the claim that properties (or, as Lowe speaks, modes) of things are topes is one that is to be rejected through an appeal to PA.48 The point is that, contrary to Lowe, one must begin with the world as we are acquainted with it in ordinary experience, it is only there that we can find categories for any ontology that can claim to be an ontology for this world, the one that is actual, the one that we actually live in and experience. But Lowe would have it that we must establish our categorial

677 scheme a priori and only after that is done do we try to find things or entities that fit those categories in those relations. But how can the terms that purport to refer to the categories that we have devised have any meaning that would enable one to discern whether or not an observed or experienced entity fell into that category? Lowe clearly assumes that the terms in the a priori categorial scheme that one has devised have meaning and referential force. This meaning can come only from the connections laid down as holding among the various categories. That is, Lowe is assuming that the inter-connections among the terms in the categorial scheme by themselves, and prior to any appeal to experience, confer meanings to the concepts the scheme contains. In other words, he is assuming that the categorial concepts receive their meaning from the scheme which implicitly defines them. As we have seen Frege insist, “...one cannot expect basic propositions and theorems to determine the reference of a word or symbol.”49 Frege was speaking about the concepts of geometry and Hilbert’s claim that the axioms implicitly define these concepts. But the same point holds for ontology: implicit definition can no more confer meaning to categorial concepts in ontology than it can in geometry. Lowe is wrong, then, in his claim that ontology begins by establishing a priori a system of categories. Ontology is about the world of ordinary experience, and it is there that ontology must begin: ontology begins with acquaintance, with PA, and only after locating the relevant categories in the world as experienced does one begin to argue, that is, arguing that the categories one has located actually do solve or re-solve or dis-solve the philosophical problems.50 Lowe gets it backwards: one does not do ontology by somehow establishing a set of categories through a priori argument and only then turning to the empirical world; to the contrary, ontology begins with PA and only then does one begin the dialectic.

678 Endnotes to Study Eighteen

1.As is well known, a true arithmetical interpretation can be given for a wide variety of axiomatic systems. See D. Hilbert and P. Bernays, Grundlagen der Mathematik, vol. 2 (Berlin: Springer, 1939; Ann Arbor: University of Michigan Press, 1944), p. 253. 2. A sharing of logical form by synthetic and analytic sentences is very important in the application of logic (arithmetic) to the world, in, for example, measurement. See G. Bergmann, “The Logic of Measurement,” Proceedings of the Sixth Hydraulics Conference (Iowa City: State University of Iowa Publications, 1956), pp. 19-34. 3.D. Hilbert, The Foundations of Geometry (Chicago: Open Court, 1902), p. 5; italics added. Note that this passage occurs before any meanings have been assigned the primitive descriptive signs; meanings are assigned for the first time in § 9 (p. 27), where an arithmetical interpretation is provided so that the axioms might be shown to be compatible, that is, that they form a (logically) consistent set of sentences (provided that arithmetic is consistent). Hilbert does say that his uninterpreted axioms express “fundamental facts of our intuition [of space]” (p. 3); this, however, I take to mean that the uninterpreted axioms have the same logical form as an interpreted axiomatic development of Euclidean geometry which purports to provide a true theory of physical space. Hilbert’s views about “implicit definition” were early criticized by Frege in a review which has been translated as “On the Foundations of Geometry,” Philosophical Review, 69 (1960), pp. 3-17. 4.Frege, “On the Foundations of Geometry,” p. 5. 5.Frege, “On the Foundations of Geometry,” p. 8. 6.See also the discussion of Carnap on meaning as introduced by his concepts of “reduction sentences”, “meaning postulates”, and “A-truth”, in F. Wilson, “The Notion of Necessary Truth in Carnap’s Later Philosophy,” in A. Hausman and F. Wilson, Carnap and Goodman: Two Formalists, (The Hague: Nijhoff, 1967), Chapters Three and Four. Where the positivists had before not used the notion of “implicit definition,” Carnap, ignoring the disastrous consequences, introduced that notion with these changes to the criterion of meaning that earlier positivists had held. In effect, with these notions, Carnap ceased to be a positivist. This occurred in 1936 with the appearance of the notion of “reduction sentences.” See Carnap’s essay, “Testability and Meaning,” Philosophy of Science, 3 (1936), pp. 417-471, and Philosophy of Science, 4 (1937), pp. 1-40; and his later “Meaning Postulates,” Philosophical Studies, 3 (1952), pp.65-75. 7.W. V. O. Quine, “Implicit Definition Sustained,” Journal of Philosophy, 59 (1964),

679 pp. 71-74. See also W. V. O. Quine and Nelson Goodman, “Elimination of ExtraLogical Postulates,” Journal of Symbolic Logic, 5 (1940), pp. 104-109; and C. H. Langford, “Note on a Device of Quine and Goodman,” Journal of Symbolic Logic, 6 (1941), pp. 154-155. 8.Quine, “Implicit Definition Sustained”, p. 74. 9.See note 1, above. 10.Quine, “Implicit Definition Sustained,” p. 72. 11.Ibid.; see also D. Hilbert and W. Ackerman, Principles of Mathematical Logic (New York: Chelsea, 1950), p. 115. 12.But for a discussion of the second of these assumptions, see the Appendix to the present essay, below. 13.Quine actually provides the definitional schemata for predicates of any number of places. 14.Quine, “Implicit Definition Sustained,” p. 73. 15.This is a very common picture, in spite of its being misleading. The major examples of definitions given in introductory and text books usually conform to it. It appears even in studies customarily taken for the best by those who should know better; thus, see for example, C. G. Hempel, Fundamentals of Concept Formation in Empirical Science (Chicago: University of Chicago Press, 1952), p. 2ff. 16.Essentially the same point is made in Langford, “Note on a Device of Quine and Goodman.” Quine never noticed, or at least never acknowledged, the relevance of Langford’s point. 17.I am grateful for the discussion and the inspiration Bergmann provided: one could not have wanted a better supervisor. 18.W. V. O. Quine, The Ways of Paradox and Other Essays (New York: Random House, 1966); Revised and Enlarged Edition (Cambridge, MA: Harvard University Press, 1976) 19.W. V. O. Quine, Time of My Life: An Autobiography (Cambridge, MA: MIT Press, 1985). 20.But Quine seemed regularly to deal with criticism by not noticing it. Gustav Bergmann developed devastating criticisms in his “Two Cornerstones of Empiricism” (in his Metaphysics of Logical Positivism [New York: Longmans, Green, 1954], pp. 78105) of Quine’s oft cited essay on “Two Dogmas of Empiricism,” Philosophical Review, 60 (1951), pp. 20-43 (reprinted in his From a Logical Point of View [Cambridge,

680 MA: Harvard University Press, 1953], pp 20-46), but Quine simply ignored them. 21.J. J. C. Smart, Critical Notice of W. V. O. Quine, The Ways of Paradox and Selected Logic Papers, Australasian Journal of Philosophy, 45 (1967), pp. 92-104. 22.Smart, p. 94. 23.Smart, p. 94. 24.Smart, pp. 94-95. 25.Smart, p. 95. 26.This point is made strongly, and convincingly, by R. Grossmann in his Ontological Reduction (Bloomington, Indiana: Indiana University Press, 1973). For a similar point, see B. Smith, “On Substances, Accidents and Universals: In Defence of a Constituent Ontology,” Philosophical Papers, 26 (1997), pp. 105-127, at p. 207; and also E. J. Lowe, The Four-Category Ontology (Oxford: Oxford University Press, 2006), pp. 5-6. 27. R. Grossmann, The Categorial Structure of the World (Bloomington, IN: Indiana University Press, 1983), p. 164. 28.See R. Hanna, Rationality and Logic (Cambridge, MA: MIT Press, 2006). 29.Hanna, p. 24ff. 30.However, Hanna, op. cit., does not recognize that his argument, if it is to be sound, must assume this principle as a premise – this is one of several gaps in his argument. 31.The roots of the must implies ought and ought implies can in standard deontic logics are explored in detail in F. Wilson, “Mill’s Proof that Happiness Is the Criterion of Morality,” Journal of Business Ethics, 1 (1982), pp. 59-72. See also F. Wilson, Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1997), Ch. II, sec. 5. 32. We don’t want it to be the case both that p is necessary (“must”) and that its opposite ‘not-p’ be either obligatory or permitted:. To put it in symbols, where ‘Np’ means that ‘p is necessary’, ‘Op’ means that ‘p is obligatory’ and ‘Pp’ means that ‘P is permitted’, what we want is that (*)

~[Np & (O~p v P~p)]

This is logically equivalent (by de Morgan’s rule) to ~Np v ~(O~p v P~p) which in turn is logically equivalent to

681

Np e (~O~p & ~P~p) But we can take as given the usual logical relations between ‘O’ and ‘P’: ‘~O~p’ ‘Pp’ and ‘~P~p’ ‘Op’, so that this conditional is logically equivalent to Np e (Pp & Op) But, given that the standard relations between ‘O’ and ‘P’ make it necessarily true that Op e Pp this last turns out to be logically equivalent to Np e Op Thus, the principle (*) logically implies that must implies ought. 33. Thus, we have justified so organizing our moral principles that (*) [see previous note] obtains; this is what the reasonable person will do. One can put the point in another way. The reasonable person, if they will the end they will also will the means; but if something is impossible, then there is no means for achieving it; in the absence of a means it will be unreasonable to will the end – one will be setting oneself u for an inevitable failure, and the reasonable person does not do that. So the reasonable person conforms their moral discourse to the principle (*) [see previous note], and therefore to the principle that must implies ought. 34.See Wilson,. Hume’s Defence of Causal Inference, p. 218ff. 35.This is the point G. Bergmann was making in his essay “Ineffability, Ontology and Method,” in his Logic and Reality (Madison, WI: University of Wisconsin Press, 1965). He was, of course, following the lead of Wittgenstein in the Tractatus – though Bergmann’s argument was always clear, where Wittgenstein’s often was not. 36.See the essay, above, on “Effability, Ontology, and Method.” 37.See our discussion, above, on “Grossmann on the Categorial Structure of the World. 38. The problem of having (a) support (b) = (b’) is discussed by Bergmann in his essay “On Non-Perceptual Intuition,” in his Metaphysics of Logical Positivism (New York: Longmans, Green, 1956), pp. 228-231. These issues are discussed above, in the essay on “Bergmann’s Hidden Aristotelianism.” For greater detail, see F. Wilson, Logic and th e Methodology of Science in Early Modern Thought (Toronto: University of Toronto Press, 1999).

682

39.John Stuart Mill, System of Logic, Bk. II, ch. ii, § 2; in his Collected Works, ed. J. Robson (Toronto: University of Toronto Press, 1974), vol. 7, p. 174. 40. System of Logic, II, ii, § 3, Collected Works, vol. 7, pp. 177-8. 41.As Bergmann at one time put it, since the truth and the necessity of logic as a set of necessary truths rest on a (deep) matter of fact about the world, “there is nothing logical about logic.” 42.See also A. N. Prior, “The Runabout Inference Ticket,” Analysis, 21 (1960), pp. 3839. 43.E. J. Lowe, The Four-Category Ontology, pp. 4-5. 44.Ibid., p. 18. 45.G. Bergmann, “On Non-Perceptual Intuition.” 46.John Earman, “Laws of Nature: The Empiricist Challenge,” in R. J. Bodan, ed., D. M. Armstrong (Dordrecht, The Netherlands, 1984), p. 221n21. 47.We have discussed this problem above, in the essay “Bergmann’s Hidden Aristotelianism.” 48.See the discussion above, in the essay “Universals, Particulars, and Tropes: The Role of a Principle of Acquaintance in Ontology.” 49.Frege, “On the Foundations of Geometry,” p. 5. See note 4, above. 50.Compare the discussions in the essays above, “Acquaintance, Ontology and Knowledge” and “Effability, Ontology, and Method.”

