Reference, Rationality, and Phenomenology: Themes from Føllesdal 9783110323542, 9783110323016

Citation preview

Michael Frauchiger (Ed.) Reference, Rationality, and Phenomenology Themes from Føllesdal

LAUENER LIBRARY OF ANALYTICAL PHILOSOPHY Edited by Wilhelm K. Essler and Michael Frauchiger

VOLUME 2

Michael Frauchiger (Ed.)

Reference, Rationality, and Phenomenology Themes from Føllesdal

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.

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

2013 ontos verlag P.O. Box 15 41, D-63133 Heusenstamm www.ontosverlag.com ISBN 978-3-86838-182-5 2013 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 CPI buch bücher.de

Contents Preface Michael Frauchiger

1

Dagfinn Føllesdal: A Personal Memoir Jon Elster

5

Reference, neither Causal nor Apostolic, but Normative Dagfinn Føllesdal

13

ESSAYS PART I: PHENOMENOLOGY Neuropsychological Foundations of Phenomenology: Is It Possible? 33 Patrick Suppes Consciousness, Modality, and Inner Awareness David Woodruff Smith

49

Noema and Reference Christian Beyer

73

Transcendental Philosophy and Modern Physics: Neo-Kantianism, Logical Empiricism, and Phenomenology Michael Friedman

89

Hume’s Phenomenological Conception of Space, Time and Mathematics Graciela De Pierris

107

ESSAYS PART II: SCIENCE On Solidity and Rigidity: Some Notes on a Paper of Dagfinn Føllesdal Wilhelm K. Essler Some Remarks on the Distinction between Basic (Theoretical) and Applied (Practical) Science and Its Importance in the Politics of Science Nils Roll-Hansen

123

137

ESSAYS PART III: LOGIC AND RATIONALITY Some Consequences of the Entanglement of Logic and Mathematics Charles Parsons

153

Validity of Inferences Dag Prawitz

179

Reason and Rationality Jon Elster

205

ESSAYS PART IV: MEANING On “Meaning and Experience” Olav Gjelsvik

221

ESSAYS PART V: REFERENCE Føllesdal and Quine’s Slingshot John Perry

237

Føllesdal and Frege on Reference Øystein Linnebo

259

COMMENTS ON THE ESSAYS Comments on Suppes, Smith, Beyer, Friedman and De Pierris Comment on Essler Comment on Roll-Hansen Comment on Parsons Comment on Prawitz Comment on Elster Comment on Gjelsvik Comment on Perry Comment on Linnebo Dagfinn Føllesdal

283 297 299 303 306 311 314 320 325

Interview with Dagfinn Føllesdal by Michael Frauchiger

335

About the Editor

Preface The papers collected in this book hearken back to the 2nd International Lauener Symposium on Analytical Philosophy held by the Lauener Foundation at the “Haus der Universität” in Bern on 22 – 23 June 2006. The symposium was dedicated to the work of Dagfinn Føllesdal (C.I. Lewis Professor of Philosophy, Emeritus, at Stanford University and Emeritus Professor of Philosophy at the Universitetet i Oslo), who was awarded the Lauener Prize for an Outstanding Oeuvre in Analytical Philosophy on the same occasion. All the essays included in the present collection thus connect directly or indirectly with Føllesdal’s work. The symposium was by many considered one of the most important philosophy conferences in Switzerland for some time and brought together a number of renowned philosophers to discuss and develop further a wide range of philosophical topics to which Føllesdal has been giving new impetus since the mid-20th century. In the period that has followed that event, all the papers initially presented there have been thoroughly reworked by their authors, based on the rich discussions that took place at the symposium. In addition, those papers have been supplemented with further substantial pieces, particularly Føllesdal’s detailed commentaries on the essays, Elster’s memoir (a revision of his laudatio at the award ceremony in honour of Føllesdal) and my interview with Føllesdal (initially dating from 2008 but the transcript has been substantially revised in 2012), all of which provide an illuminating lead into central themes from Føllesdal. All contributions have been written for the present book and have partly been completed as recently as this summer of 2012. They have merged into this collection to provide high-quality revisitation of an array of key issues in theoretical and practical philosophy which Føllesdal has been opening up and deepening over the intervening decades. The main part of the book – the essay section – starts after Elster’s “Dagfinn Føllesdal: A Personal Memoir”; it contains 14 substantial articles and begins with an essay by Føllesdal himself, in which he portrays his approach to philosophy of language and refines his influential theory of

2

reference. From there the essays are divided into five parts: Phenomenology, Science, Logic and Rationality, Meaning, and Reference. Likewise, Føllesdal’s subsequent comments on the essays – which are clarifying and advancing the debated questions as well as making clear the thematic connections to his own work and among some of the essays – are arranged into those same five groups of themes. My ensuing interview with Føllesdal – which dwells on these and further issues within his works – concludes the book. Dagfinn Føllesdal’s unremitting efforts to reduce historical unconcern, ignorance of sources and dogmatic partiality by raising awareness of complementary approaches to pertinent philosophical problems – together with his insistence on the clarity of basic concepts, on the use of evidence and on well-founded rational argumentation in each field of philosophy – have famously facilitated fresh, broad-minded discussions on a high level between philosophers of different academic backgrounds. In line with this, Føllesdal has consistently criticized the labelling of rival philosophical views, i.e. the prevalent use of vague labels such as ‘analytic’/‘continental’, ‘modern’/‘traditional’, or ‘naturalism’/‘transcendentalism’, which may simply serve to undermine and disqualify the bulk of differing, uncongenial views. Suchlike concerns underlie Føllesdal’s fruitful decades-long efforts to reconnect diverse, even adverse, philosophical traditions or movements within contemporary philosophy. Thus Føllesdal’s remarkably intelligible interpretative approach to Husserl’s philosophy, which he has elaborated and deliberated over the course of the last sixty years, has had a seminal impact on bringing “phenomenology” and “analytic philosophy” together by revealing similar guiding questions and approaches from which to start an exchange and a productive cooperation. At this point it is in order to proactively repudiate any allegations of attaching a label to Føllesdal’s undogmatic, non-partisan philosophy by bestowing the Lauener Prize for an Outstanding Oeuvre in Analytical Philosophy upon him. For “analytical philosophy” is here not narrowly conceived, say, as a kind of (neo- or post-) “logical positivist” school of thought which is opposed to “continental philosophy” or whatever. The term ‘analytic philosophy’ may in recent years have degenerated into an

3

over extensively used label applied to a jumble of styles of philosophy, denoting, if anything, a worldwide yet fragmented network of philosophers, with no trace of homogeneity to be found as far as philosophical method or subject matter is concerned. At the same time, however, a metaphilosophically more suitable usage of ‘analytical philosophy’, which is broad enough to admit a multiplicity of philosophical movements and positions, without being vacuous, continues to loosely mark out that research in philosophy which satisfies a decisive amount of items in a revisable list of heuristic and methodological criteria for the pertinent practice of philosophy. Such a reasonably broad, open concept of analytical philosophy, then, applies to that philosophical research which is (say) accurate, lucid, comprehensible to the intended readership, illustrative, critically reflective, supports claims by evidence and arguments, bewares of sweeping judgements, challenges its own assumptions, identifies common presuppositions that preclude problem solving, considers questions from various perspectives, includes proper and adequate interpretation of other authors’ views, has a clear overall argumentative structure, is informed by recent logics and semantics, is using formalisms appropriately, draws on fruitful metaphors and analogies or thought experiments, adequately takes into account the history of philosophy, proceeds by careful case studies (where appropriate), keeps up with relevant scientific research, incorporates detailed empirical findings by putting them into a broader framework, etc. For sure, Dagfinn Føllesdal is a paragon “analytical philosopher” understood in this broad yet not metaphilosophically empty sense. Acknowledgements First and foremost I wish to thank Dagfinn Føllesdal for having been so actively involved in the Symposium and, in its aftermath, so highly dedicated to the working out of the present book. Beyond his substantial contributions to the volume, including his own article and his detailed comments on the other authors’ articles, he has made judicious and candid replies in a comprehensive interview he granted me, delivering illuminative insights into unexpected connections between the manifold aspects of his oeuvre, which spans theoretical and practical problems as

4

well as different philosophical ways of looking at them which were deemed mutually incompatible for too long. Personally, I have strongly benefited both at academic and human level from the very enriching cooperation with Dagfinn Føllesdal in the preparation of this book, and I am much obliged to him for that. Moreover, I owe each contributor much. The combined and notably sustained commitment of all these very busy philosophers has facilitated me to bring this very ambitious book project to a favourable issue. Furthermore I would like to thank Wilhelm K. Essler, the president of the “Lauener-Stiftung” (Lauener Foundation) and co-editor of the series Lauener Library of Analytical Philosophy, for giving me encouragement in bringing the present substantial book – the second volume in that series – into being. I am also grateful to Stephan Hottinger, Alex Burri, Daniel Schulthess and Dieter Jordi, my other fellow members of the board of trustees of the Lauener Foundation, for looking with much favour upon the volumes of the Lauener Library. And last, but not least, I am thankful to Rafael Hüntelmann, the director of Ontos Verlag, for his patient support and helpful publishing efforts. Michael Frauchiger

Dagfinn Føllesdal: A Personal Memoir Jon Elster*

I have known Dagfinn since 1966. I was reading philosophy in Oslo when he came back from Harvard, first on a research fellowship and then from 1967 as professor and chair of the department. I served as the administrative assistant of the department under him for one year (and as his research assistant in his introductory course on logic), until I went off to France to do my doctorate. I believe he had already set his mind on a total reform of the department, which at the time was in a sadly unprofessional and underdeveloped state. It was an uphill struggle, in which he faced an unholy alliance of philosophical obscurantists and political extremists. He was caught up not only in the student movement that flourished in all Western societies at the time, but also in a fierce local “Positivismusstreit” in which opposition to analytical philosophy and to the Vietnam war somehow merged into one stance. The arrival of Dag Prawitz a few years later made his life a bit easier, but it took many more years before his plans were fully realized. Today the department is a vibrant place, where graduate students get a rigorous training and become fully integrated within the international academic community. It would not have happened without Dagfinn. Still on a personal note, I want to say a few words about how Dagfinn has shaped my own development. After I had finished my studies in France, Dagfinn was instrumental in helping me get visiting appointments at Berkeley and Stanford. I might have gone to the U.S. in any case, sooner or later – but then again I might not have. Even more decisive was his help in straightening out my way of thinking and arguing. Although I was basically committed to the analytical way of doing philosophy before I went to France, the three years I spent there – the golden age of charlatanism – undeniably did some damage to that commitment. To some extent, I adopted the local habits of elliptical *