Nineteen Dummett’s History: Critical Review of Michael Dummett’s Origins of Analytical Philosophy * Michael Dummett, in his Origins of Analytical Philosophy,1 has ventured to look at the structural connections that lie in the common roots in Frege and Brentano of his own sort of linguistic philosophy, on the one hand, and current phenomenology deriving from Husserl, on the other. The result is an important work that historians of the origins of contemporary philosophy will find indispensable. Dummett's study is not exactly history but rather what Gustav Bergmann called “structural history” − the exploration of the structural connections between historically identifiable positions.2 Bergmann was sometimes criticized for undertaking such investigations − he should have done real history instead. Here, “real history” is the narrative and explanatory order the details of which are revealed by the causal explorations of the historian. The point is, however, that any such causal explanatory story must take account of structural connections among the positions that are related in the narrative. The causal story thus presupposes the sort of work that Bergmann often undertook, and that Dummett has now undertaken. Dummett argues that any story of the origins of analytic philosophy must look beyond the standard story of a revival of empiricism. Like Bergmann and the latter’s student Reinhardt Grossmann, Dummett argues that one must also look at both Frege and Brentano, and in this context Bolzano, Meinong and Husserl.3 What is called “continental” philosophy looks back to Hegel and Kant, and we are inclined to think of these figures as characteristic of German philosophy. Far from it! The idealism deriving * Originally appeared as a Critical Review of Michael Dummett, Origins of Analytical Philosophy, Canadian Journal of Philosophy, 27 (1997), pp. 377-406

684 from Kant dominated the north and east of Germany, Protestant Germany in other words, but in the south and west, the Catholic dominated areas, the Aristotelian heritage was considerably stronger, a tradition which combined a sort of realism with a sympathy for empiricism. Historians of psychology have long recognized this division,4 and if philosophers are to do justice to the history of analytic philosophy we too must recognize it.5 In particular, the thought of Brentano was almost indispensable, not only with regard to developments in German-speaking Mitteleuropa, but also with regard to G. E. Moore and therefore Bertrand Russell, via the impact of Brentano's psychology on Moore’s teacher James Ward.6 The importance of Frege has, of course, long been recognized. There was, on the one hand, his impact on Russell and Wittgenstein, and, a little later, on Carnap. And on the other hand, there was a further impact when various “Oxford” philosophers after the Second World War, despising Russell but feeling pressure to study logic, turned to Frege for help. In the recorded interview attached to this book as an Appendix, Dummett judiciously locates his own views relative to these post-war developments of philosophical thought in England. Dummett offers the following characterization of the “analytical philosophy” the structural origins of which he is discussing: What distinguishes analytical philosophy, in its diverse manifestations, from other schools, is the belief that a philosophical account of thought can be attained through a philosophical account of language, and, secondly, that a comprehensive account can only be so attained (p. 4).

Now, from this description, it would seem that what Dummett is talking about could be more correctly characterized as “linguistic philosophy.” Certainly, it is hard to see from this characterization why it should be called “analytic.” This term in fact makes sense if we look further into the background of analytic philosophy. For, in that background in Britain in particular was Hegelian idealism, especially that of Bradley. On this version of idealism, nothing is apart from its relations to other things; these relations determine its very being. To isolate an entity, then, is to abstract it from the relational context that determines precisely what, in its true being, it really is. Analysis so isolates, and therefore falsifies. Thus, on this version of idealism, to analyze is to falsify, and knowledge can come only from synthesis, never from analysis. It was Moore’s major contribution to insist, as Locke had earlier insisted, that a thing is what it is independently of its relations to other things: relations yield facts about an entity, but do

685 not determine what the entity is in its very being.7 With this it became possible to hold that analysis not only does not falsify but can much to the contrary actually yield knowledge. When Moore argued that “A thing becomes intelligible first when it is analysed into its constituent parts,”8 he was denying something that many thought Bradley had made an irrefutable truth of philosophy. But deny it he did, and in fact made it the core of a new movement in philosophy, what came to be called, rightly, “analytic philosophy.” Dummett’s definition of ‘analytic philosophy’ is such as to ignore this aspect entirely. Be that as it may, Dummett does signal an important strand in the heritage of Moore and Russell – although his definition clearly characterizes, and is meant to characterize, his own philosophy, it also picks out a tradition that extends well beyond Dummett’s own thinking. Dummett’s definition does pick up a theme that emerged only gradually in the history of analytic philosophy. This is what Bergmann called the “linguistic turn.”9 In part, the background lies in the work of the logical positivists. Taking off from some themes of the Tractatus, the positivists argued that one could explain the necessity of necessary truths in terms of their being made true by virtue of the syntax of the language one was using. One could thus, on this view, come to understand a certain traditional philosophical but problematic issue, namely that of the nature of necessity, by talking about one’s language. The positivists generalized this notion: all traditional philosophical claims could be understood as disguised ways of talking about one’s language. But, they also argued, the choice of a language is merely a matter of taste or convenience; there are no cognitive issues involved in the choice of a language. This was the philosophical nihilism of the positivists: philosophical issues, now understood as claims about language, were non-cognitive. Of course, there were disagreements about which language was the most convenient. Any language had to contain non-logical constants, and these had to be interpreted. Specifically, given the empiricism of the positivists, these constants had to be interpreted into entities with which we are acquainted in ordinary sensory and perpetual experience and in inner awareness. Traditionally, the proper language, it was held, should contain two types of non-logical constants, zero-level or individual constants, on the one hand, and first-level predicate constants, on the other, with the latter of one, two or n places. The former were to be interpreted as designating individuals, the latter as designating the properties of, and relations among, individuals. But this was challenged by some. Goodman, for example, argued that the basic language need contain only

686 one type of constant. There were arguments for and against this proposal. Bergmann, who had made himself thoroughly familiar with the work of Moore and Russell, recognized that this disagreement was merely restating in a new way some of the traditional dialectic of the realismnominalism issue. Assume that the same predicate term, interpreted into some observable property, appears in two different but both true subjectpredicate sentences. That reflects in linguistic terms the traditional idea that properties are universals that can be exemplified by one or several particulars. In contrast, if you hold, with Goodman, that there is only one type of non-logical constant in one’s language, then one is denying the realist’s claim. Bergmann could thus reconstruct the traditional dialectic of the realism-nominalism controversy in a way that fit with the positivist reconstruction of philosophical problems.10 But Bergmann also made the further point that one now had a cognitive basis for choosing one’s language, to wit, that one choose as one’s language that which enabled one to resolve the traditional philosophical problems. Bergmann thus showed how metaphysics could escape from the nihilism which the earlier positivists such as Carnap had preached.11 This does not mean that the positivist ban on metaphysics was completely ignored. The empiricist connection remained: the non-logical constants were to be interpreted into entities that were presented to one in ordinary experience. That experience might be perceptual, in which case the referents of the non-logical constants would be physical objects, their properties and relations; or the experience might be sensory, in which case the referents of the non-logical constants would be sensory elements, their properties and relations. In addition, experience might be taken to include inner experience, so that among the referents to the non-logical constants – and therefore among the building blocks of one’s ontology – would be mental states of various sorts, e.g. cognitive (believings, supposings, disbelievings, etc.), affective (passions, moral sentiments, etc.), and conative (willings, wantings, desires, etc.). But in all cases, the appeal was to experience. Entities that some philosophers introduced that purported to lie beyond the world of ordinary experience were excluded. Gods, Aristotelian entelechies, and so on, remained excluded. Analytic philosophy in this remained, even after the linguistic turn, empiricist. The language in question was, of course, as it had been for the positivists, a philosophically and therefore logically perspicuous or ideal language, though, to be sure, one into which statements of ordinary language – that is, those which made statements of (empirical) fact – could be trans-

687 lated. On this reconstruction of metaphysics, the question of ontology, or, in other words, the question of what exists, was reconstructed as the question of which entities are referred to by the non-logical constants of that ideal language which, on the one hand, can yield a translation in principle of all ordinary statements of fact, and, on the other hand, can solve or resolve or dis-solve all the traditional philosophical problems. It thus implied a difference between, one, ordinary statements of existence, represented by the existential quantifier, which are made in one’s language, and, two, ontological claims, which are made by speaking about one’s language.12 Thus, for example, the claim of Berkeley, that only ideas exist, or of Russell, that only sense data exist, can be reconciled with the Moorean point that these philosophers do not deny the existence of chairs by reconstructing their claims as philosophical claims about the nature of the ideal language.13 There were many philosophers who continued the pattern of Carnap, arguing that there are no philosophical problems with cognitive content. Such philosophers will hold, contrary to Bergmann and Goodman, that there is no philosophical sense of ‘exist’ in addition to the ordinary one represented by the existential quantifier. This is Quine’s official position, when he argues that insofar as there are philosophical problems of existence, they can all be handled by discourse about the existential quantifier. But there is also another Quine, who argues with Goodman for the cognitive viability of a nominalist language – though in the end on this Quine parts company with Goodman and grudgingly accepts Bergmann’s argument that the ideal language will have to contain at least two types of nonlogical constant.14 Nonetheless, the general thrust of the claims of those who follow Quine is that there is no philosophical sense of ‘exist’, and that the choice of a language is based on pragmatic and non-cognitive grounds. This includes Dummett, who has recognized that the arguments of Quine and Goodman concerning the nominalist language in fact commit them to the notion that there is a philosophical sense of ‘exist’ over and above the ordinary, and has explicitly rejected such arguments on the ground – the positivist ground – that there is no relevant philosophical sense of ‘exist.’15 There was another motive behind the linguistic turn. Wittgenstein had made the point fairly explicitly in the Tractatus: thought was to be “identified” with language. Wittgenstein in fact made two claims. First, he claimed (there is in fact no argument on this point in the Tractatus) that the relation between thoughts and what thoughts are about was the same as the relation between sentences and what sentences are about. As he put it,

688 It is clear, however, that ‘A believes that p’, ‘A has the thought that p’, and ‘A says that p’ are of the form ‘“p” says p’... (Tractatus , ¶. 5.542).

But second, Wittgenstein is also claiming here that the logical structure of language coincided with the structure of mental processes. In this sense, by studying the structures of clarified or ideal languages one could get at various features of mental processes. Bergmann took up the first point: a thought is expressed by its text, that is, the sentence (in the ideal language) that represents what the thought is about, or, as one has come to say, following Brentano, what the thought intends.16 One in fact distinguishes thoughts by their texts. The thought that intends the state of affairs represented by the sentence ‘p’ is said to mean the state of affairs p that is represented by the sentence ‘p’; and this thought is referred to by means of a definite description: “the thought that is expressed by ‘p’.” Bergmann represents this definite description by +p, and argued (taking ‘M’ to mean “means” or “intends”) that +p,Mp is a necessary (linguistic) truth. This has clear affinities to Wittgenstein’s ‘“p” says p’, but in fact it is rather different: for Bergmann, ‘+ p ,’ is not the name of a complex sentence as is Wittgenstein’s ‘“p”’; rather, follwoing G. E. Moore, Bergmann has ‘+ p ,’ refer to, or name, a character of mental acts. But of the latter, more later: it is part of a response to certain problems that differs from Dummett’s response to those problems, and points to structural connections that Dummett misses. For the present, the important point is that Bergmann did not hold that the logical structure of language represented actual thought processes: he was careful to make a clear distinction between logic on the one hand and the laws of psychology on the other. For others, however, the “identification” of thought with language became a means by which materialists could talk about thought while denying its existence: one could deny the mental by speaking about thought processes in terms of overt linguistic behaviour and dispositions to such behaviour. Dummett is such a philosopher. It is a major strength of Dummett’s book on (some of the) Origins of Analytical Philosophy that it demonstrates how some structural problems in Frege’s philosophy can lead one fairly directly to adopt Dummett’s materialistic account of language as a solution to those problems. But there

689 are other solutions that are possible. It is a weakness of his book that these are not canvassed. That they are not is due, I think, to his failure to be aware of his own nominalism – a nominalism that he shares with Frege and Husserl. But Dummett does clearly recognize that if one holds on to the empiricist strain in analytic philosophy, and rejects non-empirical entities, then one cannot adopt either Husserl’s or Frege’s ontological solution to the problem of thought; and further, that if that solution is not available, then his own sort of materialism is the only way out. In this respect, Dummett’s philosophy is at one with that of Wilfrid Sellars, though, to be sure, the latter is much more ontologically articulate. Dummett takes up two aspects of Frege’s thought. The first is what he calls the “context principle.” The other is Frege’s argument for the publicity of meaning. Frege uses the latter to argue for an ontological third world of senses, that is, meanings. Dummett’s empiricism leads him to reject this third world. He uses the context principle to defend an alternative account of meaning that secures the publicity of thought. But the third realm of senses or meanings provides a neat solution to a problem bequeathed to posterity by Brentano, namely, exactly what entities are the intentions of thoughts, including perceptions? This latter is a problem that arises from the fact that we can think what is not the case and can perceive what does not exist. This problem can be solved if one argues that senses or meanings are the intentions of mental acts, including those that are non-veridical. This latter was the route taken by Husserl, and as this line was pursued by him and his followers the result was contemporary phenomenology and the basis of “continental philosophy.” But the latter is saddled with the third realm, which violates the empiricist requirement that there be no entities in one's ontology that are not given to one in ordinary perception. Given the empiricist principle, then, one must take the other route towards the publicity of meaning, that of Dummett. Dummett further sketches how this alternative can address the problem raised by Brentano, including the problems of perception and non-veridical perception. This is the structural historical story that Dummett tells. As structural history, it is convincing – as far as it goes. There are other connections that Dummett’s story misses, however, and these should be brought out. At the same time, of course, Dummett’s structural story is more than history: Dummett is also arguing for the philosophical superiority of his solution to Frege’s problem of the publicity of thought. Here one can argue in re-

690 sponse that there are real philosophical difficulties. I propose to address both the structural historical points, bringing out some structural connections that Dummett misses, and the philosophical points, suggesting where there are philosophical problems with Dummett's solution to Frege’s problem of the publicity of thought. It will be best to begin with Frege’s argument for the publicity of thought. Dummett puts Frege’s argument this way: The subjective was for him [Frege] essentially private and incommunicable; he therefore held that the existence of whatever is common to all must be independent of any. On Frege’s view, thoughts and their constituent senses form a ‘third realm’ of timeless and immutable entities which do not depend for their existence on being grasped or expressed (p. 23).