Columbia University

6 profundity and reasoning by analogy. I believe Dagfinn cured me of these tendencies – not by criticism, but by the example he set. The question “What would Dagfinn think of this argument?” provided a very effective and much needed cold shower when my imagination got overheated. That is why in the Preface to the first book I published in English I thanked him for teaching me “what intellectual honesty means and requires”.1 I cannot and shall not try to offer an appreciation of Dagfinn’s achievements in the areas of modal logic, philosophy of language and phenomenology. Other contributors to this symposium deal with these issues in much greater depth than I could possibly achieve. Before I address some of our shared intellectual concerns, I would like to make a few general remarks. Let me begin by observing that it is somewhat superficial to refer to the “areas” in which Dagfinn has worked, in the sense in which for instance Noam Chomsky has written in two “areas”, American foreign policy and linguistic theory. As I have often heard him say, Dagfinn staunchly refuses the distinction between “continental” and “analytical” philosophy. From his Master’s Thesis on “Husserl und Frege”2 onwards, these have been part and parcel of one pursuit. For the same reason, I would object to Daniel Dennett’s characterization of “the dagfinn” in his Philosophical Lexicon: One of the two possible outcomes of crossing a shark with a dolphin (the other is the follesdal). The dagfinn is tough-minded and tenderhearted, while the follesdal is soft-minded and hard-hearted; travelling together in symbiotic pairs, they are the only intelligent creatures at home in deep waters.3

Although amusing, the description is misleading. There is a remarkable unity to Dagfinn’s work, which in fact is a good illustration of the QuineDuhem thesis. His writings and theories face the tribunal of experience as a whole. This holistic nature of this thinking can be quite unnerving. I’ve often had the experience of discussing with Dagfinn something he has 1

Elster (1978). Føllesdal (1958). 3 http://www.philosophicallexicon.com/ 2

7 written and asked for a clarification of this or that point. It usually turns out that this very point is developed at some length in another piece he has written. When I turned to that piece, there might also be statements in need of clarification, which, to be sure, is provided somewhere else. The upshot is that you need more or less to grasp Dagfinn’s view of everything in order to understand his stand on anything. Let me state the point in a slightly different way. There is a “Mount Dagfinn”. You can approach it from different directions, which provide different perspectives. To use some recklessly mixed metaphors, from one angle he may look like a duck, from another like a rabbit; from one perspective he may indeed look like a shark, from another like a dolphin. But they are all the same animal, the same mountain. If we could adopt a view from nowhere, we might be able to characterize it exhaustively and precisely. But of course there is no such view. The holism of his thought is also linked, I believe, to the charity he displays in trying to understand the writings of others. I have heard him talk, for instance, about French philosophers who in my opinion were little better than charlatans, and find nuggets of sense in their writings. For Dagfinn, the principle of charity is not only something he adopts on general hermeneutic grounds. It is also related to a deep kindness in his nature that all those who know him have encountered. Yet leaving this aside, he is so aware of the underdetermination of meaning by the written word that he is reluctant, even if exposed to what others might think irrefutable evidence, to dismiss any views as nonsense. The philosophy department at the University of Oslo is only one of many Norwegian institutions that have benefited from his luminous intelligence, unfailingly good judgment and incredible capacity for work. He has been a key figure in the Norwegian Academy of Science and Letters, and for many years coordinated a forum where academics and parliamentarians could meet to discuss matters of public policy at a more thoughtful level than is possible in the usual political settings. He has taken on these tasks out of a sense of public duty that, together with the personal kindness that I mentioned, makes him such an admirable human being. He has performed them, moreover, in what one might call a spirit of selffulfilling idealism. By refusing to acknowledge certain imperfections of

8 human nature, he has caused his interlocutors and interaction partners to rise above their normal level. I want to conclude by some remarks about our shared interests, notably with regard to rationality and to literary interpretation. The fact that his writings on these topics are not those for which he is best known constitutes in itself a reason for drawing attention to them. In his writings on rationality and its role in the explanation of action Dagfinn has made a number of important contributions.4 I want particularly to mention three of them, which have not got the attention from social scientists that they deserve. The first is the idea that rationality is a norm. Human beings internalize the requirements of rationality because they want to be rational. We take no pride in our occasional selfdeception or weakness of will. On the contrary, if we fall victim to these tendencies and become aware of it, we deplore the fact and try to avoid similar behavior in the future. An important implication of the idea of rationality as a norm is therefore that it offers a built-in limitation on the current and very exciting work that is going on in behavioral economics. This literature shows how human beings deviate from the prescriptions and predictions of rational-choice theory in a very large and rapidly increasing number of ways. To take an example at random, subjects in experiments are notoriously bad at figuring out the logic of backward induction in sequential bargaining – not because it’s difficult, but because it doesn’t come naturally.5 Yet once the logic is explained to them, they understand that it makes sense and apply it correctly. The second point can be stated in terms of the standard paradigm of preferences and a set of feasible actions. The rational agent chooses the option in his feasible set that is top-ranked according to his preferences. In many cases, as when we are trying to explain consumer behavior, this is a useful working hypothesis. In other cases, as Dagfinn has pointed out, the value of the approach is limited by the implicit assumption that the feasible set is given. Because of the subjective nature of rational choice, the “feasible set” can only refer to the set of actions that the agent believes to be feasible. Although the objective set of feasible actions may be given in 4 5

See notably Føllesdal (1982). Johnson et al. (2002).

9 some absolute sense, what is seen as subjectively possible depends, among other things, on the agent’s imagination and creativity. Rational-choice theory, therefore, is severely limited in situations where creativity matters. This explains, for instance, the failure of once-popular theories of technical change that relied not only on the objective existence but also on the subjective reality of an “innovation possibility frontier” that constrains the new inventions that can be made.6 The third point I’d like to mention relates to decision-making under uncertainty. In an article on ethical aspects of recombinant DNA research, written at a time when there was a great deal of uncertainty about the risks of such research, Dagfinn made the general point that the larger the number of competing theories, the larger the probability that they are all false.7 The existence of many different theories may be a sign that their common object is a part of the universe that we simply do not understand very well. It may then be rational or prudent to assume that something even worse may happen than what is predicted by any of the theories, or that the probability that the worst will happen is even greater than predicted by any of them. (In the memorable words of former Secretary Rumsfeld, we sometimes have to pay attention to the “unknown unknowns”.) In the current debates over modeling climate change, this fundamental point is often overlooked. The theory of rational choice is an area where Dagfinn’s work has very much influenced my own. The theory of literary interpretation is one in which, if my memory serves me well, we had the same idea more or less at the same time, independently.8 The idea is simple. The “hermeneutic method” does not rest on mysterious operations of empathy or intuition. Rather, as Dagfinn writes, “the hermeneutic method is nothing but the hypothetico-deductive method applied to meaningful material” such as texts, works of arts, and actions. Dagfinn has spelled this out, for example, in a wonderful analysis of five different interpretations of the Stranger, or the Strange Passenger, in Ibsen’s play Peer Gynt. His work in this area has

6

For an exposition of this idea and some objections to it, see Elster (1983), p.104-5. Føllesdal (1979 b). 8 Føllesdal (1979 a). 7

10 been highly influential, as reflected for instance in a recent book by Chris Mantzavinos, Naturalistic Hermeneutics.9 I would like to take the occasion, however, to challenge Dagfinn on this point. One of the places where he develops his views on interpretation is in a textbook he wrote on philosophy of science to which I also contributed a few chapters. Written in Norwegian, it’s been published in German10 and translated but not yet published in English. It deserves close attention because it is perhaps the one piece of writing that gives the best overall view of “Mount Dagfinn”, although on specific points one has to go to the more specialized articles. In the book, he sharply distinguishes interpretation of a work of literature from the analysis of authorial intentions. A satisfactory interpretation, he writes, has to fit the text. In addition, he continues, “we commonly interpret and judge literary works on the basis of the preconceptions we believe were those of the public for whom the author was writing”. I would argue that we should rewrite as follows: “on the basis of the preconceptions we believe the author believed were those of the public”, since the actual preconceptions of the public would presumably be irrelevant for the interpretation if the author was wrong about his readers, just as the actual constraints on innovation are irrelevant if the innovator is unaware of them. In other words, I claim that authorial intentions and beliefs are central to interpretation itself.11

9

Mantzavinos (2005). Føllesdal, Walløe and Elster (1988). 11 Elster (2009). 10

11 References Elster, J. (1978), Logic and Society, Chichester: Wiley Elster, J. (1983), Explaining Technical Change, Cambridge University Press Elster, J. (2009), “Interpretation and rational choice”, Rationality and Society 21, 533. Føllesdal, D. (1958), Husserl und Frege, Oslo: Aschehoug. Føllesdal, D. (1979 a), “Hermeneutics and the hypothetico-deductive method”, Dialectica 33, 319-36. Føllesdal, D. (1979 b), “Some ethical aspects of recombinant DNA research”, Social Science Information 18, 401-20. Føllesdal, D. (1982), “The status of rationality assumptions in interpretation and the explanation of action”, Dialectica 36, 301-16. Føllesdal, D., Walløe, L. and Elster, J. (1988), Rationale Argumentation, Berlin: Gruyter. Johnson, E. et al. (2002), “Detecting failures of backward induction”, Journal of Economic Theory 104, 16-47. Mantzavinos, C. (2005), Naturalistic Hermeneutics, Cambridge University Press.

Reference, neither Causal nor Apostolic, but Normative Dagfinn Føllesdal∗

Abstract Frege’s approach to semantics was dominant for several generations. It gave a unified, very elegant picture of singular terms, general terms and sentences working in basically the same way, through sense and reference. It was followed up and developed further by Carnap, Church and many others. In Word and Object (1960) Quine used basically this kind of semantics to show that quantification into modal contexts would lead to a collapse of modal distinctions: everything that is true is necessarily true. An examination of Quine’s argument showed that it was too catastrophic, and that this reflected a serious flaw in Frege’s unified semantics. Frege’s “one-sorted” semantics had to be replaced by a “two-sorted” one, where names and other singular terms have a very different semantics from general terms and sentences. This new approach to names and other singular terms is sketched briefly and differentiated from “causal” theories of reference (Evans) and transmission theories (Geach/Kripke). The paper relates to Perry’s discussion of the “slingshot” and of sequestered information in his essay in this volume. Also Linnebo’s ideas in his essay on the distinction between semantics and meta-semantics helps to elucidate the issue.

In this paper I will discuss some arguments about reference, mainly stemming from my 1961 dissertation, Referential Opacity and Modal Logic, but with a more detailed examination of some issues that I did not go into in my dissertation. Frege on sense and reference Frege contributed more than anybody else to developing semantics, primarily in “Über Sinn und Bedeutung” (1892) and “Der Gedanke”

∗

Stanford University and Universitetet i Oslo

14 (1918), but also in many of his other writings.1 Four of his observations will be particularly important for what follows: 1. Frege distinguishes between sense (Sinn) and reference (Bedeutung) of expressions. This distinction is partly motivated by trying to account for three phenomena: How can identity statements be both true and informative? How can empty names be meaningful? How do we determine what a name refers to? 2. A unified semantics: there are three basic kinds of expressions, singular terms, general terms and sentences, which all function in the same way: they have a sense and this sense determines their reference – in a complicated interplay where also material factors play a role. According to Frege, sentences refer to truth values, general terms to concepts, and singular terms to objects. It has become common to treat general terms as referring to their extension, that is the class of objects of which the term is true, and we will do so in what follows. This is not required for the arguments that follow, but it will simplify our discussion. Carnap’s method of intension and extension basically functions in the same way. 3. Universal substitutivity: expressions with the same reference may be substituted for one another without changing the reference of the whole expression wherein the substitution takes place. The same holds for expressions with the same sense, they may be substituted for one another without changing the sense of the total expression. (This must be read in conjunction with the next point, in some contexts expressions refer to their ordinary sense.) 4. Opaque contexts: Within some contexts, expressions refer to their normal sense, rather than to their normal reference. Frege makes use of this in order to deal with modalities. Later we will come back to his proposal.

1

A very good discussion of Frege’s semantic views is found in Künne (2010).

15

Quine We will compare and contrast some of Frege’s views with those of Quine and we will see that there are some important points where neither of them can be right. Quine criticized a notion of sense that gave no identity criteria for sense. ”No entity without identity” he said in 1957 and he insisted on this point throughout the rest of his life.2 According to Quine, a satisfactory notion of sense has to tell us how expressions get their sense and how we grasp these senses. This has to take account of the social nature of language. In Word and Object and later work Quine took some steps in this direction, but much, much more remains to be done. Quine was more satisfied with the theory of reference, including a theory of truth. Carnap argued that necessity and analyticity are closely connected: a statement is necessary if and only if it is analytic. This gave Quine an opportunity to strengthen his criticism of meaning and analyticity. From 1941 on, he used his views on reference to criticize the modal notions and through them the notions of analyticity and meaning. Quine presented a number of arguments against the modalities. In the beginning he pointed to their lack of clarity: necessity can be defined in terms of possibility and both notions in terms of possible worlds. However, possible worlds in turn can hardly be defined without appeal to possibility and thereby we are back in a small circle again, similar to the one that arises in the theory of meaning when we define synonymy and analyticity in terms of meaning and the latter term again in terms of the former. In both cases one moves in a small circle that should satisfy no 2

Quine first used this famous phrase in his presidential address to the Eastern Division of the American Philosophical Association in December 1957 [Quine (1957)], reprinted in Ontological Relativity and Other Essays [Quine (1969)], where the passage occurs on page 23. The idea is there much earlier. Thus in "On what there is" (1948), Quine writes: "But what sense can be found in talking of entities which cannot meaningfully be said to be identical with themselves and distinct from one another?” [Page 4 in the reprint in Quine (1953)]. Quine used the phrase again in many later writings, for example in Theories and Things [Quine (1981)], p. 102, and in From Stimulus to Science [Quine (1995)], pp. 40 and 75.

16

philosopher. Carnap and others who defined necessity in terms of analyticity connected the two circles, but still we just get interdefinability between a handful of words. The supposed tie between modality and meaning enabled Quine to reinforce his criticism of meaning and analyticity by pointing to ontological obscurities in modal logic, particularly quantified modal logic. These arguments brought out a severe lack of clarity about what the objects are that we talk about in modal contexts. Quine took this as an indication that there is something fundamentally wrong with the modal notions, like necessity and possibility, and thereby also with meaning, synonymy and analyticity. We shall not review these arguments here, but will mention the last of them, which Quine presented in Word and Object 1960. Quine here gave an argument to the effect that modal distinctions collapse when we quantify into modal contexts. That is, not only is everything that is necessary, true, but also conversely, everything that is true, is necessary. So there was no point in the modal notions. Carnap and others had held that without such quantification modal notions would be of no use. Quine therefore felt that he had clinched his case after nineteen years of criticism of the modal notions. Quine’s argument was a variant of what Barwise and Perry later called “the slingshot”3 since it achieves great results with meager means. Perry, in his contribution to this symposium, discusses the slingshot in more detail. It had been used by others before, notably Alonzo Church,4 and also later, by Davidson,5 and was generally accepted at the time. The only exception seems to have been Kurt Gödel. In his contribution to the Library of Living Philosophers volume on Russell, he used the argument: For if we admit the further apparently obvious axiom, that the signification of a composite expression, containing constituents which have themselves a signification, depends only on the signification of these constituents (not on the

3

Barwise, J. and J. Perry (1981). See John Perry’s contribution to this symposium. Church, A. (1943), Church, A. (1956), pp. 24-25. 5 Davidson, D. (1967). 4

17

manner on which this signification is expressed), then it follows that the sentence “Scott is the author of Waverley” signifies the same thing as “Scott is Scott”; and this again leads almost inevitably to the conclusion that all true sentences have the same signification (as well as all false ones).6

He then wrote: I cannot help feeling that the problem raised by Frege’s puzzling conclusion has only been evaded by Russell’s theory of descriptions and that there is something behind it which is not yet completely understood.7

Quine’s argument was disastrous for the philosophical approach to the modal notions at that time. Formalizing it shows that all assumptions were universally shared by Carnap and everybody else writing on the modalities. So there must be something wrong with the argument. Church had accepted a similar argument, but he avoided the collapse for the notions of necessity and possibility by developing Frege’s idea that in modal contexts expressions refer to their normal sense and not to their normal reference. In my dissertation I have an appendix on Church’s proposal where I show that although it contains quantifiers and what looks like modal constructions it avoids the collapse by treating these constructions in a purely extensional way. All constructions in Church’s system are referentially transparent, as they have to be in order for quantification to make sense. However, they are also extensionally transparent. It is therefore misleading to regard Church’s system as a modal system. He himself called it “the logic of sense and denotation.” Frege and Church were able to deal with the logical, so-called “alethic” modalities by help of Frege’s idea of a reference shift, whereby inside modal contexts of this kind, expressions refer to their ordinary sense. However, Quine’s argument is too catastrophic to be overcome by 6

Gödel, K. (1944). Gödel uses ‘signification’ and ‘signify’ for Frege’s ’Bedeutung’ and ’bedeuten’. Charles Parsons has written a very good introductory note to this paper in the Collected Works edition, where he discusses these issues on page 104. 7 Ibid., p. 130 = Collected Works, p. 123.

18

Frege and Church’s reference shift. An examination of the argument showed that it leads to a similar collapse of all non-extensional notions, not only the logical modalities, but also constructions that involve causality, probability, knowledge and belief and legal and ethical notions. Frege’s idea of a reference shift does not readily carry over to this kind of constructions. We would not like to associate with each linguistic expression a wide variety of entities to which the expression refers in nonextensional contexts, modeled on the way expressions refer to senses in contexts of logical modality. An analysis of Quine’s argument indicated that the only feasible way of undercutting it was to give up the traditional Fregean view on semantics, the view that singular terms, general terms and sentences all have basically the same kind of semantics. The analysis showed that it can be undercut by a ”two-sorted semantics”: singular terms behave quite differently from general terms and sentences. This leads us to the so-called “New theory of reference,” which we shall now discuss. ”New theory of reference” The basic idea in the new theory of reference is the following: Referring expressions, typically proper names, are ”rigid.” (This apt expression was introduced by Saul Kripke in his writings on this subject in 1971-728. In my 1961 dissertation I called them “genuine singular terms.”) They refer to the same object in all different circumstances, or as one would say it in modal logic, in all “possible worlds.” The fact that names and other referring expressions have a radically different semantics from general terms and sentences, reflects that objects play a very special and crucial role in our daily lives and in our communication with one another. They are important for our gathering of knowledge and for our actions. This gives them a special place in semantic theory: One should expect that keeping track of objects is so important that it has given rise to an important feature of language. We shall now look at the emergence of reference and how it gives rise to this special non-Fregean feature of referring expressions. 8

Kripke, S. (1971) and (1972).

19

Three features of objects Objects have three features that are crucial for reference: 1. “Transcendence” First: They are the bearers of a (usually) large number of properties and relations. Normally we know only a small number of these, but the object is conceived of as having numerous further properties that we do not know yet, but which are there to be explored. They transcend our knowledge, to use Husserl’s phrase. This aspect of objects plays a central role in Husserl’s phenomenology, but we shall not go into this here. 2. Change Secondly: Objects, except mathematical ones and a few others, change over time: One and the same object can have a property at one time and lack it at another time. The object remains identical through changes. Modalities come in at this point; not only are there the actual changes, there are also possible ones, there are accidents and there are necessities. Or, at least, so we say. 3. Fallibility Finally: There is our fallibility. We may have false beliefs about objects. We may seek to correct these beliefs, but all the while our beliefs, true or false, are of the objects in question. A belief, or set of beliefs, is not about whichever object happens best to satisfy our beliefs. A semantics that would just seek to maximize our set of true beliefs would reflect poorly the role that objects play in epistemology. Quine on reification Starting in Word and Object Quine discussed some features of the role of objects in our lives. In the 1990s this became one of his main concerns. Here is an example, from 1995: As Donald Campbell puts it, reification of bodies is innate in man and the other higher animals. I agree, subject to a qualifying adjective: perceptual reification

20

(1983). I reserve ‘full reification’ and ‘full reference’ for the sophisticated stage where the identity of a body from one time to another can be queried and affirmed or conjectured or denied independently of exact resemblance. [Distinct bodies may look alike, and an identical object may change its aspect.] Such [discriminations and] identifications depend on our elaborate theory of space, time and of unobserved trajectories of bodies between observations.9

This distinction between perceptual reification and full reification Quine made already in 1990, although without this terminology. Here is a passage from a lecture Quine gave that year: I wonder whether a dog ever gets beyond this stage. He recognizes and distinguishes recurrent people, but this is a qualitative matter of scent. Our sophisticated concept of recurrent objects, qualitatively indistinguishable but nevertheless distinct, involves our elaborate schematism of intersecting trajectories in three-dimensional space, out of sight, trajectories traversed with the elapse of time. These concepts of space and time, or the associated linguistic devices, are further requisites on the way to substantial cognition.10

Three reasons for using names If I am right in what I have been saying so far, names are normally introduced for the following three purposes: (i) When we are interested in further features of the object beyond those that were mentioned in the original description that was used to draw our attention to the object. (ii) When we want to follow the object through changes. (iii) When we are aware of our fallibility and want to refer to an object without being certain that we are right about its properties.

9

Quine, W.V. (1995), p. 350. The italics are Quine's. Quine, W.V. (1990), page 21 of the manuscript. See also Quine, W.V. (1995), pp. 35-40.

10

21

Given that objects play an important role in our attempts to explore and cope with the world, and given that objects have the features I have listed, we should expect these features to be reflected in our language. We should expect a language to have a category of expressions that are especially designed to refer to these objects and stay with them through all these changes that they and our beliefs about them undergo. This is just what genuine singular terms, or rigid designators, are supposed to do. These terms are hence inseparably tied up with the notions of change and fallibility and not just with the modal notions. Genuine singular terms comprise, in addition to proper names, the variables of quantification and correspondingly the pronouns of ordinary language, which within each context of use stick to the same reference. Also indexicals and demonstratives are in this sense genuine singular terms. These latter expressions, like ‘this’, ‘she’, ‘here’ and ‘now’, are used at numerous occasions with ever new references. However, at each occasion of use, the expression is used in order to refer to a specific object, and within that particular situation it is used as other genuine singular terms to keep on referring to that object. Within that situation these terms function like other genuine singular terms, they signal to the listener that we now intend to keep on referring to the same object. Given our concern with objects and other constancies in the world, we should in fact expect a lot of expressions to have this feature. For example, we should expect this to be the case for mass terms, natural kind terms, properties, etc., as has been pointed out by Kripke and Putnam. Even terms that refer to events will have these features. Events, in spite of their sometimes short duration, are objects that we often want to say several things about, and find out more about. However, one must be careful not to believe that all or most terms are rigid. Terms that are not purporting to refer to objects cannot be expected to have this kind of stability. There is, for example, no reason to expect that general terms will keep their extension from one possible world to the next. On the contrary, if they did, then again modal distinctions would collapse. It is crucial that our semantics be two-sorted, we need expressions that keep their reference, but we also need expressions that change their extension.