This is what Dummett refers to as the “extrusion of thoughts from the mind.” The Fregean argument that he has in mind is this: The reference and sense of a sign are to be distinguished from the associated idea. If the reference of a sign is an object perceivable by the senses, my idea of it is an internal image .... The same sense is not always connected, even in the same man, with the same idea.... This constitutes an essential distinction between the idea and the sign's sense, which may be the common property of many and therefore is not a part or a mode of the individual mind.17

Frege’s argument begins from an important premise. He takes for granted not merely that images are private, but further that minds are individual objects, that images are modes of these objects, and that any such mode is inseparable from the object of which it is a mode: like the object, that is, the mind of which it is a mode, the mode is an individual or particular, something that the mind cannot share with any other object, and specifically cannot share with any other mind. One version of nominalism holds that the properties of things are as particular as the things of which they are properties. Such properties, understood as particulars, have been called “tropes” or “perfect particulars.” The point is that Frege’s argument presupposes not only the privacy of images but also the nominalistic thesis that the properties of objects are perfect particulars. In fact, Frege’s argument is that the images are private precisely because they are particular and therefore inseparable from the minds of which they are modes. Dummett’s characterization of the argument picks up the point concerning pri-

691 vacy, but it neglects the nominalistic premise. Nonetheless, Dummett does surely put it correctly that the conclusion of Frege’s argument is that there is a realm of entities – a third realm beyond the physical on the one side and the mental on the other – which he (Frege) calls senses. These senses are grasped by the mind when it thinks about things, and, therefore, when it understands a sentence. Husserl, too, introduces a third realm of ideal entities. Husserl, too, uses the entities of this third realm to account for the fact that thought is public. Dummett clearly lays out this common feature of the ontologies of Frege and Husserl.18 Dummett also correctly recognizes that Husserl uses these entities to solve a problem with which Frege was not directly concerned. This is the problem that Brentano bequeathed to his successors: what are the objects of thought when thought is false, when its referent does not exist. Thought, Brentano quite correctly pointed out, is intentional in the sense that it is about entities other than itself. When I consciously judge that Toronto is east of Vancouver, there is a certain state of consciousness. This is the judging. The judging is an act of the mind, where ‘act’ is contrasted to mere potentiality.19 This state of consciousness has a certain quality, the quality that distinguishes that state from one of imagining, or supposing, or hoping, or disliking, etc.20 We may call this the species of the act. But the act also has a feature by virtue of which it has as its object the state of affairs that Toronto is east of Vancouver. That object itself is of course not an ingredient in the act; it is not a real part of the act. The act has that state of affairs as its intention; the act is about that state of affairs. There is an ingredient in the act, a certain matter, which makes it about that state of affairs. If the same sort of matter is present in an act of another species of mine, e.g. a disliking rather than a judging, then that second act is also about the same state of affairs. Moreover, if there is an act of yours that has the same sort of matter, then that act too is about the same state of affairs. When Husserl makes these points – and quite correctly makes them, since they are perfectly commonsensical – we see him agreeing with Frege that there is an important sense in which thought is indeed public: thoughts are such that the same thought can be shared by two individuals, or by the same individual on different occasions. The problem of false belief is simply this: We can not only think of existing states of affairs, but also can think of states of affairs that do not exist. For example, some people who live south of the 49th parallel likely

692 think that Toronto is west of Vancouver. Their thought is false. In the case of the judgment that Toronto is east of Vancouver, there seems to be no problem: the act of consciousness is related to that state of affairs. In the case of the false thought, however, there is no (existing) state of affairs for the thought to be related to. In each case the thought seems to be related by the relation of intentionality to a state of affairs. In one case, however, the case where the thought is false, there is no relatum for the relation to relate the thought to. Intentionality is therefore no ordinary relation, no relation like, for example, kicking where there can be no kicking unless there is an actual exiting object there, there being kicked. The problem of false belief is thus this: What exactly is this very special relation of intentionality that relates thoughts to their objects? What is it about it that enables it to relate two things when one of those things need not exist? Husserl uses the third realm of entities to solve this problem. Some have appealed directly to entities in the third realm to solve the problem of false belief.21 As Moore once did, we may refer to the entities in the third realm as “propositions.”22 A proposition is true if there is a fact which corresponds to it, and false if no such fact exists. As for thoughts, these have propositions for their direct objects. Thoughts are about facts or states of affairs in the world by virtue of intending propositions which are in turn about the states of affairs in the world. But this solution clearly will not do. In the first place, it does not seem to solve the problem. For, a false belief does not involve a cognitive attitude towards something that is (a proposition) but rather involves an attitude towards something that is not.23 In the second place, the problem of false belief is not really solved, but merely displaced. It becomes the problem of stating what the relation is between a proposition and the state of affairs that it is about.24 And in the third place, it puts a veil between the thinker and the objects he or she is supposedly thinking about. For, what we think about are the states of affairs in the world and not the propositions that, upon this view, lie between our thoughts and what we are thinking about. In fact, this view of propositions in effect re-creates the traditional doctrine of representative realism. The only difference is that the representing entities, the entities of the third realm, are not Locke’s subjective ideas but non-mental entities which can provide for the publicity of thought. The view thus solves the problem with which Frege was concerned, but as a way of solving the problem of false belief, it is simply a failure: not only does it not solve it but it re-introduces all the problems – all the insoluble problems – of representative realism.25 Dummett suggests that Husserl introduces the entities of the third

693 realm in order to solve the problem of intentionality and of false belief, the problem bequeathed by Brentano to his successors. Dummett, in other words, construes Husserl as holding to something like the propositional view that we have just noted. But this is not entirely accurate. To be sure, Husserl does appeal to the third realm to solve Brentano’s problem of intentionality, that is, the problem of false belief.26 But this is not the primary task for the entities of the third realm to perform. In fact, in the Logical Investigations, the third realm is introduced much before the problem of intentionality is raised: it is introduced in the first place in order to solve a quite different, and more basic, problem.27 Nor does Husserl introduce the third realm in order to solve the problem of the publicity of thought, as does Frege. It is one of the shortcomings of Dummett’s history that he does not clearly bring this out, nor recognize clearly the central problem that Husserl aims to solve with his third realm. Husserl distinguishes three things with regard to a red object. There is, in the first place, the red object. Then, second, there is the aspect of red which appears before us. This aspect is as particular as the object itself: the red aspect of this red object is different from the red aspect of that other red object. These aspects are thus, ontologically, what we previously called tropes or perfect particulars. Husserl distinguishes, third, the Species.28 Where some philosophers argue that one can account for the fact that two perfect particulars are both reds in terms of their exact similarity, Husserl argues that there is a third entity, the Species under which the two aspects are subsumed.29 This Species is an “ideal unity”, a Form that exists outside the spatio-temporal world of the object and its aspects.30 It exists in the third realm, neither mental nor physical. In other words, Husserl introduces the entities of Frege’s third realm in order to solve the problem of sameness. This is the problem why the properties or aspects of things are characterized as the same. The answer is that they are the same by virtue of falling under the same form or ideal species, falling under the same entity in the third realm. Husserl shares with Frege the assumption that the properties of things are tropes or perfect particulars. But Husserl recognizes, where Frege is not so clear, that this poses the problem of sameness. It is at this point, to solve this problem rather than of the publicity of thought, that Husserl argues for the existence of a third realm. It should clearly be noted that Husserl’s problem of sameness arises only given the assumption that the properties or aspects of things are tropes or perfect particulars. Suppose, to the contrary, that one holds that the

694 properties of things are not particulars. That would be to hold, for example, that the red in this thing is literally the same as the red in that other thing. The two things are particulars, but the property red that is an aspect of each, does not as it were divide into a multiplicity of reds: it simply is one and the same entity in the two things. In that case the problem of sameness which Husserl addresses simply does not arise: the properties in things are already universals, and there is no need to introduce entities in a third realm to account for the sameness.31 But Husserl does in fact share with Frege the assumption that the properties of things are as particular as the things of which they are properties. Dummett does not clearly identify this point of agreement. In any case, Husserl’s third realm solves a number of additional problems – always an argument with some force when it comes to defending an ontology: solving several problems with the same sort of entity establishes that the introduction of such an entity is not merely ad hoc. In particular, the entities of the third realm provide Husserl with an account of necessary truth: necessary truths are eternal truths, and eternal truths are truths about the (ideal) relations that hold among the (ideal) forms.32 Such relational truths about essences or forms are not purely tautological, purely formal and independent of all content. They are, rather, synthetic. Since they are also eternal truths, they must be counted as “synthetic a priori”. Now, among the things of the world are mental states and their objects. Husserl is concerned about the case where the object does not exist, that is, with what we have called the problem of false belief. The mental state of thinking is a particular. But it has a certain species: it is a believing, for example. The quality or aspect of the act that is the species is itself a perfect particular or trope. This particularized quality falls under a certain form or essence located in the third realm. The act will also contain a particularized quality by virtue of which it intends one object rather than another. This particularlized quality also falls under a certain form or essence.33 The object which the act intends will also consist of particularized qualities, perfect particulars. These particulars that constitute the object will in turn fall under certain forms or species. The essence of the act will be a certain complex essence, as will the essence of the object.34 Let us call these the act form and the object form, respectively. Husserl now quite reasonably takes intentionality to be a relation that connects an act form to an object form. Intentionality is therefore like an ordinary relation: it always has relata. Indeed, necessarily so, since the fact consisting of the intentional relation relating certain forms is an eternal

695 truth. Thus, the problem of intentionality being a very special relation that relates existing to non-existing things is neatly avoided. As for truth and falsity, that will depend of course upon whether the object exists or not. If it does exist, the object form will have an entity falling under it; if the object does not exist, the object form will have no entity falling under it. The crucial point is that the object form will exist (in the third realm) even if it has no object falling under it. Thought can thus have a non-existent object. Moreover, the third realm does not stand between the knowing mind and the object known, as in the case of the proposition account we sketched above. The task of the forms is to establish a necessary tie between the mind and its object; it is by virtue of this tie that the object is presented to the mind. The entities of the third realm serve to present an object, not to re-present it. In general we are not directly presented with the objects which we think: the object is thought but is not present to us. But we can directly confront these particulars in intuition, in certain cases at least. In such a case the intention is said by Husserl to be “fulfilled.”35 The truth of a belief or judgment is discovered by directly confronting the particulars falling under the object form; “knowledge,” as Husserl puts it, “refers to a relationship between acts of thought and fulfilling intuitions.”36 But however Husserl arrives at the third realm, he does also insist that it fulfills the same role that Frege assigns to its entities, as the immediate, and public, objects of thought, and Dummett objects to it on essentially empiricist grounds: the entities are simply not given to us in experience. Dummett puts the point this way, that if there are such entities, then it is possible to give an account of thought that is independent of language. Both Frege and Husserl tend to approach thoughts – the entities of the third realm – by way of language. Frege insists that it is senses that refer and not linguistic expressions, but in practice he begins with expressions and their references, and proceeds to argue that the expression has such and such a sense. In practice, the distinguishing of senses as parts of thoughts is parasitic upon the recognition of the structure of the sentence expressing the thought. This was in fact one of the central objections by Russell to Frege’s account of sense and reference in his well known paper “On Denoting.”37 Dummett rather uncharitably remarks that Russell’s arguments against Frege in this paper “are obscure and contorted,” and that “they have resisted attempts to elucidate them or render them cogent” (p. 81). In fact, however, Herbert Hochberg, with whose work Dummett seems unfortu-

696 nately to be unfamiliar, has shown in a brilliant paper that a careful reading of Russell reveals his arguments to be fairly clear, certainly cogent, and quite successful.38 Russell considers the sense or meaning and the denotation of a phrase. Let the phrase be ‘C’, and let us refer to its meaning as “MC” and its denotation as “DC”. Thus, we have (i) ‘C’ denotes DC and (ii) ‘C’ means MC Russell is concerned with the relation (iii) R(MC, DC) If it is held that (iii) is to be understood as the conjunction of (i) and (ii), then the relation between the sense and the referent is merely linguistic. But this will not do: if meanings really are independent, in a third realm, then there must be a logical relation by means of which sense determines its denotation which is independent of us and of our contingent language. As Russell puts it, ...the relation of meaning and denotation is not merely linguistic through the phrase: there must be a logical relation involved, which we express by saying that the meaning denotes the denotation.39

But, Russell proceeds to argue, in general if we keep a logical connection, then it turns out that meaning and denotation in crucial cases collapse into one another; and further that in general the meaning connection to a referent cannot be explained save by means of denoting expressions, that is, linguistic expressions. Again as Russell puts it, ...the difficulty which confronts us is that we cannot succeed in both preserving the connexion of meaning and denotation and preventing them from being one and the same; also that the meaning cannot be got at except by means of denoting phrases.40

So Russell is arguing the same point that Dummett argues, that Frege has no way of getting at senses save via language. Nowhere, then, does Frege give us a way to deal with thoughts in their own right. Frege’s theory, of meanings or senses in a third realm, conforms to the first axiom or principle that Dummett takes to define analytic philosophy – that “a philosophical account of thought can be attained through a philosophical account of language” – when it undertakes to study these objective thoughts by means of language. However, the Fregean philosophy

697 does not conform to the second axiom – that “a comprehensive account can only be so attained” – since it holds that in principle at least thoughts are independent of language. However, Frege’s practice of never analyzing thoughts save through language points elsewhere: it clearly implies that we can obtain a comprehensive account of thought only through a comprehensive account of language. Thus, according to Frege’s practice “we have no account of thoughts save by reference to a language” (Dummett, p. 11), and both of the axioms that Dummett has taken to be definitory of analytic philosophy are satisfied. For Dummett, the significance of Frege’s work does not lie in its argument for the third realm, but rather in its insisting that thoughts be public. The importance of the denial of the mental character of thoughts, common to Bolzano, Frege, Meinong and Husserl, did not lie in the philosophical mythology to which it gave rise – Frege’s myth of the ‘third realm’, or Husserl’s of ‘ideal being’. It lay, rather, in the non-psychological direction given to the analysis of concepts and propositions (p. 25).

For both Frege and Husserl there is the problem of the linkage between thoughts considered as entities in the third realm, and language.41 Now, ...viewed through the imagery of the third realm, [this problem of linkage] appears acute. But when we come down to earth, and consider the expression of thoughts in language, it evaporates. Frege believed that the only access we human beings have to thoughts is through their verbal expression; so the question how we grasp thoughts resolves into the question how we understand sentences (p. 63).

If we thus simply identify the grasping of thoughts with the using of linguistic expressions, “The mystery has vanished: indeed, the third realm has vanished, too” (p. 63). Moreover, so long as we insist that language is essentially public, we continue to have what Frege so strongly insisted upon, namely, the publicity of thought. One can achieve the objectivity of thoughts, in a way that fulfills both of Dummett’s axioms for analytic philosophy, by construing thought as essentially linguistic. The accessibility of thoughts will ... reside in their capacity for linguistic expression, and their objectivity and independence from inner mental processes in the common practice of speaking the language, governed by agreement among the linguistic community on the standards of correct use and on criteria for the

698 use of sentences (p. 25).

One’s “grasp of the sense [of an expression] is ... an ability” (p. 104), that is, a linguistic ability, the capacity to use an expression correctly. The grasp of a sense is thus transformed from somehow standing in a relation to an entity in the third realm to the ability to use a word according to intersubjectively accepted rules. This hardly makes the account of thought “non-psychological” as Dummett claims. To be sure, thoughts become public rather than private. But to suppose that the psychological is limited to the private is to continue to exist with Frege and Husserl in the context of 19th century psychology. But psychology has changed; it is now an objective science. How it came to change from the science of the private to an objective science of human behaviour is a long story.42 But change it did, and its subject matter is now human behaviour broadly understood, and in particular including linguistic behaviour and communication. So Dummett’s proposal that thoughts be construed as linguistic does not make them any the less psychological, though it does secure their publicity.43 Wilfrid Sellars for many years argued the same view with regard to thoughts and the mental. Sellars defines meaning in terms of linguistic role.44 Thus, ‘rot’ (in German) means .red. where ‘ .red. ’ is a common noun which denotes any linguistic item in a language that plays the role in that language that ‘red’ plays in English. As for the role, this is the set of regularities and dispositions that relate the word in question to (a) the world, (b) other words, and (c) actions.45 The regularities by which things and features of things in the world introduce words into discourses are “language entry” or “world-word” transitions; the regularities that pattern the introduction of further words into discourse upon the appearance of a given word are “intra-linguistic” or “word-word” transitions; the regularities that connect words to actions, including such actions as asserting, denying, supposing, etc., are “language exit” or “word-world” transitions. These regularities and dispositions are not merely patterns but are, as Sellars says, “rule governed.” That is, the patterns are brought about by rules: normative rule expressions play a special role in maintaining the patterns within a linguistic community. The structures here are complex; at one point Sellars describes them this way: the role which a linguistic item plays is defined ...in terms of the uniformities and propensities which connect the utterance with

699 (1) other utterances (at the same or different levels of language), (2) with the perceptible environment, and (3) with courses of action (including linguistic behavior).

However, he goes on, the uniformities to which one appeals in defining a linguistic role ...are not mere uniformities, for they are grounded in rules in a way most difficult to analyze, but which involves the causal efficacy of rule expressions.46

Like Sellars, Dummett takes the notion of “same meaning” to be identifiable with the notion of inter-translatability: ...the true unit for a fully adequate description of linguistic practice would not even be a single language....It would be a maximal set of languages connected by the existence of a standard translations between them (Dummett, p. 153).

In fact, as it turns out, Dummett’s account of meaning is largely of a piece with that of Sellars. Sellars restricts the notion of “cognitive meaning” to those aspects of the linguistic role that are given by the “word-word” transitions47; as Sellars put it, “the specific nature of a factual concept is determined by the material rules of inference governing it, as its generic nature is determined by formal rules of inference” (“Inference and Meaning,” p. 317). In fact, according to Sellars’ account of meaning, even so-called observation predicates have their sense, their cognitive meaning, given by word-word connections, even when their reference or, as Sellars says, their application is otherwise determined. ...that at least some of the descriptive predicates of a language must be learned responses to extra-linguistic objects in order for the language to be applied, is obvious. But that not even these predicates (“observation predicates”) owe their conceptual meaning to this association should be reasonably clear... (ib., p. 334).

The cognitive meaning is thus given not only by the rules of syntax that determine how words go together to form sentences but also by the statements of general fact in which the words appear. So, on Sellars’ account of meaning, ...material transformation rules determine the descriptive meaning of the ex-

700 pressions of a language within the framework established by its logical transformation rules (ib., p. 336).

In terms that Carnap once made familiar, the cognitive meaning of a word or concept is given not only by the L-rules but also by the P-rules, the statements of physical or natural law in which those words occur.48 The cognitive meaning is thus given by the statements of general fact that "implicitly" define it.49 Sellars characterizes the world-word transitions in terms of entities in the world evoking words the configurations of which picture those entities.50 But these picturing relations, while contributing to the meaning of the term in the sense of linguistic role, do not contribute to the cognitive or conceptual meaning of the term – that, to repeat, is given by the word-word connections. In the context of Dummett’s account of language, the world-word transitions correspond to the relation of reference and the word-word transitions which define the cognitive meaning of the word correspond to its sense. Dummett is in fact explicit that the sense of an expression is given by its relations to other words. This is, of course, his famous context principle, adopted for his own purposes from Frege’s dictum that only in the context of a sentence does a word have meaning. As Dummett re-phrases Frege’s principle, thoughts now become public patterns of usage: ...the senses of subsentential expressions are determined by the role they play in the context of the sentence as a whole. Because the referent of a sentence is a truth-value, it follows that the sense of a subsentential expression is identifiable as the contribution this expression makes to determining the truth-value of the sentence in which it occurs (p. 5).

One of the strengths, according to Sellars, of his view of meaning is that it enables one to give a comprehensive account of abstract entities without having to introduce any new sorts of entity over and above the ordinary things of the ordinary world: there is no need for a third realm, as with Frege’s Senses or Husserl’s Species, or a fourth realm as in the case of such Fregean entities as numbers or the twin monsters the True, which is the virtuous twin, and the False, which is the evil twin. Dummett makes much the same point. Frege held that there are indeed such entities as numbers, and truth values, quite outside the realm of ordinary things. We have no experience of such entities, no idea or intuition of them as we have ideas and intuitions

701 of ordinary physical things. This forces upon Frege the question, as Dummett puts it, “how are numbers given to us, granted that we have no idea or intuition of them?” (p. 5). The Fregean answer to this question is that “numbers are given to us through our grasping whole thoughts concerning them” (ib.). Once the turn is made, as Dummett makes it, to construing thoughts as the use of linguistic expressions, it follows that grasping whole thoughts concerning numbers is correctly using sentences containing number expressions. The senses of number expressions are given by the sets of axioms or rules that implicitly define them.51 As Dummett has put it elsewhere: ...it seems that we ought to interpose between the Platonist and the constructivist [or “subjective idealist”] picture an intermediate picture, say of objects springing into being in response to our probing. We do not make the objects but must accept them as we find them (this corresponds to the proof imposing itself on us); but they were not already there for our statements to be true or false of before we carried out the investigations which brought them into being.52

“What is truth?” asked Pilate. Dummett’s Fregean question is: how are truth and falsity given to us, granted that we have no idea or intuition of them? Dummett's answer is that they are given to us through our grasping whole thoughts concerning them. And, given Dummett’s construal of the grasping of thoughts as the use of linguistic expressions it follows that grasping whole thoughts about the True and the False consists in correctly using sentences containing these truth-value expressions. The sense of the terms ‘true’ and ‘false’ are given by the set of axioms or rules that implicitly define them. These terms are of course related to thoughts themselves; it is sentences considered as thoughts that serve to introduce the terms ‘true’ and ‘false’ into discourse. But truth and falsity are in turn connected to assertion and denial. This connection is not made by Frege: What Frege fails to stress as heavily as that sense is bound up with truth is that the concept of truth is in its turn bound up with that of assertion (p. 12).

It is nonetheless essential. Truth is implicitly defined, then, by its connection to sentences on the one hand, and by its connection to assertion on the other. These connections are, of course, normative for the use of the relevant expressions; the connections are rule governed. Thus, the truth consists of those sentences that are assertible in the sense of being WORTHY OF ASSERTION. And a sentence will be counted as true just in case that it con-

702 forms to the rules that determine what is worthy of assertion: truth consists in conformity to a rule. Dummett is not the first to make truth be the conformity to a rule. Sellars has done much the same. He argues that the statement The colour white (whiteness) is a property ought to be construed as The (or a) .white. is a predicate while a exemplifies whiteness ought to be construed as The (or a) .white. is true of a Finally, the latter holds just in case that . a is white. is true where this is to be understood as saying that . a is white.s are semantically assertible53 Truth is thus defined to be what is semantically assertible (“S-assertible”), and, more specifically, ...S-assertible by us. For truth in the ‘absolute’ sense is, in its own way, language relative, relative to our language.54

This in turn is to say that truth is relative to our conceptual structure: truth means “S-assertible in our conceptual structure.” ...the fundamental form of ‘true’ is true in conceptual structure CSi. The ‘unqualified’ sense of ‘true’ pertains to the special case where CSi is our conceptual structure....55

The positions of Dummett and Sellars are both redolent of neoKantian views. (Sellars at least is certainly aware of this connection.) Brentano once quoted Windelband to this effect, that truth is conformity to a rule. According to Brentano’s Windelband, What [Kant] did...was to recast the concept of truth. According to him truth is what corresponds with the norm in our mind; not what corresponds with the object....Moreover, according to Kant, truth is not restricted to judgements or to thought; it may be found equally well in all the other areas of mental activity, in volition and in feeling, provided only that they conform to certain norms or rules.56

This points to an important difference between the views of Sellars and

703 Dummett. If it is indeed the case that truth consists in conformity to appropriate rules in our conceptual scheme, then there may well be different conceptual schemes, thus allowing for different bodies of truth. Sellars is careful to limit his conceptual scheme to language entry transitions where the language patterns that we have learned are such that when we confront the world perceptually, the world we perceive brings it about that we produce instances of sentences that picture that world. Truth, as we have seen, is not a relation. Picturing, on the other hand, is a relation, indeed a relation between two relational structures. And pictures, like maps, can be more or less adequate. The adequacy concerns the ‘method of projection’. A picture (candidate) subject to the rules of a given method of projection (conceptual framework), which is a correct picture (successful candidate), is S-assertible with respect to that method of projection.57