22

Ever since Bertrand Russell in 1905 proposed that proper names could be construed as definite descriptions in disguise, many have held that the main role for names is to serve as shorthand for descriptions. Names save us from repeating the whole description. They could be called ‘names of laziness’, just as Geach talked about ‘pronouns of laziness’. There does not seem to be any other role for names for the Russellian. However, I think that names, like pronouns, are not usually introduced for reasons of laziness. They are introduced for the three reasons I mentioned above, they signal that we are interested in future features of the objects, we want to keep track of it through changes and we know that we may be mistaken about the various features of the object. If we are not moved by any of these three reasons we will normally not introduce names or pronouns. Two examples Let me illustrate this by two examples. Compare the two descriptions: the balance of my bank account the person with the glasses Here, the first of the descriptions may be by far the most frequently used. On the “laziness” view it would therefore be likely to be replaced by a name. However, I doubt that any of you have ever introduced a name instead of this description, which you use so often. The reason is that you are not interested in any other feature of this number than that it happens to be the present value of your bank account. If you were particularly interested in other features of that number, perhaps because you are a number theoretician, you would switch to using a numeral. However, the second description, ‘the person with glasses’, or others like it, you may have used occasionally in order to pick out an object or a person. If you want to say something more about that person, regardless of whether the person keeps the glasses on or not, then you introduce the person’s name, if you know it, or you may use a pronoun, like ‘he’ or ‘she’.

23

In fact, as I noted earlier, pronouns, demonstratives, indexicals, and variables of quantification are genuine singular terms, and they may help us to see how genuine terms work. These terms are typically introduced in order to refer to an object, as we just did in the example with the person with the glasses. As soon as we are confident that our listener has picked out the right person, we start referring to this person by ‘she’, or ‘he’, as the case may be, and we then use this pronoun to refer to this person, regardless of whether she is carrying glasses or not, whether she disappears from sight, etc., as long as the situation and circumstances make it likely that all participants in the conversation keep on referring to this person by ‘she’. As John Perry points out in his contribution to this symposium, the information we made use of when we picked her out is sequestered during our conversation, we keep on referring to the same person even if the original information we used to pick her out, no longer applies to her or to her uniquely; there is no call for new evaluation of what the reference is every time the name is used. The situation is similar with the other genuine singular terms I mentioned: demonstratives, indexicals, and variables. Although their behavior differs in important ways, they all serve the same purpose: they introduce an object, which they then keep on referring to, normally within the whole rest of the conversation. Proper names are similar, except that they are introduced in order to keep the same reference permanently, not just within the confines of a short conversation. All these genuine singular terms, pronouns, variables, etc. and also proper names, have a sense when we use them to pick out their reference. Once their reference has been fixed, their sense will gradually change. My view is that their sense will be steadily modified in such a way as to help us keep track of the object. These modifications happen within a framework of individuation. In order to catch on to the use of genuine singular terms we have to have a background conception of individuation, so that we know what the objects are that we are aiming to keep track of. The person with the glasses for whom we introduced the pronoun ‘she’ is an example. In order to understand and use personal pronouns we need a conception of personal identity, and also of different genders.

24

In my view, the notion of a genuine singular term is not fundamentally a modal notion; it is not a notion that requires appeal to necessity or essences for its definition or clarification. There is no need to use the modal jargon of “possible worlds” in our definition of genuine singular terms. What defines them, is that they signal that we aim to refer to the same object in all the situations where we use the term, not just through modal changes. Preservation of reference is basic to our use of language far outside modal logic, for example in connection with our talk of an object reappearing or as changing over time. The tie between word and object, causal and transmission views I have argued that we use genuine singular terms to signal our intention to track some particular object through changes etc. This view is normally taken to mean that genuine singular terms are guaranteed to keep their reference. However, the argument for this view in my dissertation is a conditional argument: it says that if quantification is to make sense, then one has to keep on referring to the same objects “in all possible worlds.” There is no guarantee that one succeeds in this, and the argument says that if one gets confused about the references of one’s terms, so that they sometimes relate to one object, sometimes to another, then quantification ceases to make sense. If genuine singular terms were guaranteed to keep their reference, then appeal to a causal tie which connects the term with its object would seem an attractive solution. The tie would hold regardless of confusions and mistakes. Gareth Evans was, I believe, the first to propose such a causal theory of reference.11 The same guaranteed preservation of reference is characteristic of many versions of the transmission theory, which was proposed by Peter Geach and Saul Kripke. Geach is very clear on this. He wrote in 1969: I do indeed think that for the use of a word as a proper name there must in the first instance be someone acquainted with the object named. But language is an institution, a tradition; and the use of a given name for a given object, like other 11

Evans, G. (1973)

25

features of language, can be handed down from one generation to another; the acquaintance required for the use of a proper name may be mediate, not immediate. Plato knew Socrates, and Aristotle knew Plato, and Theophrastus knew Aristotle, and so on in apostolic succession down to our own times; that is why we can legitimately use ‘Socrates’ as a name the way we do. It is not our knowledge of the chain that validates our use, but the existence of such a chain, just as according to Catholic doctrine a man is a true bishop if there is in fact a chain of consecrations going back to the apostles, not if we know that there is.12

However, I regard these accounts as unsatisfactory, for several reasons that I will not discuss here, one of them being their inability to account for reference change. A normative view on reference As Linnebo points out in his contribution to this symposium, two questions have to be distinguished with regard to singular terms and reference: one is the general question whether singular terms behave semantically like general terms and sentences and change their reference from occasion to occasion just as general terms and sentences change their extension and truth value. Here my answer is no. Singular terms are radically different from general terms and sentences in this respect. They stick with their reference. So the old one-sorted semantics has to be replaced with a twosorted one. The main division goes between what I have called genuine singular terms on the one side and general terms and sentences on the other. Definite descriptions, in their normal uses, are classified together with the general terms and sentences. They behave like general terms; they just happen to fit one object. For this reason I will sometimes for simplicity use ‘singular term’ instead of the longer ‘genuine singular term’. Linnebo’s second question arises from this and is more particular: How are these singular terms connected with their reference? I will end by sketching my own view on the tie between genuine singular terms and their objects. Adherents to causal theories of reference and transmission theories of reference of the kind described by Geach tend to regard the tie 12

Geach, P. (1969), My italics.

26

between singular terms and their objects as a matter of ontology. According to them, we must distinguish between the ontological issue of what a name as a matter of fact refers to and the epistemological issue of how we find out what it refers to. I look upon the ontological and the epistemological issue as much more closely intertwined. This is largely because language is a social institution. What our names refer to – and not only how we find out what they refer to – depends upon evidence that is publicly available in situations where people learn and use language. Preservation of reference is not something that is achieved through the introduction of a genuine singular term in our language. Sameness of reference is never guaranteed. There is always a risk that in spite of the best of our efforts, we get mixed up, and if the mix-up is persistent and community-wide, the term may even come to change its reference. A name does not always continue to refer to the object on which it was bestowed at the original “baptism”. Rigidity, or genuineness, as I see it, is not incompatible with such a reference shift. Instead, I look upon rigidity as an ideal, something like a Kantian regulative idea, which prescribes the way we use language to speak about the world. When we use a name or another genuine singular term, we signal to our listener that we want to refer to a particular object and will do our best to keep on referring to that same object as long as we use the term. There is in our use of names and other genuine singular terms a normative pull towards always doing our best to keep track of the reference and keep on referring to it. Sometimes we go wrong, it then is unclear both what we believe, and what our beliefs are about until a new usage has been established. All our talk about change, about causation, ethics, and knowledge and belief, as well as about the other modalities, presupposes that we can keep our singular terms referring to the same objects. To the extent that we fail, these notions become incoherent. To conclude this brief discussion of reference: I hold that there are genuine singular terms, or rigid designators, in our language and that they are indispensable for our talk about change, causality, modality, etc. However, my view differs from other current views on reference mainly in that I do not regard preservation of reference as automatically achieved

27

through our use of singular terms, but as something we try to achieve. This is what I mean by ‘normative pull’ and also by what I call a ‘regulative idea.’ This view on reference could therefore also be called a normative view on reference.

References Barwise, J. and J. Perry (1981). “Semantic Innocence and Uncompromising Situations.” Midwest Studies in Philosophy, 6, 387–403. Church, A. (1943) “Carnap’s Introduction to Semantics.” The Philosophical Review 52, 298-304. -“-

(1956) Introduction to Mathematical Logic. Princeton: Princeton University Press.

Davidson, D. (1967). “Truth and Meaning.” Synthese 17, 304-322. Evans, G. (1973). “The Causal Theory of Names.” Aristotelian Society Supplementary Volume xlvii, 187-208. Follesdal, D. (1961). Referential Opacity and Modal Logic. Harvard dissertation 1961. Oslo: Oslo University Press, 1966. London: Routledge, 2004. Frege, G. (1892). “Über Sinn und Bedeutung.” Zeitschrift für Philosophie und philosophische Kritik, 100, pp. 25–50. Reprinted in Frege (1962), pp. 40–65, and Frege (1967), pp. 143–162. Black, M. (trans.), 1948: “Sense and Reference,” The Philosophical Review 57, pp. 207–230, and many later translations. -“-

(1918). “Der Gedanke. Eine Logische Untersuchung.“ in Beiträge zur Philosophie des deutschen Idealismus I (1918–1919), pp. 58–77. Reprinted in Frege (1966), pp. 30-35, Frege (1967), pp. 342-362, and in Künne, pp. 40–65. Quinton, A., and Quinton, M. (trans.), 1956, “The Thought: A Logical Enquiry”, Mind, 65, pp. 289–311, and many later translations.

-“-

(1962). Funktion, Begriff, Bedeutung: Fünf logische Studien. Patzig, G. (ed.). Göttingen: Vandenhoeck & Ruprecht.

28

-“-

(1966). Logische Untersuchungen. Patzig, G. (ed.). Göttingen: Vandenhoeck & Ruprecht.

-“-

(1967). Kleine Schriften. Angelelli, I. (ed.), Darmstadt: Wissenschaftliche Buchgesellschaft and Hildesheim: Olms, 1967, 1990.