But there seems to be no reason, if truth is not a relation and is rather conformity to a rule, why there should not be other conceptual schemes, other languages, in which language entry transitions of other sorts occur. Why, for example, could we not have facts, truths, about entities not of this world, provided only that we had a set of rules that licensed their assertibility? Why could we not licence the assertibility of statements of fact provided that they were taught by, say, the Church? In fact, Dummett does exactly this. He proposes that we are “able to place entire confidence in the Church's solemn pronouncements on the content of our faith.”58 This licences us to believe not only in such ordinary facts as that Jesus was a man but also in such extra-ordinary facts as that Jesus was also God and that God is Trinity. It is perhaps ironic that Dummett on the one hand has empiricist objections to the mysterious third realm of Frege and Husserl but is, on the other hand, prepared to accept as facts such mysteries as Jesus being both wholly human and wholly God and of God being Trinity. Locke’s empiricist scruples, in contrast, led him to deny the Triune nature of God and to adopt an Arian view of Jesus. How are we to decide between the language of Sellars that allows for only one kind of language entry transition and the language of Dummett that allows for at least two kinds? So long as truth is merely relative to rules, it is hard to see how we can do anything more than leave the two parties to their different language games. This is not a refutation of the view, but it does strike me as something of a reductio – much as Socrates attempted to produce a similar reductio of Protagoras’ relativism. But there is another objection to the view that truth is a matter of as-

704 sertibility according to rule. This objection was presented by Brentano in his remarks on Windelband. This reply applies equally well, it seems to me, to the views of Sellars and Dummett. Brentano puts it this way: If truth were no more than judging according to rule, then every judgement which is made on insufficient grounds or which is completely blind would have to be erroneous. But this is certainly not the case. Insight ... must always be true; but a frivolously made assumption, a mere prejudice, or a view adopted by mere appeal to authority or because it is fashionable, may turn out to be true or may turn out to be erroneous.59

This does seem to me to be decisive. But be that as it may there are other consequences of the view of truth provided by Dummett and Sellars. On this view, as we have seen, a sentence will be counted as true just in case that it conforms to the rules that determine what is worthy of assertion: truth consists in conformity to a rule. It is important to recognize that this view makes reference to something very different from pointing to an entity. In deciding what should be reckoned as the reference of an expression, Frege had a precise question to ask: what contribution is made to determining the truth-value of any sentence in which it occurred, where that contribution must be something had in common with any expression whose substitution for it would in all cases leave the truth value unaffected (p. 55).

The contribution of an expression to determining the truth value of a sentence in which it occurs will be the contribution to determining whether that sentence is worthy of assertion. Unless what makes a sentence worthy of assertion in fact requires pointing, an expression will be said to refer so long as it makes a contribution to the sentence being worthy of assertion. We can thus have number expressions construed as referring expressions without having to take them, as Frege did, as referring in the sense of pointing at certain abstract non-physical entities. Similarly for the True and the False: these, too, can be construed as referring expressions without having to take them, as Frege did these also, as referring in the sense of pointing at certain abstract, non-physical entities. Thus, upon Dummett’s view, such terms as number expressions or the terms expressing the two truth values can be held to be referring expressions without having to hold that there are any entities outside the world of ordinary experience. Many would of course argue that, upon this view, the abstract entities do not exist – even though, within the rules or axioms of the language,

705 it can be said that there are numbers or that there are truth values. To hold that one can say both these things requires one to distinguish a philosophical and an ordinary sense of ‘exist’. A philosopher such as Wilfrid Sellars would make such a distinction, and use as an argument in favour of the position he and Dummett both defend that it does provide us with an ontology in which abstract entities such as numbers and truth values do not exist.60 But Dummett, as we have noted, rejects the notion that there is a philosophical sense of ‘exist’ beyond the ordinary sense, and so he can reject any argument based on his supposed denial of the existence of abstract entities. As a third case of abstract entities, we should notice Husserl’s Ideal forms. On Dummett’s view, Husserl creates the third realm in order to ensure that general concepts such as ‘horse’ have a single sense or meaning even though they refer to many individual things. Husserl, we are told, holds that “an expression may have the same meaning, but different references, in different contexts.” This of course is true of indexical expressions. Husserl, however, applies it not only to these expressions “but ..., appealing to a far more primitive conception of reference than Frege’s, applies it to a general term like ‘a horse,’ considered as predicated of different individual animals” (Dummett, p. 52). This is hardly fair to Husserl. The latter does not begin from a “primitive notion of reference” but instead offers an argument that his Ideal forms must exist; and, having so argued, he then quite reasonably holds that species terms refer primarily to these entities and derivatively to the entities that fall under these Ideal forms. But be that as it may, Dummett does attribute this view concerning the primitive notion of reference to Husserl, and holds, not unreasonably, that it is wrong. According to Dummett, the term ‘horse’ indeed has a sense; that sense is given by the axioms that implicitly define it. There is no need to assume a third realm wherein exists a special Ideal form to which the term refers. Rather, it refers simply to individual horses. Here the world-word transitions are fairly direct: to put it a bit simplistically, an individual horse introduces the term ‘horse’ into discourse. Dummett does not lay out the world-word connections for ‘horse’ but he does address the issue of perceptual language. He does this in particular with reference to colour terms. Colour terms involve, in the first instance, a “propensity to apply colour-names to surfaces of bright uniform hue” (p. 91). This is how we first learn to use them. As Sellars put the same point, “at least some of the descriptive predicates of a language must be learned responses to extra-linguistic objects in order for the language to be ap-

706 plied” (“Inference and Meaning,” p. 334), that is, in order for the language to have referents. We then learn to extend these terms to other surfaces, with the propensities regarding the use of the terms becoming gradually more complex. The same holds, one presumes, for other perceptual terms, including terms such as ‘horse’. In any such case, however, what constrains the reference is the sense, that is, the set of word-word connections into which the term enters. Dummett raises an objection to Husserl’s introduction of Ideal forms. To grasp, he tells us, “the concept of a type, we have first to know what any of its tokens is, and then to understand what equivalence relation makes them tokens of that type” (p. 48). This suggests that what makes two tropes or perfect particulars to be of the same kind is a relation of similarity or resemblance that holds between them. It seems that Husserl goes wrong by trying to account for resemblance in terms of falling under a single Ideal form, where in fact what accounts for the reference of a single term is the relation of resemblance, the relation that makes the two tropes or perfect particulars equivalent.61 It would presumably be what we noticed above, Husserl’s “primitive conception of reference” (p. 52), that misleads him into substituting the single concept, the single Ideal form, for the multiple reference determined by the resemblance relation. Dummett hardly does justice to Husserl in this discussion. Husserl offers a sustained argument against the resemblance account of why we attach the same term to several perfect particulars. He not only offers his own argument based on the phenomenology of judgments of sameness, he also appeals to the arguments of John Stuart Mill62 against the resemblance account usually adopted by empiricists such as Spencer following e.g. Berkeley and Hume, and defended by Spencer.63 Dummett seems to be as unfamiliar with this traditional dialectic as he is with Husserl’s arguments on the topic. But Dummett is moved more by his own account of sense than he is by the traditional arguments and the traditional dialectic. What constrains reference is sense and sense is a matter of word-word connections. Husserl takes reference to particular things to be constrained by the Ideal form under which things fall: if a term is associated with a form then that association constrains it to refer to the things that fall under it. Berkeley, Hume and Spencer take reference to a particular this to be constrained by the resemblance relation that connects the several things: if a term is associated with a resemblance relation then that association constrains it to refer to things that so resemble. But for Dummett it is not the entities that constrain

707 reference: to the contrary, it is the sense of an expression that constrains reference, the word-word relations. For Dummett, then, it is linguistic connections that determine the extension of a term, its multiple reference. The point can be put another way. Dummett seems to locate himself on the side of Berkeley, Hume, and Spencer and against Husserl when he states that to “grasp the concept of a type, we have first to know what any one of its tokens is, and then to understand what equivalence relation makes them tokens of that type” (p. 48). He seems to be saying that the basis of any predication of particular things is a relation of resemblance. However, for Berkeley, Hume and Spencer, the basic relations of resemblance are all primitive; they cannot be any further analyzed. They can only be designated, or as it were pointed to, but they cannot be explained. Dummett, however, seeks such an explanation: Husserl, he complains, is “far more anxious to establish ... his theory..., than to explain” what it is for a particular to be a token of a type, while “of the nature of the similarity relation that constitutes [instances of a type] as belonging to the same species we receive even less enlightenment” (p. 48). Since Dummett is seeking a linguistic explanation of what upon the view of Berkeley, Hume and Spencer cannot be given a linguistic explanation, it follows that he (Dummett) is rejecting the account of relations of resemblance given by Berkeley, Hume and Spencer. Dummett is seeking, rather, a linguistic explanation of sameness, a linguistic account of the similarity of particulars, an account in terms of sense (word-word connections) for the reference of kind terms to different individuals. Dummett has expressed this radical nominalism elsewhere as follows: What objects we recognize the world as containing depends upon the structure of our language. Our ability to discriminate, within reality, objects of any particular kind results from our having learned to use expressions, names, or general terms, with which are associated a criterion of identity which yields segments of reality just that shape: we can, in principle, conceive of a language containing names and general terms with which significantly different criteria of identity were associated, and the speakers of such a language would view the world as falling apart into discrete objects in a different way from ourselves...for Frege, the world does not come to us articulated in any way; it is we who, by the use of our language (or by grasping the thoughts expressed in that language), impose a structure on it.64

Or again, The picture of reality as an amorphous lump, not yet articulated into discrete

708 objects, thus proves to be a correct one, so long as we make the right use of it ....Such a picture corrects the naïve conception ... [which] presupposes that the world presents itself to us already dissected into discrete objects, which we know how to recognize when we encounter them again, in advance of our acquiring any grasp of language at all.65

Thus, the samenesses and differences of things are conferred on things by language. It is this that distinguishes Dummett from Berkeley, Hume and Spencer. For them, it is resemblance that determines the correct range of application of a term; resemblance is prior to language. For Dummett, on the contrary, language determines the reference of terms and thereby determines resemblance; language is prior to resemblance: prior to language, there are no oysters or cows, no clouds or shadows, no structure objectively there, no prior fact of any matter, all one “amorphous lump.” Fraser Cowley rightly asks, he seems to presume rhetorically, “Does he really mean what he seems to mean? Does he really believe it?”66 But Dummett does seem to believe it.67 What we have then, in Dummett's ontology, is a radical nominalism, one in which there are only particulars, and in which all sameness is a matter of language. To be sure, Dummett would not so express it. To put it this way is to insist upon what Dummett denies, that besides the ordinary use of ‘exist’ there is a philosophical use of ‘exist’. For Dummett would insist, quite reasonably from his point of view, that he allows for ordinary usage to say that there are colours, there are species, etc., and to insist, therefore, that these things all exist and are all referred to. It is only by importing a second, philosophical, sense of ‘exist’ that Dummett can be taken as denying that there are any objective grounds in the real resemblances of things for the correct application of kind terms. From his own point of view, Dummett is not a nominalist. But for all that, he really is one, in a very traditional way. Once we attempt to read him as doing ontology by talking about a certain philosophically perspicuous language, even though he does not so describe his enterprise, it is clear that he is committed to a radical materialistic nominalism. And this radically nominalisitc position is one that is, quite clearly, untenable. Dummett is not the first philosopher to come out of Oxford who has defended such a nominalism -- David Pears did so many years ago68 – but as Butchvarov has conclusively shown, the position will not withstand philosophical scrutiny.69 Not even Sellars takes so strong a nominalistic position (though he does come close – he takes the characteristics of things to be tropes or per-

709 fect particulars which turn out to be pretty close to being blobs.70). Although Sellars insists, as we have noted, that abstract entities, including species considered as single entities, are of a linguistic nature, he does not go so far as to eliminate all objective, non-linguistic grounds as underlying the correct application of kind terms to individual things. Sellars elaborates upon the nature of the world-word connections, unlike Dummett, who does little more than label such connections as one kind of reference. For Sellars, as we have previously noted, the world-word connections involve a relation of picturing.71 Holding the world to consist of tropes or perfect particulars and of complex individuals constructed as it were out of perfect particulars, he also holds that the perfect particulars (sort of) objectively, in themselves, fall into certain kinds. There are the reds, the blues, the b-flats, etc. What Dummett refers to as our “propensity to apply colour-names” (Dummett, p. 91) to particulars is correctly exercised, on Sellars’ view, only if the colour names are applied to just those particulars that are (sort of) objectively the same colour. The picture of the world that the exercise of the propensity produces must correctly represent the way colours are (sort of) objectively distributed in the world. Notice, however, that even though Dummett’s nominalism seems to be stronger than that of Sellars and certainly stronger than that of Frege or Husserl, he shares with all these philosophers the principle that all the entities in the world are particular.72 As we shall see, this shared nominalism leads him to overlook a very different response to the problem of the objectivity of thought. The problem for Dummett is the problem that Frege had, how to secure the objectivity of thought. But for Dummett there is a further constraint: it must be done without appealing to an imaginary third realm of entities. The move that Dummett takes to be crucial is that of making thoughts essentially linguistic. A different solution to the problem is possible, however. It is one that Dummett overlooks, I shall suggest, because of the nominalism that he shares with Frege, Husserl, and Sellars. What Dummett finds puzzling is the step from the claim that thoughts are objective to the claim that the grasp of thoughts, the grasp of senses, is something that is essentially linguistic. Given the initial step taken by Bolzano, and followed by Frege, Meinong and Husserl, whereby thoughts were removed from the inner world of mental experience, the second step, of regarding them, not merely as transmitted, but as generated by language, was virtually inevitable: it is puzzling only why it took so long (pp. 25-26).