Geach, P. (1969). “The Perils of Pauline.” Review of Metaphysics 23 (1969), pp. 287300. Reprinted in Geach (1972). -“- (1972). Logic Matters. Oxford: Blackwell. Gödel, K. (1944). “Russell’s mathematical logic.” Paul A. Schilpp, ed., The philosophy of Bertrand Russell. La Salle, Ill.: Open Court, 1944, pp. 128-129. Reprinted in Solomon Feferman and others, eds., Kurt Gödel: Collected Works, Vol. II. Oxford: Oxford University Press, 1990. Kripke, S. (1971). “Identity and Necessity.” In Identity and Individuation, edited by M. K. Munitz. New York: New York University Press, 1971. -“- (1972) “Naming and Necessity.” In Semantics of Natural Language, edited by D. Davidson and G. Harman. Boston, Dordrecht: Reidel, 1972. Also Cambridge, Mass.: Harvard University Press, 1980. Künne, W. (2010). Die Philosophische Logik Gottlob Freges. Frankfurt a.M.: Klostermann, Rote Reihe 30. Quine, W.V. (1948). “On what there is.” Review of Metaphysics 2, No. 5, pp. 21-38. Reprinted in Quine, W.V. (1953). -“-

(1953). From a Logical Point of View (Nine Logico-Philosophical Essays). Cambridge: Harvard University Press. Second edition 1961, 2. edition revised 1996.

-“-

(1957). Presidential address to the Eastern Division of the American Philosophical Association in December 1957, printed in the Association’s Proceedings and Addresses for 1958, reprinted in Quine 1969.

-“-

(1969). Ontological Relativity and Other Essays. New York: Columbia University Press.

29

-“- (1981). Theories and Things. Cambridge, Mass: Harvard University Press. -“- (1990). “From Stimulus to Science,” lecture at Lehigh University Oct. 15. -“- (1995). From Stimulus to Science. Cambridge, Mass.: Harvard University Press. -“-

(1995). “Reactions.” In Paolo Leonardi and Marco Santambrogio, eds., On Quine. Cambridge: Cambridge University Press.

Essays Part I Phenomenology

Neuropsychological Foundations of Phenomenology: Is It Possible? Patrick Suppes*

Introduction First, let me say that it is a pleasure to dedicate this lecture to Dagfinn Føllesdal. I have known him for four decades, and we have given a large number of seminars together at Stanford. If I know anything at all about phenomenology, it is entirely due to him. He has carefully explained to me many features of Husserl’s work. If you find that I have missed the point in various matters, do not blame him, but put it down to my being a poor listener, and at times a poor student. This last remark is not one of pseudohumble apology. In fact, I think that in the kind of matters that lie at the intersection of scientific psychology and neuroscience, on the one hand, and phenomenology on the other, I probably have the weakness of being too confident that whatever phenomenologists have to say I can ultimately find, for the parts that I am willing to accept, some reasonable scientific account. This lack of neutrality will shine through in what I have to say today. Yet, I will do my best to be quite explicit about those aspects of the scientific side of the ledger that are not well worked out and do not deal adequately with your own favorite phenomenological concept. I also understand that by concentrating on the work of Husserl, transmitted to me mainly by Dagfinn, not only in lectures, but in his wonderfully clear articles, I am taking a particular route into phenomenology that may be different from the one some of you cherish. Nevertheless, I hope what I have to say today will still contribute in a general manner to understanding the potential and the limitations of giving an account of phenomenological thought and concepts in scientific terms. I also want to stress what I consider a very positive aspect of phenomenology, in contrast especially with much modern philosophy of mind. This is the emphasis on the details of experience, and the recognition that *

Stanford University

34 how the subtleties of perception are dealt with is critical to any successful phenomenological theory. In a very different way, a similar emphasis is to be found in my two favorite classical philosophers of mind, namely, Aristotle and Hume. I will occasionally have something to say about them in comparison with Dagfinn’s clarifying version of Husserl. This lecture has three parts. First, I want to go through the instructive article about phenomenology that Dagfinn wrote for psychologists in 1974 for the Handbook of Perception. I stress this handbook is one that is written for psychologists, not philosophers. The first volume is on historical and philosophical roots of perception. I outline how I propose, from a neuropsychological standpoint, to analyze the most important specific concepts, such as that of noema. In the second part, I turn to a brief discussion, of two more general concepts of phenomenology that it seems to me are of central importance and also controversial. These are the concepts of intentionality and consciousness. Finally, in the third part, I turn to a general concept of meaning, one that is psychologically based, and, I think, relevant to phenomenology. I. Phenomenology à la Føllesdal with emphasis on perception Dagfinn begins with a discussion of intentionality, but I want to reserve that for later and to begin with one of the specific technical concepts of Husserl, namely, that of noema. Dagfinn initially quotes a brief statement from Husserl (1952, p. 89). The noema of a physical object is a “generalization of the notion of meaning to the realm of all acts”. This is certainly the kind of remark that needs a lot of unpacking, but Dagfinn says something immediately thereafter that provides the needed explanation. Just as the meaning of a linguistic expression determines which object the expression refers to, so the noema determines what the object of an act is—if the act has an object; some acts have a noema for which there is no corresponding object. (Føllesdal, 1974, p. 379)

Dagfinn next gives in an enlightening paragraph an example I will not quote, but summarize. The noema of a tree that a person is looking at is a complex structure, which consists of many features or aspects of the tree. Now, these features are not all of the features, but the ones that, as I understand it, are the kinds of features that are known to the perceiver. This does

35 not mean that in a given act of visual perception all of these features or properties are ones that can be immediately available to perception. A favorite example of Dagfinn’s, one that I have always liked, is the back side of a tree that we are looking at. It is part of the general structural view we hold of trees that they are not simply paintings on a wall but have mass and volume in three dimensions, and therefore, when we are looking at them from a given viewpoint there is also a back side because they are also not transparent. Dagfinn and Husserl emphasize that this structure is not an abstract structure simply given in the physical world, but is one related to the act of perceiving. I take this to mean that the structure is one that is built up in a complex fashion from individual experiences, and the experiences of others communicated to us, that fix the structure of a tree or whatever object we are considering. It does not mean that we include in the noema all of the physics we know about trees, such as the behavior of the particular atoms that make up the tree. One of the remarks I would make is that, as I understand it, a foundational account of the features of a tree that make up the structure or the noema is not something that is given in a mathematically exact form. Also, I would surmise that the structure that constitutes the noema of a physical object, such as a tree or a chair, can be different for different individuals and different cultures. What is important is that even though the noema is an abstract structure, it is not a particular structure at a given time and place. It is like the type of a token in linguistic usage. This remark takes us to the closely related concept of noesis. I do not think the parallel is exact but, as already suggested, the noesis of a structure stands to the noema as a token of a word stands to its type—the noema is abstract and timeless, the noesis temporal and concrete. The concrete, but strongly structural character of a noesis is, in many respects, close to the structural process neuroscientists are looking for to solve the binding problem. How does the brain put together nearly simultaneous perceptual input to associate this electromagnetic congeries of signals with the concept of seeing a tree? The psychologist J. J. Gibson, whom I knew fairly well, pushed hard for a concept of “direct” perception. Yet we can ask from a neuroscience perspective, “What could possibly be meant by direct reference?” By “direct” we usually think of x going to y without intermediate considerations, but if anything was ever obvious, no

36 act of perception is ever direct in the brain. Anyone who looks at the complexity of our auditory or visual systems, not to speak of the other senses, knows how ironic the description of perception as being direct is. It will be a long time before we have understood the intricate computational details of any one of these neural systems. It is rightly often said that, after the human brain itself, the next most complicated structure in the universe, as we know it, is the human visual system. These skeptical remarks apply, with little change, to philosophical theories of direct reference. The third concept of this trio is that of the hyle, which is immediately reminiscent of Aristotle’s ΰλη, his concept of matter. It seems to me useful to think of Husserl’s hyle as being close to Aristotle’s concept, even though differences can be found. For example, he certainly wants to introduce hyle in connection with acts, not simply in terms of the world as it is, independent of any act. But the remarks about it are close to saying that the hyle of an experience is the givenness of that experience, that which makes it, in its particular time and place, exactly what it is. It is the hyle that supports, so to speak, the inexhaustible features of the object toward which the act is directed. I will not try to put this other ways. Many of you can do a better job than I can. It is more important for me to make as transparent as I can my modification of these notions to put them within the scientific framework of psychology and neuroscience. First, let us begin with the concept of a noema, which is the easiest for my analysis. It is close to the ordinary concept of structure in modern mathematics and logic. But it is characteristic of the use of structures in any scientific theory that the concepts included in the structure are very far from offering a full description of the physical or psychological object or process being considered. Only certain limited features are actually brought in to a particular theory, but in phenomenological terms the structure is that of a determinable X. One way of thinking about this is that the neuropsychological scientific approach can in any given theoretical set up deal with only part of the noemata. What phenomenology supplies, in its concentration on the richness of experience, and its inexhaustible nature from a conceptual standpoint, is a vast collection of features that can be the focus of a given particular scientific theory. A well-worn but classic and easily understood example is that in standard forms of classical mechanics the color of objects plays no role and is not formulated directly as a physi-

37 cal concept, even though it is one of the most striking perceptual features of physical objects that we ordinarily encounter. So, my first point is that the structures that we use in scientific theories, where we are focused on certain properties or relations of objects, form substructures of various noemata. The substructures that we use in formulating about classical particle mechanics or something very different, electromagnetic theory, are substructures coherent with the richer phenomenological noemata of the experiences involved in either classical mechanics or the more subtle phenomena of electricity and magnetism. The second point is that for such identified substructures of a given scientific theory we then have well defined concepts of isomorphism for particular realizations of these concepts, especially for what I would now term substructures of noeses, that is, substructures that are particular realizations in some concrete place and time, of the particular scientific theory being considered. I also remark that because of the inexhaustible character, and our inability, with which I entirely agree, to express in a completely exhaustive way the nature of experience, there is no natural notion of isomorphism for noemata or noeses, because we are not able to list all the features that should be included in a possible isomorphism. However, it seems to me there is an essential difference between the structural isomorphism of noemata and noeses. A noema is an abstract structure with a natural equivalence relation that is, in fact, a congruence relation, as in algebra and pure geometry. A concrete noesis, on the other hand, has to characterize approximately equivalent perceptions, which is naturally a similarity relation, i.e., a relation that is reflexive and symmetric, but not necessarily transitive, or in some cases not even symmetric (Tversky, 1977). Such a notion of similarity is widely used in psychology, where the need for thresholds in recognizing the approximate equivalence of all kinds of signals and other phenomena in perception requires a generalization of equivalence to something weaker. Moreover, the extensive literature on these subjects, summarized in some detail in Suppes et al. (1989), includes a probabilistic version as well which goes back to the early work of Thurstone (1927a, b). A recent good analysis of the importance of such a concept of similarity in the early work of Carnap is that of Leitgeb (2007).

38 Finally, the hyle is for me the givenness in experience. It provides a conceptual basis for realizing that, no matter how detailed the isomorphism between two noeses of a given noema, there is still a difference, because of the uniqueness of the particular time and place of each noesis. Notice that this recognition of the given lies outside the formal structure of scientific theories, as ordinarily considered. What they have to say about this point is not very interesting. On the other hand, I emphasize that I am not stressing for one moment just the nature of scientific theory itself. It is when we turn to scientific experimentation that the notion of hyle has a proper place. A given experiment, confirming or disconfirming some aspect of a particular scientific theory, must take place at a given time and place with given equipment and given individuals performing the experiment—performing as I would want to say here, various particular acts. The particularity of the givenness of an experiment is what constitutes the hyle, and is an essential part of any serious account of science, that is, any account that necessarily deals with experiment as well as with theory. What I said about the given very much fits in with Aristotle’s concept of hyle, for matter provides the content of form. In the De Anima, for example, the psychological mechanisms for perception and higher intellectual functions are forms without matter or hyle. The forms, or noemata, are what are abstracted from the hyle. These forms are sensible forms in the case of perception and intelligible forms in the case of higher functions. A second point to stress about my own views of both scientific theory and experiment is that it is the experimental side of science that is closest to the phenomenological account of experience, for the openness and the inexhaustible character of the concrete acts needed in performing experiments are clearly recognizable. Any attempt to fully formalize experimentation is, for too many reasons to be enumerated here, a foolish undertaking. All experiments of any deep scientific merit contain relatively detailed accounts of what the experimenters did and, in fact, are written in phenomenological terms: what specific acts as experimenters did they perform, in what context and with what equipment? And that equipment itself was used and evaluated in terms that are not merely scientific, but also include the phenomenologically given.

39 We must not be misled by the fact that there are significant parts of experimentation that can be formalized. A good example is statistical design. In the case of medicine, where in many cases, fundamental scientific theories are often completely lacking, the most important aspect of the scientific method is the application of statistical-design principles. Even though these design principles can be stated in very sharp mathematical form, it does not mean the experimentation itself can be put into such mathematical form. The experimentation will continually be a matter of practice. The joke used to be at the Stanford Linear Accelerator Center (SLAC), that the theoretical physicists complained that the experimental physicists never understood what they were talking about; the experimental physicist complained that the engineers and technicians, who ran all of the equipment, never quite satisfied what they wanted; and the engineers and the technicians continually complained that the experimentalists would do very much better experiments if they only understood the equipment—and God help them if a theoretician was ever permitted to touch the equipment. This joke says something serious about scientific experimentation at a high level. It is very much a phenomenological kind of experience. Only a very feeble summary is given on the intricacies of equipment behavior in any modern physical experiments, in the publications reporting them. It is not possible to give what you might think of as a complete account. All we can do is summarize some of the features. A favorite resort in the case of complicated equipment is to give a reference to the manual produced by the manufacturer as to what its performance characteristics are, or should be. Hopefully, they match approximately what occurred in the experiment. Such scientific experimentation is the great large-scale phenomenological aspect of modern science. When I stand in my garden looking in any given direction, the perceptual detail I am aware of, not to speak of that which I am not, defies complete description. The necessary vocabulary to accurately describe the spatial patterns of shape, color, texture, mass and the like does not exist in any natural language. Even Proust, that master of perceptual detail, was crude in comparison with the swift but rich survey my eyes (and others’, of course, as well) can make. Perhaps an even better example is the richness of unspoken and indescribable visual, auditory, and sometimes tactile per-

40 ception used to control and manage our physical motions in any part of the natural world. Scientifically, we can focus on some small piece of the action, so to speak, but a full synoptic analysis is hopeless and a mistaken, as well as unachievable, scientific goal. The absence of a complete foundation does not mean that the relation between phenomenology, on the one hand, and psychology and neuroscience, on the other, is hostile or competing. But a recognition that, of course, it is a scientific task to keep enlarging the range and depth of the theories and experiments developed and confirmed as having some validity. And yet from any serious viewpoint of analysis on the nature of experience, the subject matter of phenomenology will remain inexhaustible and not be covered in any complete way by scientific results. Indeed, I am including scientific experiments as part of phenomenology, because of their essential, uneliminable givenness. I also want to stress that what I just said does not mean that I think phenomenology has some privileged access to knowledge that is not available in any scientific fashion, something that has often been claimed mistakenly in the history of philosophy. The openness of phenomenology and the acceptability of it in the modern world are based upon its splendid emphasis on the richness of experience and the recognition that our scientific endeavors will never be complete. In emphasizing this viewpoint phenomenologists need not transcend experience in any Kantian fashion by developing pure apriori synthetic concepts and propositions about nature, or a moral philosophy whose central concepts lie outside experience entirely. II. Intentionality and consciousness I focus on intentionality, with an afterword on consciousness. Intentionality. All animate matter is intentional. Why? Consider plants and bacteria. Their genetic evolution depends upon accidental and probabilistic mutations, as their main mechanism. Of course, the accidental mutations themselves are not individually intentional. It is, rather, a necessary feature of the selection process that a fundamental goal of all life, namely, survival of the species, is facilitated by this process. It is important to recognize that accidental means, themselves intrinsically nonintentional, can, and often do, contribute to the highest intellec-

41 tual goals, such as the development of a new theory for some part of physics or biology, not to speak of mathematics. It is empirically untenable, I would argue, on the basis of many famous incidents in the history of science to claim that new discoveries are reached by calm, deliberate and conceptually explicit chains of reasoning. Accidental and probabilistic associations, whose process cannot be brought to conscious awareness, are central to finding new results of any intellectual depth. I have expanded on these ideas in an earlier article (Suppes, 2003). But linking intentionality up and down the phylogenetic tree of species implies no claim that all kinds of intentionality are equal in the pantheon of nature. It is also a mistake to think that there is a sharp, identifiable cleavage between animate and inanimate matter, the intentional and nonintentional. The slope from the lofty peaks of human intentionality to the valleys of none is slippery, with no natural stopping point. Were the Australian stromatolites that flourished 3.4 billion years ago as perhaps the earliest form of life really to be classified as such (Allwood et al., 2006)? It surely does not deeply matter, for the historical creation of new organic molecules was a necessary preliminary, and wherever we draw what we may call the line of life in evolution, the differences for species close to the line, but on different sides will seem small in the context of the long and complex ascent to our species. The next step up from plants and bacteria is to the species that clearly demonstrate some form of elementary associative learning. Little worms, like Aplysia or C. Elegans, whose DNA is so surprisingly close to our own for long sequences of the universal four-letter alphabet, are good examples. Their genetic evolution has given them a repertoire of acts: approach stimuli favorable for finding food, withdraw from a range of stimuli that are noxious and deadly. Given this repertoire, they can now be conditioned to approach some new stimuli and avoid others, toward which they have no instinctive bias to approach or avoid. Such simple associative learning will modify their cell structures, but not in ways that will be genetically inherited. They can be taught under supervised, i.e., experimental, procedures to learn new intentional acts of approach and avoidance. In principle, such associative learning regimes are of the same general kind that are used to teach, in a super-

42 vised way, many things to children. And these same kinds of associations are used in an unsupervised way in scientific discoveries of all sorts. Many phenomenologists and cognitive psychologists will strongly disagree with the claim I have been making about the central role of associative processes for almost all species, high and low, but above those species that are incapable of association. I cannot take the time to argue this point in detail here, but I have done so in several other places, e.g., Suppes (2002). Consciousness. Like William James, Husserl has a very liberal use of the concept of consciousness. In many ways, a better term for both of them is “mental phenomena”. For example, when in the discussion of the transcendental reduction Husserl refers to directing one’s attention to the sphere of consciousness, I would, from a scientific psychological perspective, speak of the directedness of intention. Examples would then include many acts performed on the edge, or outside of, the edge of consciousness, but ones that only make sense in terms of their intentional directedness, not at all in terms of their description in ordinary physical language, for instance, my regular walking from one office to another in Ventura Hall at Stanford, where I have had my office for fifty years. There may be in texts of Husserl I do not know proper detailed reference to the intentional mental phenomena that are unconscious and dominate much of our mental activity. From a cognitive standpoint alone, I have in mind the unconscious associations so central to creative activity in both theoretical and practical endeavors, from the outward reaches of modern mathematics to the high art of reaching compromise in all serious political matters. To believe that conscious rational deliberation is central to any of these creative efforts is to have, in my view, a scientifically unsupportable view of human thinking processes. Now in William James’ case, where I know the texts well, I find it easy to modify his usage of the term “consciousness” to restrict it to the modern psychological sense of direct awareness. As in the past, I leave it to Dagfinn to instruct me on how I should reconstruct Husserl on this matter. There is another point about the directedness of intentions that I need to clarify. This is the issue of intentional acts that do not have a physical object or process as their focus, because of a cognitive mistake or a hallu-

43 cination, for instance. I agree with Husserl that this is a problem that must be dealt with. But I have no difficulty with it. From a neuroscientific standpoint, intention is not directly focused on external physical objects or processes, but on their brain representations. In other words, attention requires perception and perception requires brain images of what is perceived. There can be, in the process, quite naturally, responses of the neural system that we would externally count as errors of perception. Such as if acts are in fact much studied in scientific psychology, and the view mainly adopted seems compatible with Husserl’s. In cases even of direct awareness—consciousness in the strong sense I use—of the perception, the intentional sense of directedness, even in error, can remain very strong. (As I have gotten older, I am increasingly aware of this in the greater number of mistaken visual perceptions I have, even when in a state of very strong focused attention.) In a recent article (Føllesdal, 2006), Dagfinn gives a detailed analysis of Husserl’s concept of anticipations in experience, which can be characterized as features of the physical objects or processes we encounter. So when we see a car driving by its color and shape are anticipated features. Such anticipations are the constituents of the noematic sense of the noema, which following the Fregean tradition we may think of as the meaning of the noema. Dagfinn stresses that anticipations may be experienced consciously or unconsciously, and there are also many features of the noema of an object or process that we have not yet experienced, but potentially may do so in the future. He says that Husserl regards “anticipations as largely sedimentations from past experience. They might hence differ from person to person, dependent upon their personal history. The dynamics of this sedimentation process is a major theme in Husserl. An important subtheme is the adaptations that yield intersubjectivity.” (p. 378). This setup seems most congenial to my own scientific theorizing about how the brain works, as a vast dynamic adaptive associative network. More on this in the next section.