710 This is misleading. The move had already been made many years before by Wilfrid Sellars. But Dummett seems to ignore Sellars’ work. Nonetheless, Dummett is not wrong: it took some years before people ventured to try to solve the problem of the objectivity of thought by appeal to a nominalistic, materialistic account of language. Part of the reason for this delay, Dummett argues (p. 13), is the fact that Frege took no notice of the notion of the force of an expression. Frege’s theory of linguistic meaning, his account of sense and reference, makes no appeal to the social character of language. But this is needed if we are to account for the force of an expression, for example, its being interrogative, assertive, etc. We may well wonder about this argument. After all, Brentano and Husserl both acknowledged that mental acts come in species, cognitive, affective and conative species, and among the cognitive Husserl at least was quite prepared to admit such things as questionings, supposings, assertings, and denyings. Moreover, it seems clear to most that Frege, too, admitted such attitudes; he has both the assertion stroke and the supposition stroke as part of his notation, and relates these explicitly to the same cognitive attitudes that Husserl considered. Nor is there any reason to suppose that these attitudes are somehow essentially involved in the public expression of thoughts. Nonetheless, Frege did not develop his account of assertion. Having insisted on a sharp distinction between sense and force – between the thought expressed by a sentence and the force attached to an utterance of it – Frege was content to leave it at that, without attempting any more detailed account of any particular type of force (Dummett, p. 13).

And this was a mistake that held back the move to Dummett’s account of thought. For, “force can hardly be conceived to exist save as attached to sentences uttered in the course of linguistic interchange” (p. 13). Again, this would seem to be mistaken. Why should we not hold, with Brentano and Husserl, and, indeed, Frege, that assertion, supposition, wishing, etc., are species of mental act that are often expressed by the force of the expressions we utter in linguistic acts – that is, in such linguistic moods as the assertoric, the subjunctive, and the optative. Since such mental acts often lie behind, and explain, our speech acts in linguistic interchanges, Dummett is surely correct in holding that a full account of speech and thought can hardly ignore such things. But whether considering the force of an expression, the mood of the verb, will actually force one to consider the intersubjective aspects of language is dubious.

711 Of course, for Dummett all thought is objective: there is nothing of the private about it, as there is for Brentano and Husserl. And once one thinks of thought as wholly objective then the notion that the mood of a verb expresses a species of mental act makes little sense. At the least a tale must be told – that Frege does not tell – about how such acts connect up with objective thoughts. One would guess that we would have to have different sorts of relation – assertoric, imperatival, etc. – to the objective thoughts or senses. But even so, such relations would have to have one foot as it were in the private, which would go against the notion that all thought, including such attitudes, is objective. So to this extent Dummett makes good sense: had Frege thought about the moods of verbs, the force of an expression, then the move towards a position on thoughts akin to that of Dummett might well have come sooner. But Dummett is mistaken in holding that the objectivity of meaning requires one to deny that there are thoughts in the sense of private episodes. Dummett notwithstanding, the objectivity of meaning, even when it is understood linguistically as Dummett understands it, does not require one to deny the inner and the private. Sellars has shown us how this is possible.73 Thoughts, Sellars argues, are private episodes containing items that play the same role in the ongoing inner discourse that various other items play in public language. Since they play the same role, they share the meaning with the latter. There is as it were an inner language – “mentalese”, Sellars calls it – the expressions of which can be translated into the public language that we share with others. We can have, for example, ‘red’ (in my mentalese) means .red. Meaning remains objective since what defines meaning is the notion of a linguistic role, a set of regularities and propensities governed by various rules. The regularities and propensities stay as they are, patterns shared by both the public episodes and the private episodes. So one can consistently maintain, on the one hand both that meaning is essentially linguistic and that this meaning is objective, and on the other hand that there are inner, private episodes which share in that meaning. What is central to the Sellarsian account of meaning that is defended by Dummett is that meaning is a matter of regularities and propensities. This means that in order to determine the meaning of a physical mark – or an item in mentalese – it is necessary to locate a mark or sound or twitch among a pattern of marks or sounds or twitches. In Brentano’s terms, it is essentially non-intentional items that carry meaning, and to determine that meaning it is necessary to locate that non-intentional item among a pattern

712 of non-intentional items. The meaningfulness is given by the pattern, the “linguistic role”, and by the items which make up the pattern. This view of the meaningfulness of thought is that of Titchener.74 The items that carry the meaning are non-intentional sensory contents of various sorts, according to Titchener, and their meaningfulness is a matter of their “context”, that is, the connections to other sensory contents. Titchener’s views are essentially those of the associationists, from Berkeley and Hume onwards,75 and, as Boring points out, are continuous with the behaviouristic theories of meaning of subsequent psychologists such as Holt and Tolman.76 If Sellars and Dummett – and Titchener – are correct, then to identify an item as expressing a thought, to identify it as meaning something, it is necessary to identify the pattern from which its meaningfulness derives. This pattern connects the meaningless content or item to other meaningless contents or items, past, present and future, to the extra-mental and extralinguistic world, and to bodily movements and actions. The pattern is exemplified by not just a present item but by this item together with all the other items to which it is connected by the pattern. But these other items are not given to us all at the same time. Thus, on the Sellars-DummettTitchener account of meaning, the meaning is never given to us wholly at a moment. This is clear as soon as one includes as one must various propensities in the definition of a linguistic role. For, propensities as such are not presented to us. (We see things dissolving but we do not see their solubility.) Titchener explicitly recognized this, and suggested that the meaning was carried unconsciously, physiologically. If the dispositional and lawful connections to non-present sensory contents were not presented, these dispositions could at least (he argued) be correlated with some (unconscious) physiological state. In this way Titchener hoped to account for the paucity of sensory contents in the case of higher mental processes. Indeed, he saw this as his crucial contribution to the context theory of meaning, as his improvement of that theory over the usual associationist version deriving from Berkeley. It is ingenious. But it will not work. For, it misses the important point that meaning is wholly presented to us. Consider the state of consciousness of a person hearing the word ‘bell’: the state of consciousness of a person who understands the language is very different from the state of consciousness of a person who does not understand the language. The person who understands the language is aware of the meaning of the word. This very simple point was made, basically in defence of Brentano and against Titchener, by the Würzburg psychologists gathered around Külpe.77 Titchener had his students work on

713 this problem, and they performed some remarkable feats of introspective analysis, but they were never able to make sensory contents into imageless thoughts: meaning simply was wholly present and that was all there was to it. Phenomenology requires that meaning and therefore thought be analyzed in such a way that it is wholly present in consciousness. The account of meaning given by Dummett, Sellars and Titchener fails precisely for not squaring with the phenomenological fact. Meaning is irreducible.78 Which was Brentano’s essential point against the associationists,79 be they the British associationists or the physiologically oriented associationists of Wundt’s laboratory and of Wundt’s student Titchener. However, this seems to bring us back to Frege’s point: if meaning is in consciousness, then, since what is in consciousness is private, it follows that meaning is private. But, to emphasize with Frege and Dummett and Sellars, meaning is public: several persons can think the same thought. How can we reconcile this point about the public nature of meaning with, on the one hand, meaning being present in consciousness and, on the other hand, not introducing entities such as Husserl’s Ideal forms inhabiting some non-empirical and fictitious third realm? There is a way out. This way out was identified by G. E. Moore. Frege to achieve publicity of thought is required to take thoughts outside consciousness and into a third realm because he begins from the assumption that what is present to the mind is particular. This is the assumption that he shares with Husserl, and which Dummett uncritically accepts, that the characteristics of things are as particular as the things of which they are characters. This assumption is, of course, not by any means required. John Stuart Mill denied it for one. Spencer had held with Berkeley and Hume that the characteristics of things are particular, and that their sameness is to be accounted for in terms of a relation of (exact) resemblance. It is precisely this that Mill denied. To the contrary, he argued, the characteristics of things are not particular, and one and the same character can be in several things. This makes the characteristics of things to be universals, in one sense of that term. But they are not Platonic forms nor are they Husserlian Ideal forms or species. For the characteristics of things that Mill is talking about are the characteristics that things are presented as having in ordinary experience. What is presented to us in ordinary experience are complexes – facts, if you wish; and the characteristics of things are simply presented as parts of these complexes. Several complexes contain the same characteristic. Characteristics of things are thus universal, but they do not inhabit a third realm; they are simply (ontological) parts of ordinary things, parts

714 therefore of the world of ordinary things, and parts furthermore that are presented to us in ordinary experience.80 This Millian view of the characteristics of things was later defended by G. E. Moore against the view of G. F. Stout that the characteristics of things are tropes or perfect particulars, and therefore as particular as the things they characterize.81 Dummett simply fails to notice this line of thought in analytic philosophy. This is partly because he tends to ignore the role of Moore and Russell in the tradition of analytic philosophy.82 It is partly because he identifies analytic philosophy with philosophy that has made the linguistic turn. But mostly it is because he accepts uncritically the nominalistic assumptions of Frege and Husserl. Moore specifically defends the view that the characteristics of things are universals in the context of a discussion of meanings and minds. He thereby provides a solution to Frege’s problem. If the characteristics of things are universals, then one and the same characteristic can, on the one hand, be given in ordinary experience, and can, on the other hand, be in several things. This will be as true of the mental as of anything else. Thus, there is nothing to prevent a character that characterizes my mental state from characterizing another, later mental state of mine; nor is there any reason why that character should not also characterize a mental state of yours. If we make meaning a universal, then it can characterize several mental states, several of mine and both mine and yours. If meaning is a universal, then that will enable meaning to be public. Moreover, since it is simply part of a complex given in experience, something given in inner awareness, to hold this is not to hold with Frege and Husserl that meanings must somehow be located in a third realm. To the contrary, meanings simply are part of the world of ordinary experience, as we ordinarily experience it. Moore makes the point of the Würzburg psychologists. He contrasts what happens when one hears one’s own language spoken, and what happens when one hears an unknown language spoken. ...in the first case, there occurs, beside the mere hearing of the words, another act of consciousness – an apprehension of their meaning, which is absent in the second case.83

He then contrasts the apprehension of the meaning of two different sentences, “twice two are four” and “twice four are eight”. There certainly are such things as the two different meanings apprehended. And

715 each of these two meanings is what I call a proposition.84

Moore quite clearly rejects the third realm view of meaning. He raises the “objection to the supposition that there are such things as propositions at all, and that belief consists merely in an attitude of mind towards these supposed entities.” ...if you consider what happens when a man entertains a false belief, it doesn't seem as if his belief consisted merely in his having a relation to some object which certainly is. It seems rather as if the thing he was believing, the object of his belief, were just the fact which certainly is not – which certainly is not, because his belief is false.85

But Moore, in his typical way, and without quite realizing it, introduces a second account of propositions which fits with his point about the consciousness of meaning without introducing the problems of the third realm. This is the view that a proposition is an element in the mental act which is the awareness of the meaning or proposition. He considers two beliefs, one the belief that a tree is an oak and the other the belief that the tree is an ash. We should, he suggests, ...distinguish, in the case of these two beliefs, between the element in respect of which they are both alike, the element which we express by saying that they are both acts of belief, and the elements in respect of which they differ; and to say these latter elements are in the one case, the proposition that the tree is an oak and in the other the proposition that it is an ash.86

The proposition thus turns out to exist in the mental act, in the state of consciousness which is the awareness of it, and not in some mysterious third realm outside of ordinary experience. A proposition is simply a characteristic of a mental state. It is that matter in the mental act about which Husserl was concerned, the matter that makes the act about one state of affairs rather than another. But where Husserl makes this matter as particular as the act of which it is an element, Moore is prepared to treat it as he treats of other characteristics of things, as a universal. Thus, propositions, the characteristics of mental acts that constitute their matter, are like other characteristics of things that are presented in ordinary experience: they can be common to several things. For, Moore clearly rejects the nominalist view that the characteristics of things are as particular as the things of which they are characteristics: “...when we say that two things resemble one another, what we mean by this is always merely that they have some property in common.”87 Considering

716 property in common.”87 Considering specifically the characteristic of being white, he points out that “...we mean by ‘pure white’ something different from any particular patch of pure white or from the sum of them all – something which is itself a universal common to them all.”88 This characteristic is one that we are aware of in ordinary experience: it would be a mistake to “hold that, in a case where we see two patches of exactly the same shade of colour, or one patch of the shade in question, this shade of colour is something which we don't see....”89 Propositions or meanings are characteristics of just this sort, universals which we experience as parts of the things that we are aware of in ordinary experience. That being so, we can hold that meanings are indeed public: the same meaning, being a universal, can be in my acts of awareness and also in yours. Frege’s problem is thus solved without the introduction of a third realm. This Moorean solution to the problem was pursued with some vigour by Gustav Bergmann in a number of important essays.90 Bergmann develops the views of Külpe and Moore much as Sellars and Dummett develop the views of Titchener. In Bergmann’s case, in the formula for meaning or intentionality, +p,Mp the definite description +p, refers to the proposition exemplified in the mental act, where this proposition is argued to be a universal. Bergmann also argued this formula is a necessary truth; and, more strongly, that ‘M’ is analogous to a relation. This latter led him back to something like the propositions theory, where even a false proposition has an existing object.91 All the standard objections to the proposition theory apply to this ontology that Bergmann came eventually to develop. One can, however, reject Bergmann’s account of ‘M’ and the rich ontology that flows from it while still recognizing how he was attempting to do justice to the objectivity of thought, that is, of the character denoted by the definite description ‘+ p ,’.92 Dummett, then, through his uncritical acceptance of the nominalism of Frege and Husserl, and through his failure to look at the stream of analytic philosophy following upon the work of Moore and Russell, fails to see that his own, Sellarsian account of meaning is not the only viable response to Frege’s problem of the objectivity of meaning.93 These are the structural connections that Dummett does not explore in his structural history of the genesis of analytical philosophy. His account, therefore, of some of the “origins of analytical philosophy” is severely limited. It must be said, however, that Dummett did not set out to write a detailed account

717 of the origins of analytical philosophy, only an account of some of its origins. However, he did set out to give an account of analytical philosophy. By this he means, as we have seen, a very narrow interpretation of what constitutes analytical philosophy. He means his own philosophy and, one presumes, views such as those of Sellars that are akin to it and even antedate it. It is perhaps this narrow definition of what constitutes analytical philosophy that prevents Dummett from recognizing that there are responses to Frege’s problem other than his own. One must say in the end, however, that whatever the limitations of the book, and whatever the limitations of Dummett’s own philosophy, his Origins of Analytical Philosophy is a very good book, laying out briefly but in a very perspicuous form some important structural history. I have suggested that the book does not say everything that needs to be said. But few books do. And Dummett’s book says much more than most. That is what makes it a good book.

718 Endnotes to Study Nineteen

1.This book grew out of a set of lectures Dummett delivered in Italy in 1987. The lectures appeared originally in Lingua e Stile, 13 (1988), pp. 171-210. They were subsequently translated into German together with a transcript of an interview with the translator, Dr. Joachim Schulte. The present volume is a revised form of the lectures together with a translation of the interview. Page references to this book will appear in parentheses in the text. 2.See for example, G. Bergmann’s introductory remarks to his essay “Some Remarks on the Philosophy of Malebranche,” in his Meaning and Existence (Madison, WI: University of Wisconsin Press, 1960), pp. 189-204. 3.See Bergmann’s essays on Frege, “Frege’s Hidden Nominalism” in his Meaning and Existence, pp. 205-224, and “Ontological Alternatives,” in his Logic and Reality (Madison, WI: University of Wisconsin Press, 1964), pp. 124-157; and also the essay “The Ontology of Edmund Husserl” in the latter, pp. 193-224, as well as his important study Realism (Madison, WI: University of Wisconsin Press, 1967). See also R. Grossmann’s The Structure of Mind (Madison, Wisc.: University of Wisconsin Press, 1965), and his study Reflections on Frege’s Philosophy (Evanston, Ill.: Northwestern University Press, 1969). See also Grossmann’s essay, “Frege’s Ontology,” in E. B. Allaire et al., Essays in Ontology (The Hague: Nijhoff, 1963), pp. 106-120. 4.Cf. E. G. Boring, A History of Experimental Psychology, Second Edition (New York: Appleton-Century-Crofts, 1957). 5 Barry Smith, “On the Origins of Analytic Philosophy,” Grazer Philosophische Studien, 35 (1989), pp. 153-73, makes this distinction terms of German and Austrian traditions in philosophy. 6.See in particular James Ward’s article on “Psychology” in the 1886 edition of the Encyclopaedia Britannica, Peale (American) Reprint (Chicago: R. S. Peale Co., 1892), vol. XX, pp. 37-85. 7.Cf. G. E. Moore, “The Nature of Judgment,” Mind, n.s. 8 (1899), pp. 176-193.. On Locke, cf. F. Wilson, “The Lockean Revolution in the Theory of Science,” in S. Tweyman and G. Moyal, eds., Early Modern Philosophy: Epistemology, Metaphysics and Politics (New York: Caravan Press, 1986), pp. 65-97. For the connection between Locke and Moore, cf. F. Wilson, “Perceptual Ideality and the Ground of Inference: Comments on Ferreira’s Defence,” Bradley Studies, 1 (1995), pp. 139-52.

719

8.G. E. Moore, “The Nature of Judgment,” p. 182. 9.As R. Rorty suggests, it was Bergmann who first introduced this phrase. See G. Bergmann, Logic and Reality, p. 177; and R. Rorty, Introduction to his anthology, The Linguistic Turn (Chicago: University of Chicago Press, 1967), p. 9. 10.Cf. G. Bergmann, “Particularity and the New Nominalism,” in his Meaning and Existence, pp. 91-105. 11.Bergmann’s The Metaphysics of Logical Positivism, first edition (New York: Longmans Green, 1954) contains a record of his efforts to break free from Carnap’s nihilism. For details of the journeys of both Bergmann and Carnap, see H. Hochberg, The Positivist and the Ontologist: Bergmann, Carnap and Logical Realism (Amsterdam: Rodopi, 2001). Bergmann was a colleague of Wilfrid Sellars and Everett Hall for some years at the University of Iowa. These three important analytical ontologists had a significant impact on each other. 12.See G. Bergmann, “Two Criteria for an Ideal Language”, Philosophy of Science, 16 (1949), pp. 71-74; reprinted in Rorty, The Linguistic Turn. 13.Cf. Herbert Hochberg, “Professor Quine, Pegasus, and Dr. Cartwright,” in his Logic, Ontology, and Language (Munich: Philosophia Verlag, 1984), pp. 86-100. 14.Cf. W. V. O. Quine, “Soft Impeachment Disowned,” Pacific Philosophical Quarterly, 61 (1980), pp. 450-451. 15.Dummett, “Nominalism,” Philosophical Review, 65 (1956), pp. 504-5. 16See G. Bergmann, “Intentionality,” in his Meaning and Existence, pp. 3-38. 17.G. Frege, “On Sense and Reference,” in Translations from the Philosophical Writings of Gottlob Frege, second edition, ed. and trans. P. Geach and M. Black (Oxford: Blackwell, 1960), p. 59. 18.The central text is Edmund Husserl, Logical Investigations, First German Edition, 1900. References will be to the translation of J. N. Findlay (London: Routledge and Kegan Paul, 1970), 2 vols., and will be to Investigation number, section number, volume of the Findlay translation, and page. 19.Cf. Logical Investigations, V, sec. 13, vol. 2, p. 563. 20.Cf. Logical Investigations, V, sec. 20, vol. 2, p. 586ff. 21.Gilbert Ryle, “Are There Propositions?”, Proceedings of the Aristotelian Society,

720 n.s. 30 (1930), p. 92ff, referred to this argument for propositions as “the argument from the intentionality of acts of thinking.” 22.G. E. Moore, Some Main Problems of Philosophy, p. 258. 23.Ibid., p 263; Ryle, “Are There Propositions?”, p. 105. 24.Ryle, pp. 106-7. 25.Ryle, p. 106. 26.Logical Investigations, V, sec. 11, vol. 2, p. 557ff. 27.Intentionality is not introduced until Investigation V; the third realm is introduced and defended in Investigation II. 28.Husserl, Logical Investigations, II, Intro., vol. I, p. 339. 29.Ibid., II, sec. 3, vol. I, p. 343. 30.Ibid., II, sec. 4, vol. I, p. 343f. 31Cf. R. Grossmann, “Conceptualism” and “Sensory Intuition and the Dogma of Localization,” in E. B. Allaire et al., Essays in Ontology, pp. 50-63. 32.Cf. Husserl, Logical Investigations, III, sec. 11, vol. 2, p. 455f. 33.Ibid., v, sec. 16, sec. 20, vol. 2, p. 576ff, p. 586ff. 34.Ibid., V, sec. 17, vol. 2, p. 578ff. 35.Ibid., VI, ch. 1, vol. 2, p. 675ff. 36.Ibid., VI, sec. 67, vol. 2, p. 837. 37.B. Russell, “On Denoting,” in his Logic and Knowledge, ed. R. C. Marsh (New York: Allen and Unwin, 1968), pp. 45-55. 38.H. Hochberg, “Russell’s Attack on Frege’s Theory of Meaning,” in his Logic, Ontology, and Language, pp. 60-85 See also the discussion of the same topic in his Thought, Fact and Reference: The Origins and Ontology of Logical Atomism (Minneapolis, MN: University of Minnesota Press, 1978). 39.Russell, “On Denoting,” p. 49. 40Ibid. 41.Cf. Barry Smith, “On the Origins of Analytic Philosophy,” p. 163, p. 169.

721

42.Some of the story is told in F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill (Toronto: University of Toronto Press, 1990). 43.Others have made the same mistake of concluding that since their account is behavioural and therefore objective that it is therefore non-psychologistic; cf. M. Brodbeck, “Explanation, Prediction, and 'Imperfect' Knowledge,” in her Readings in the Philosophy of the Social Sciences (New York: Macmillan, 1968), pp.363-98. 44.For a brief discussion, see W. Sellars, “Notes on Intentionality”, in his Philosophical Perspectives (Springfield, IL: Charles C. Thomas, 1967), pp. 308-320. See also his “Some Reflections on Language Games,” in his Science, Perception and Reality (London: Routledge and Kegan Paul, 1963), pp. 321-358; and “Language as Theory and Language as Communication,” Philosophy and Phenomenological Research, 29 (1969), pp. 506-27. For discussion of some of Sellars’ views, see the study above, “Marras on Sellars on Thought and Language.” 45The account is essentially behaviouristic, as Sellars recognizes; see his “Behaviorism, Language and Meaning,” Pacific Philosophical Quarterly, 61 (1980), pp. 3-30. Because thoughts are identified by Dummett with language behaviouristically understood, his views can correctly be characterized as materialistic. For some, I suppose, that is a good thing. 46.W. Sellars, “Notes on Intentionality,” in his Philosophical Perspectives (Springfield, Ill.: Charles C. Thomas, 1967), p. 310. 47.W. Sellars, “Inference and Meaning,” Mind, n.s. 62 (1953), p. 313-338. 48See also Sellars, “Synthetic A Priori,” in his Science, Perception and Reality. There are serious objections to this account of meaning, however; see the essay above, “Acquaintance, Ontology and Knowledge”; and also F. Wilson, “Perceptual Ideality and the Ground of Inference: Comments Ferreira’s Defence.” 49.Concerning implicit definition, see the essay above, “Implicit Definition Once Again.” 50.W. Sellars, “Naming and Saying,” in his Science, Perception and Realit., pp. 225246. 51For this notion, see the study above, “Implicit Definition Once Again.” 52.M. Dummett, “Wittgenstein’s Philosophy of Mathematics,” Philosophical Review, 68 (1959), p. 348.