III. A psychological concept of meaning The meaning of a word or a feature of a noema is given by a subnetwork of associations activated when the word is heard or read, and this same pro-

44 cess also occurs when an object or process is perceived. Anticipations in Husserl’s sense are immediately activated in the network. This is not the place to offer a battery of details, but I stress two conceptual points. First, this account of meaning leaves no natural place for a sharp distinction between the analytic and synthetic. Second, it also leaves no place for a sharp distinction between the intentionally loaded meaning of a perceptual anticipation, for instance, of color or shape, and the meaning of a word or phrase. From an evolutionary standpoint, one of the most important aspects of language is surely our ability to use it to communicate anticipations to each other. So I challenge the idea that the fundamental concept of meaning of an expression or perception is abstract and external. Dagfinn has pointed out to me that people are sensitive to this use of abstract. I am also sensitive, because I think one of the characteristic problems in going from psychology to neuroscience is that there is not any clear sense of abstract that applies directly to the brain. Everything in the brain is physically given activity or anatomy. So what in neural activity can be properly and directly regarded as abstract? Well, I think abstract meaning must be a derivative notion. If we take the brain seriously we are not going to have any immediate sense of the abstract, even though, psychologically there is a tendency to talk about abstract concepts. But in order to talk about the abstract, as we ordinarily like to talk about it, we must have a sense of representation, and how are we to think of representations in the brain? There must be something concrete there. The brain has only purely physical processes. There must be something concrete to be the representation of that which is to be represented. For cognitive relevance, the abstract concept must have, in this neural sense, a concrete brain representation. So, for example, if I talk about the brain representation of a word, it will be something physical that is a representation, probably of the spoken structure of the word, and will have an approximate structural isomorphism to that structure. Given this neuropsychological viewpoint, it would be mistaken to hold that meanings are abstract and eternal objects. So I would find the phenomenological sense of meaning to be one with which I can be quite sympathetic. In addition, a point, as I understand it, of Husserl’s that I very much agree with is this. It is a mistake to think of meanings being restricted to language. Perception and experience generally are saturated with

45 meaning. If you ask me for the meaning of the word Paris, I would want to respond and I would want to say something about the meaning of the word. But for me the meaning of Paris, the city, is much richer than the meaning of the word. And I think this is related, but is related to a distinction that is too often sharp. I also insist on an evolutionary picture. The perceptual system is something that is millions of years old in mammals, and along with it is the development of Husserlian anticipations. In contrast, natural language, as spoken by humans, is a very recent development. The richness and depth of the apparatus that is used in perception, as a whole, is evident, compared to the relative newness and restricted character of language. I want to make a series of remarks that are focused on this contrast. I start with the neural phenomena of the very large area of association in the human brain and the proper emphasis on the centrality of association in the theory of mind, starting with Hume in his treatise (1739/1951) and followed up by an endless body of research since then. The crucial point about meaning for me is to use the concept of associative network to give the meaning dynamically of a word or a city, or anything else in my experience. This network is huge and continually changing. The next important concept is activation. Why do we need a notion of activation related to the notion of meaning? Because we have, each of us, such a large complex associative network. If when I said Paris all nodes and branches of your associative networks lighted up, your heads would look like light bulbs because there would be so much electrical activity. The critical point is that it would be impossible to compute anything seriously. And so what is important is that only a subnetwork is activated. What nodes are activated? Those with the strongest associations to the brain “image” of Paris, often also adjusted to the context in which the word is being used. It is this subnetwork of associations that we can think of as being the neural representation of meaning. I expand on aspects of these neural associative networks relevant to meaning. I begin by emphasizing the dynamical nature of meaning and belief. Over two meetings, one in California and now this one in Bern, I have had an extended discussion with Dag Prawitz about whether or not we can come to complete agreement on the nature of mathematical proof. It has been a wonderful extended discussion. My associations have been modi-

46 fied throughout these conversations with him and that means the meanings, and also my beliefs, have been continually modified dynamically. Dag is very persuasive and careful in arguing various points. So as we talk, I am continuing to modify internally my beliefs. One of the most important aspects that comes out of a neural scientific view of how the brain is working is that there is this continued modulation of associations that has in its own way a holistic effect. Changing the strength of one association changes the relative strength of others. Of course, the associations on which strongly held beliefs principally depend can only be changed by major changes in the strengths of some other key associations, which are unlikely in the ordinary course of events. You might ask, “Is there anything besides the associative network carrying meaning?” Well, the broad answer is negative, but it is very important to immediately emphasize what is additional. This is also a neuroscientific point even more than a psychological one. The great problem in the philosophy of mind, as expounded by modern philosophers, is the mind, as conceived by them, is not able to do anything. There is no doing. It is very quiet and reflective, meant to be deliberate in the best cases, but the actual way of doing anything is not laid out. If I give such a model of mind a question and it has to find an answer, no account is given in any detail of how this might be computed, or how the organization of the necessarily complicated reasoning in some cases, is to be accomplished. In another direction, given the necessity of driving a car or doing anything physically coordinated, how does the mind work, and how does the mind think about how it works in such temporally constrained cases? The standard philosophy of mind is missing the essential ingredients to handle any of these problems. In the case of the associative view of mind as fundamental, many cognitive scientists have asserted it is too elementary and too simple an idea, and some philosophers have seconded this view. I want to explain why this response to the claim that association is universal and computationally powerful is wrong. Well, my first rhetorical retort is the kind that one philosopher likes to say to another, “You just didn’t think about this matter hard enough.” I want to be more circuitous in my substantive answer. In many ways the single most fundamental result in logic in the twentieth century was discovering the set of computable functions and un-

47 derstanding how that set could be defined in multiple extensionally equivalent ways; above all, how any computable function can be computed. Equally important is the realization, that the apparatus for computing such function is extraordinarily simple. The product of the number of states and the number of symbols is a well-recognized measure of the simplicity of a universal Turing machine. Still one of the very best results is Marvin Minsky’s UTM’s product of twenty-eight, seven internal states and four symbols. Well, exactly the same thing can be said about associative networks. It is very easy to have an associative network mimic or represent any universal Turing machine. So the account I want to give of meaning is, I think compatible with what I hear phenomenologists saying, even though I have a very naturalistic view about both meaning and belief. Remember what marked belief for Hume, a certain feeling, a certain vividness. I used already the modern term for this, activation. The way we produce a current belief is by activating some subnetwork, which is activated usually by incoming stimuli. Prior to activation, beliefs are only potential, in a sense not too far from what Aristotle says of knowledge in the De Anima (417b5– 417b20). “We don’t have a little database, Ah, here’s belief 1 and here’s belief 2, here’s belief 3, let’s go look up belief number 300, 561.” It isn’t like that at all. Beliefs are dynamic, only computed at the last moment as they should. Now many, of course, will not change much at all. I am not saying, for example, that you, or anybody you know, is changing views about the flatness of the earth from moment to moment—nothing like that. But in the overall dynamic structure, belief changes, sometimes suddenly, and with it meaning, a point vividly illustrated by dramatists and novelists since ancient times, as a mirror of something that occurs continually in human relations.

References Allwood, A. C., M. R. Walter, B. S. Kamber, C.P. Marshall, and I. W. Burch (2006). Stromatolite reef from the Early Archaean era of Australia. Nature, 441: 714– 718. Aristotle (1975). De Anima (On the soul). Cambridge, MA: Harvard University Press, 4th edn. English translation by W. S. Hett. First published in 1936.

48 Føllesdal, D. (2006). Hintikka on phenomenology. In R. E. Auxier and L. E. Hahn (Eds.), The Philosophy of Jaakko Hintikka, pages 373–387. Peru, IL: Open Court Publishing Company. Føllesdal, D. (1974). Phenomenology. In E. C. Carterette and M. P. Friedman (Eds.), Handbook of perception, Volume I, Historical and philosophical roots of perception, pages 377–386. New York: Academic Press. Hume, D. (1739/1951). A treatise of human nature. London: John Noon. Quotations taken from L. A. Selby-Bigge’s edition, Oxford University Press, London, 1951. Husserl, E. (1952). Ideen zu einer reinen Phänomenologie und phänomenologischen Philosophie. Drittes Buch: Die Phänomenologie und die Fundamente der Wissenschaften. Ed. by M. Biemel (Husserliana V). The Hague: Nijhoff. Leitgeb, H. (2007). A new analysis of quasianalysis. Journal of Philosophical Logic, 36: 181–226 Suppes, P. (2002). Representation and Invariance of Scientific Structures. Stanford, CA: CSLI Publications. Suppes, P. (2003). Rationality, habits and freedom. In N. Dimitri, M. Basili and I. Gilboa (Eds.), Cognitive processes and economic behavior, pages 137–167. New York, NY: Routledge. Suppes, P., D. H. Krantz, R. D. Luce, and A. Tversky (1989). Foundations of measurement, vol. II: geometrical, threshold and probabilistic representations. New York: Academic Press. Thurstone, L. L. (1927a). A law of comparative judgment. Psychological Review, 34: 273–286. Thurstone, L. L. (1927b). Psychophysical analysis. American Journal of Psychology, 38: 368–389. Tversky, A. (1977). Features of similarity. Psychological Review, 84: 327–352

Consciousness, Modality, and Inner Awareness David Woodruff Smith*

Abstract The themes of intentionality and modality are prominent in the work of Dagfinn Føllesdal. Here I integrate them in an analysis of (self-) consciousness, specifically, the form of awareness we characteristically have of our own current experience. The structure of an act of consciousness may be explicated, I propose, in a phenomenological description that articulates inter alia the form of inner awareness of experience that is sometimes called self-consciousness. We distinguish, accordingly, certain forms of phenomenological structure in a simple visual experience: (i) The mode of presentation of the object of consciousness is reflected in the sense (Sinn) of the experience (per Husserl, Frege, Føllesdal). Indeed, the sense prescribes a particular object that can be re-identified in further experiences presenting the same object with different properties (per Husserl, Føllesdal, Smith and McIntyre). (ii) The presentation of the object is further modified or modalized by such characters as visuality, attentiveness, certitude, etc. (thetic characters, per Husserl). (iii) The modality of presentation in consciousness, on my analysis, includes not only such thetic characters, but also the specific characters of egocentricity, reflexivity, and phenomenality, as well as spatiotemporal locus. These modal characters, I propose, define the form of awareness the subject has of an act of consciousness. (Husserl speaks of modalities of judgment, as opposed to modalities of being. I want to appropriate the wide notion of intentional modalities introduced by Jaakko Hintikka, but I want to parse the modality of an experience into the several phenomenological characters indicated above.)

*

University of California, Irvine

50 THE MODEL: (SELF-) CONSCIOUSNESS AS MODALITY 1. Factoring Consciousness: the Modal Model of (Self-) Consciousness. What makes a mental act or experience conscious, according to a classical view, is one’s awareness of that experience as it transpires. When I consciously see some object, for instance, I am aware of seeing it: an awareness wherein my consciousness-of-the-object includes a consciousness-ofitself, and therewith a consciousness-of-myself. What is the form of that awareness? That is our problem. A variety of recent theories of consciousness have proposed that this awareness consists in some form of higher-order monitoring, where an experience either includes or is accompanied by some form of secondary mental activity that monitors the experience. This may be a kind of higherorder perception or thought or observation concerning the experience, as if the brain is doing two things at once, where the second activity monitors the first and thereby renders the first conscious. (See Kriegel and Williford eds. 2006 for analyses of the problems with these theories.) A significantly different theory of consciousness is that which I have called the modal model of (self-) consciousness. On this view, consciousness involves an inner awareness that is an integral part of the way an act of consciousness is performed, part of what I’ve called the modality of consciousness, i.e., the modality of presentation in an act of consciousness, as opposed to the mode of presentation of an object in the act. This model turns on a distinction among several basic forms or dimensions in the structure of consciousness. Accordingly, our task is to factor the structure of a typical act of consciousness. Specifically, we want to factor out those forms that define the type of awareness of experience, and therewith of self, that we find in our everyday consciousness. We distinguish two senses of “self-consciousness” here: consciousness of one’s consciousness-ofsomething, and consciousness of oneself in being conscious-of-something — conscience de soi and conscience de moi, to parody Sartre (albeit within a different framework). (This model of (self-) consciousness was proposed in Smith 1986 and is amplified in subsequent works cited under Smith in the bibliography below.)