722

53.Cf. W. Sellars, “Hochberg on Mapping, Meaning and Metaphysics,” Midwest Studies in Philosophy, 2 (1977), pp. 219-20. 54.W. Sellars, Science and Metaphysics: Variations on a Kantian Theme (London: Routledge and Kegan Paul, 1968), p. 132. 55.Ibid., p. 133. 56.F. Brentano, The True and the Evident, trans. R. Chisholm (London: Routledge and Kegan Paul, 1966), p. 9. 57.Sellars, Science and Metaphysics, p. 136. 58.M. Dummett, “A Remarkable Consensus,” New Blackfriars, 68 (1987), p. 427. This essay gave rise to a lively exchange. A list of the contributions can be found at the end of Dummett’s final contribution, “What Chance for Ecumenism?” New Blackfriars, 69 (1988), pp. 544-55. 59.Ibid., p. 14. 60.See W. Sellars, “Abstract Entities,” in his Philosophical Perspectives. 61It seems to be the position of G. F. Stout that what accounts for sameness is a primitive relation of resemblance; see his contribution to the symposium “Are Characteristics of Things Universal or Particular?” Aristotelian Society, Supplementary Volume III, 1923, pp. 114-122. The other contributors to the symposium were G. E. Moore and G. Dawes Hicks. Russell offered in his “On the Relations of Universals and Particulars” in his Logic and Knowledge (ed. R. C. Marsh [London: Allen and Unwin, 1956], pp. 105124), a classic argument against this view – it must take exact resemblance itself to be a primitive relation which itself must be a universal, since, if a, b, and c are all the same shade of red, then a must resemble b in exactly the same way that b resembles c, and in exactly the same way that a resembles c. The relation of resemblance characterizes each of these pairs, which means that it is a universal. But if one is going to allow this one universal, one might just as well allow others, and take the characteristics of things to be universals. For defences of Russell’s argument, see A. Donagan, “Universals and Metaphysical Realism” (Monist, 47 [1963], pp. 211-146), and H. Hochberg, “Russell’s Proof of Realism Reproved” (in his Logic, Ontology and Language [Munich: Philosophia Verlag, 1986], pp. 196-203). See also H. Hochberg, “Logical Form, Existence and Relational Predication, ”in his Logic, Ontology and Language, pp. 204-230. 62Cf. J. S. Mill, System of Logic, Eighth Edition (vol.. 7 of his Collected Works, ed. J.

723

62Cf. J. S. Mill, System of Logic, Eighth Edition (vol.. 7 of his Collected Works, ed. J. Robson [Toronto: University of Toronto Press, 1978] originally published, London: Longmans Green, 1872), Book II, Ch. ii, sec. 3, note. Mill puts the argument that characteristics of things are not perfect particulars in this way. He states Spencer’s view thus: he (Spencer) “...maintains that we ought not to say that Socrates possess the same attributes which are connoted by the word Man, but only that he possesses attributes exactly like them...”. Mill then goes on to object that Mr. Spencer is of the opinion that because Socrates and Alcibiades are not the same man, the attribute which constitutes them men should not be called the same attribute.... If every general conception, instead of being ‘the One in the Many,’ were considered to be as many different conceptions as there are things to which it is applicable, there would be no such thing as general language. A name would have no general meaning if man connoted one thing when predicated of John, and another though closely resembling thing when predicated of William. .................. ...The things compared are many, but the something common to all of them must be conceived as one ... 63 Herbert Spencer, Principles of Psychology, 2 vols. (New York: D. Appleton and Co., 1902), vol. II, p. 294. 64 M. Dummett, Frege: Philosophy of Language (London: Duckworth, 1973), pp. 503-4. 65 Ibid., p. 577. 66 Fraser Cowley, Metaphysical Delusion (Buffalo: Prometheus Books, 1991), p. 118. 67But so does Sellars; see his “Concepts as Involving Laws and Inconceivable without Them,” in his Pure Pragmatics and Possible Worlds: The Early Essays of Wilfrid Sellars ed. with intro. by J. Sicha (Reseda, CA: Ridgeview Publishing Co., 1980), pp. 95-124, where he proposes that “universals and laws are correlative, same universals, same laws, different universals, different laws” (p. 104). See the discussion in of Sellars’ view in “Universals, Bare Particulars and Tropes: The Role of a Principle of Acquaintance in Ontology,” above. 68.D. Pears, “Universals,” Philosophical Quarterly, 1 (1950-51), pp. 218-227.

724

69.P. Butchvarov, Resemblance and Identity (Bloomington, IN: University of Indiana Press, 1966), Ch. One. See also H. Hochberg, “Mapping, Meaning, and Metaphysics,” and “Sellars and Goodman on Predicates, Properties and Truth,” both in his Logic, Ontology and Language, pp. 157-84 and pp. 185-95, respectively. 70.Once again, see the discussion in of Sellars’ view in “Universals, Bare Particulars and Tropes: The Role of a Principle of Acquaintance in Ontology,” above. 71.Cf. Sellars, “Naming and Saying.” 72 This is the principle of localization discussed by R. Grossmann in his “Sensory Intuition and the Dogma of Localization.” 73.W. Sellars, “Empiricism and the Philosophy of Mind,”in his Science, Perception and Reality, p. 186ff: and also “Notes on Intentionality.” 74.Cf. the selections from Titchener in R. Herrnstein and E. G. Boring, A Source Book in the History of Psychology (Cambridge, Mass.: Harvard University Press, 1966), p. 186f. 75.This is the received opinion, but in the case of Hume it must be somewhat qualified; cf. F. Wilson, “Association, Ideas and Images in Hume,” in P. Cummins and G. Zoeller, eds., Minds, Ideas and Objects (Atascadero, CA: Ridgeview Publishing Co., 1990), pp. 255-74. 76.Cf. Herrnstein and Boring, op. cit., p. 193, p. 197. 77.G. Humphrey, Thinking (New York: Wiley, 1963), Ch. I-IV summarizes excellently the whole discussion. In particular, Würzburg is defended against Titchenerian criticisms in Ch. IV, esp. p. 122ff. 78.This in fact led to the downfall of the research programme of classical psychology based on the method of systematic introspective analysis; cf. F. Wilson, Psychological Analysis and the Philosophy of John Stuart Mill, Ch. 8. 79.Cf. F. Brentano, Psychologie vom empirischen Standpunkt (Hamburg: Felix Meiner, 1955), vol. I, p. 190ff. 80.To put it another way, properties, even when they are understood as universals, capable of being literally the same entity present several objects, are given to one in sense experience; but neither Platonic Forms not Husserlian Ideal Forms are given in sense, they are the objects rather of a non-sensory rational intuition. There is an argument against Platonism that goes something like this: “it cannot

725 be true that there are Forms that ground necessary connections among things, since they are abstract entities occupying a Third Realm and therefore not in space and time; but all causally relevant entities are in space and time; the Forms are therefore causally inert and useless philosophically.” (See, for example, P. Benaceraf, “Mathematical Truth,” Journal of Philosophy, 70 [1973], pp. 661-679; and, following him, R. Hanna, Rationality and Logic [Cambridge, MA: MIT Press, 2006], p. 33.) But one must be careful. First, on the Platonic-Aristotelian scheme of things (cf. “Bergmann’s Hidden Aristotelianism,” above) the Forms are causally relevant, but the account of causation and of explanation is very different from the “covering law” model of science. Second, they are certainly not among the ordinary this-worldly causes of ordinary things. Third, sensible qualities, understood as universals, are among the causally relevant entities – for the causal regularity, “Water, when heated boils”, the property of having a certain degree of heat is certainly causally relevant, but it is also a universal (at least on the view defended by Mill), and in this sense universals are causally relevant. Moreover, fourth, the heat is something given in our sensible experience of the world, so in this sense universals are given in sense experience, though Platonic Forms or Husserlian Ideal Forms are not given in such experience. But, fifth, properties as are not localized in space and time – they are present in ordinary objects, ordinary particular things, which are localized, but as universals they can occur in several such particulars – they are present but not wholly present in the objects they characterize. Finally, sixth, to hold that an entity to be causally relevant must be localized in space and time is to fall into a nominalism in which sensible properties are tropes, entities which are as particular as the entities which they characterize – it is to fall into the position defended by Spencer rather than the one defended by Mill. 81.G. E. Moore, “Are the Characteristics of Things Universal or Particular?” in his Philosophical Papers (London: Allen and Unwin, 1959), pp. 17-31.. 82.It is part of the Oxford tradition to loath Russell: it comes from John Cook Wilson, through H. A. Prichard to more recent “Oxford philosophy.” 83.G. E. Moore, Some Main Problems of Philosophy (London: Allen and Unwin, 1953), p. 59. 84.Ibid. 85.Ibid., p. 263. 86.Ibid., p. 308. 87.Ibid., p. 358. 88.Ibid., p. 340. 89.Ibid., p. 375.

726

90.See Gustav Bergmann, “Intentionality,” in his Meaning and Existence; pp. 3-38, and also “Acts,” in his Logic and Reality, pp. 3-44. Sellars criticized Bergmann’s views in his “Notes on Intentionality.” For other critical comments, see also the essay above, “Effability, Ontology and Method.” 91.See G. Bergmann, New Foundations of Ontology (Madison, Wisc.: University of Wisconsin Press, 1992). 92 For some critical discussion, see the essays above, “Effability, Ontology and Method” and “The Aboutness of Thought”; and, for a different perspective, L. Addis, Natural Signs: A Theory of Intentionality (Philadelphia: Temple University Press, 1989). 93 Dummett ignores Bergmann’s work as much as he ignores that of Moore and Russell, and, for that matter, just as much as he ignores the work of Sellars. This is unfortunate; he could have learned much from all these thinkers. Among other things, he would have benefited from Bergmann’s important essays on Frege and Husserl. Dummett would undoubtedly have also benefited from R. Grossmann, Reflections on Frege’s Philosophy. He seems to prefer the dabblings of such philosophers as P. T. Geach to those like Bergmann, Sellars, and Grossmann who have a sounder grasp of ontological issues.

PhilosophischeAnalyse PhilosophicalAnalysis 1 Herbert Hochberg Russell, Moore and Wittgenstein The Revival of Realism

8 Rafael Hüntelmann Existenz und Modalität Eine Studie zur Analytischen Modalontologie

2

9 Andreas Bächli / Klaus Petrus Monism

ISBN 3-937202-00-5 334 pp., Hardcover € 94,00

Heinrich Ganthaler Das Recht auf Leben in der Medizin Eine moralphilosophische Untersuchung

ISBN 3-937202-01-3 167 pp., Hardcover € 58,00

3 Ludger Jansen Tun und Können Ein systematischer Kommentar zu Aristoteles’ Theorie der Vermögen im neunten Buch der „Metaphysik“ ISBN 3-937202-02-1 302 pp., Hardcover € 70,00

4 Manuel Bremer Der Sinn des Lebens Ein Beitrag zur Analytischen Religionsphilosophie ISBN 3-937202-03-X 134 pp., Hardcover € 58,00

5 Georg Peter Analytische Ästhetik Eine Untersuchung zu Nelson Goodman und zur literarischen Parodie ISBN 3-937202-04-8, 332 pp. Hardcover € 94,00

6 Wolfram Hinzen / Hans Rott Belief and Meaning Essays at the Interface ISBN 3-937202-05-6 250 pp., Hardcover € 58,00

7 Hans Günther Ruß Empirisches Wissen und Moralkonstruktion Eine Untersuchung zur Möglichkeit von Brückenprinzipien in der Natur- und Bioethik ISBN 3-937202-06-4 208 pp., Hardcover € 58,00

ISBN 3-937202-07-2 189 pp., Hardcover € 58,00

ISBN 3-937202-19-6 340 pp., Hardcover € 70,00

10 Maria Elisabeth Reicher Referenz, Quantifikation und ontologische Festlegung ISBN 3-937202-39-0 ca. 300 pp., Hardcover € 89,00

11 Herbert Hochberg / Kevin Mulligan Relations and Predicates ISBN 3-937202-51-X 250 pp., Hardcover € 74,00

12 L. Nathan Oaklander C. D. Broad's Ontology of Mind ISBN 3-937202-97-8 105 pp., Hardcover € 39,00

13 Uwe Meixner The Theory of Ontic Modalities ISBN 3-938793-11-2 374 pages, Hardcover,€ 79,00

14 Donald W. Mertz Realist Instance Ontology and its Logic ISBN 3-938793-33-3 252 pp., Hardcover, EUR 79,00

15 N. Psarros / K. Schulte-Ostermann (Eds.) Facets of Sociality ISBN 3-938793-39-2 370 pp., Hardcover, EUR 98,00

16 Markus Schrenk The Metaphysics of Ceteris Paribus Laws ISBN 13: 978-3-938793-42-8 192pp, Hardcover, EUR 79,00

EditedBy • HerbertHochberg • RafaelHüntelmann ChristianKanzian • RichardSchantz • ErwinTegtmeier

PhilosophischeAnalyse PhilosophicalAnalysis 17 Nicholas Rescher Interpreting Philosophy The Elements of Philosophical Hermeneutics ISBN 978-3-938793-44-2 190pp., Hardcover € 89,00

18 Jean-Maurice Monnoyer(Ed.) Metaphysics and Truthmakers ISBN 978-3-938793-32-9 337 pp., Hardcover € 98,00

19 Fred Wilson Acquaintance, Ontology, and Knowledge Collected Essays in Ontology ISBN 978-3-938793-58-9 XX, 726., Hardcover, EUR 159,00

EditedBy • HerbertHochberg • RafaelHüntelmann ChristianKanzian • RichardSchantz • ErwinTegtmeier