51 On the modal model, the structure of a sample visual experience is articulated in a phenomenological description in the first person: < Phenomenally in this very experience I now here see this green frog I’m jumping over >

\________ modality of presentation _______/ \__ mode of presentation __/

The angle quotation marks are a device used to “quote” the content or noema of the experience described by the words between the marks. (Husserl used quotation marks in this way, as I have long emphasized: Smith 1970, 2007.) The underlinings indicate distinguishable formal slots or positions in the overall content of the experience, what Husserl might call “dimensions of variation” in the structure of the experience (noesis + sensory hyle), reflected in the structure of the noema of the experience. Thus: I might see not a frog, but a dog, and I might see not a green but a red frog; I might not see, but hear, the frog; I might see it not clearly, but vaguely; I might see it not focally, but peripherally; and so on. As indicated, we distinguish two basic formal parts of the content or structure of the experience: the mode of presentation of the object, and the modality of presentation. Our primary concern here is with elements in the modality of presentation. Importantly, to say an object is presented or represented in an act of consciousness is to say it is “intended” in a certain way: talk of presentation and talk of intentionality are equivalent. We begin, then, by marking out the indicated slots of phenomenological structure, returning to specific issues or problems below. On the proposed model, the phenomenological structure of the experience described is factored into several characters informed by certain elements of content: • phenomenal character: < Phenomenally … > • reflexive character of inner awareness: < … in this very experience … > • egocentric character of awareness (of self): < … I … > • temporal character of awareness: < … now … > • spatial character of awareness: < … here … > • species character: < … see … > • mode of presentation (sense) of object: < … this green frog I’m jumping over >

52

On this account, my inner awareness of my visual experience has the form of the structured modal character or content: < Phenomenally in this very experience I now here see … > . That character defines self-consciousness, or consciousness-ofconsciousness, and eo ipso consciousness of self or subject and of spatiotemporal-centricity. On this modal analysis, my awareness of my ongoing experience, and of myself as its subject, is intrinsic to the experience. It is not achieved in any separable higher-order monitoring. In particular, phenomenological reflection (through which we develop the preceding account of phenomenological structure) takes the form of a further, reflective activity in which we develop an account of the phenomenological structure of the exampled experience. That reflective activity is something above and beyond the activity of performing or living through my visual experience itself. Yet, in living through the experience, I am already aware of having the experience, or performing the act of consciousness. The inner awareness that resides in the experience is thus a pre-reflective form of awareness, as phenomenologists have long emphasized, from Husserl, extending Brentano, to Sartre and on to various contemporary philosophers looking to phenomenology. (Compare Smith 1986, Siewert 1998, Zahavi 1999, Thomasson 2000.) The awareness is “inner” in the sense that it is internal to, an intrinsic and integral part of, the experience, as opposed to an external observation or reflection or monitoring. My idiom ‘inner awareness’ is a variation on Brentano’s ‘inner consciousness’ as opposed to ‘inner observation’. Still, my modal analysis of consciousness differs significantly in detail from Brentano’s, Husserl’s, and Sartre’s, as I conceive the “modality” of consciousness. We emphasize, fundamentally, the distinction between mode and modality of presentation, or “intention”. Thus, the mode of presentation of the object of consciousness is carried by the content, < this green frog I’m jumping over >,

53 which Husserl called the noematic sense (Sinn) of the act of consciousness, reflecting the “way” the object is meant or intended. By contrast, the modality of presentation in the act is carried by the modifying content, < Phenomenally in this very experience I now here see … >. The modal part of the content of the act is an articulated extension of what Husserl called the thetic character of the act. Husserl did not mark precisely the distinctions indicated above, nor the indexical forms charted above; yet his detailed phenomenological analyses addressed many relevant aspects of these forms of awareness, of the temporality of one’s experience, of the spatiality of one’s “lived” or “living” body (Leib) in perception, and more. The overall content of the act Husserl called the noema of the act. Our concern here is the form and structure of content, rather than the ontology of content and its role in intentionality. (See Føllesdal 1969/1982; Dreyfus ed. 1982; Smith and McIntyre 1982; Smith 2007. For an alternative interpretation of noema, see Drummond 1992 and Sokolowski 2000.) On the present analysis, then: My experience in seeing the frog is conscious – in the form of consciousness typical of our everyday adult experience — in virtue of my inner awareness of this passing experience. And that form of inner awareness is achieved through the content, < Phenomenally in this very experience I now here >, which is part of the modality (not mode) of (re)presentation in the experience. Importantly, this inner awareness is not achieved in a second, higherorder act monitoring the experience. Rather, the inner awareness is an integral part of the experience. How does it work? If you will, inner awareness is a form of self-reflexive monitoring of the experience, achieved through the reflexive content , which is further modified — brought to the “light” of consciousness, as it were — by the content , and further articulated by the content . We should emphasize that, on this analysis, inner awareness is not a form of representation of the experience — whether higher-order or sameorder or whatever. Representation proper is achieved through the mode of presentation (= representation) of the object of experience; whereas inner

54 awareness is achieved through the reflexive character in the modality of presentation. Consequently, inner awareness does not “make an object” of the experience (as Jean-Paul Sartre stressed). Rather, I experience or live through the act of consciousness, with a certain reflexive awareness of the transpiring experience. Nor does inner awareness “make an object” of the subject, myself; I am the subject having this experience, not an object of the experience. Before going into further details of this modal model of consciousness, we post a caveat. The classical view, reinvigorated in recent higherorder monitoring theories, holds that a mental state is conscious if and only if the state includes or is accompanied by a second state of higher order. But that is too strong by far. Lower animals (if not plants) are conscious, newborn humans are conscious, dogs and birds are conscious. But it is implausible that all these varieties of consciousness bear all the elements of structure indicated. The target of the present, modal analysis is rather the form of awareness characteristic of everyday conscious experience in beings like us, that is, normal, modern, adult human beings. The analysis itself is an exercise in phenomenology, wherein we reflect “transcendentally” on the forms of experience with which we are familiar. (Contrary to a common view, I do not find “transcendental” analysis of consciousness incompatible with “naturalistic” accounts of consciousness. But that is a long and further story. Compare Smith 2004 and Ford and Smith 2006.) 2.

Levels of Structure Regarding Consciousness.

Phenomenological analysis of experience is in certain respects like logical analysis of a piece of language. Accordingly, the above analysis of the structure of an act of consciousness looks graphically rather like the logical analysis of a sentence reporting the visual perception at hand: ‘I see this green frog I’m jumping over’. That style of logical analysis of perception sentences was inaugurated by Jaakko Hintikka (1969), as Hintikka conceived the logic of such sentences along the lines of modal logic. Indeed, my choice of the term ‘modality’ in ‘modality of presentation’ is meant to echo the conception of intentional attitudes as “modalities”. (See Smith and McIntyre 1982 for details of the comparison of intentionality and the logical semantics of sentences ascribing intentional attitudes.) Recently,

55 Hintikka (2006) has re-emphasized the epistemic modalities, remarking that “the notions of knowledge and information are much more familiar and much more important in practice than philosophers’ semi-technical notions of necessity and possibility [the alethic modalities]” (p. 20). For our purposes, in any event, I want to assume an intuitive conception of phenomenological structure and intentional content, and an intuitive conception of “modality” applied therein. The structure or content of an experience is an ideal structure in the sense that it can be shared by other experiences of the same type. Within that broad notion of content, we factor out the indicated elements of structure in a typical experience, using the language of phenomenological description as our guide. Along those lines, and before turning to intentional modalities (perception, etc.), we look to the “Fregean” interpretation of Husserl’s theory of intentionality. Launched by Dagfinn Føllesdal (1969/1982), that model draws a parallel between reference via meaning in language and intentionality via meaning in consciousness. (See also Smith 1971, Dreyfus ed. 1982, Smith and McIntyre 1982, Smith 2007, and Beyer in the present volume). Following out this approach to phenomenology, we align coordinate levels of structure in act, content, and object of consciousness — whereby an act is directed via its content toward its object. We use language in forming a phenomenological description of an experience or act of consciousness: a first-person description that articulates or expresses the content or noema of the act. In the present model of the structure of consciousness, then, we align act description and act content, factoring the structure of an act’s content in correlation with the structure of the act’s phenomenological description. As noted, we use angle “quotes” to articulate structured content, within a structured phenomenological description. Briefly, the correlated structures in our running example are: a.

An experience or act of consciousness wherein (say) DWS sees a certain amphibious organism.

b. A phenomenological description of the experience (from the firstperson perspective):

56 ‘ Phenomenally in this very experience I now here see this green frog I’m jumping over ’. c. The ideal intentional content (noema) of the experience: < Phenomenally in this very experience I now here see this green frog I’m jumping over >. d. The real intentional content (noesis + sensation or hyle) of the experience: { Phenomenally in this very experience I now here see this green frog I’m jumping over } The phenomenological description (b) characterizes the experience (a) by articulating the ideal content or noema (c) instantiated in the real or concrete structure (d) in the experience (a). Here we use standard quotation marks to quote the expression, angle brackets to “quote” the ideal content, and braces to “quote” the real content. For Husserl, the real content of the experience consists of the relevant parts of the experience flowing off in time, the ideal content is instantiated or realized in the concrete experience (but shareable with structurally similar experiences), and the relevant parts of the ideal content are instanced in correlative parts of the experience. (The details of the ontology here are reconstructed in Smith 2007.) Accordingly, the underlinings indicate the relevant formal slots in the structure at each level.

THE DETAILS: FACTORS OF (SELF-) CONSCIOUSNESS 3.

Mode of Presentation, or Sinn: < this green frog … >.

The mode of presentation — reflected in what Husserl called the noematic Sinn of the experience — prescribes the object of the experience in a certain way: here, the object is intended as “this green frog …”. We shall not address this much-discussed element of content. But we note the different roles of indexical contents in the mode and modality of presentation: in the mode, and , , , in the

57 modality of presentation. (Smith 1989 appraises the demonstrative content in perception and its intentional appeal to context. Smith 2005 explicates the difference between the indexical content that specifies the object of perception and the reflexive content that reflexively indicates the experience itself.) 4.

Act Species Character: < … see … >.

All we need note about the act species content is that it belongs in the modality of presentation, starting with what Husserl called the thetic character of the act of consciousness. I may see this frog, or I may hear this frog. The difference lies in the modality of presentation, clearly not the mode of presentation that specifies the object as such. Within the modality of presentation, then, we mark out the content , reflecting the type of consciousness bearing the character of visual experience. 5.

Egocentric Character: < … I … >.

A certain egocentricity further characterizes the experience at hand. For the experience is not only intentionally directed as if toward a particular object, through the content . It is also intentionally directed as if from a particular subject, “I”, through the content . I have a form of “inner awareness” of myself in performing this act of consciousness, an awareness that is “inner” in that it is an internal part of the form of experience. However, I myself am not normally the object (or among the objects) toward which my current experience is directed. In seeing this frog, I do not see myself — unless, say, I see my reflection in a pond in which the frog I see is jumping from one lily pad to another. Rather, in seeing the frog I am aware of myself only reflexively insofar as I am aware of my experience’s being enacted by me. As we naturally say, “I see this frog …”, and so report on my seeing the frog: phenomenological description follows just this form. You may also see the frog, but the character of my experience is that “I see this frog”. (Drummond 2006 argues that the “case” should be “my seeing this frog”. However, that case loses the sense of activity wherein “I see …”, with ownership implicit.)

58 Accordingly, the content is an element of content belonging to the modality of presentation. To see that such a content is normally present in our experience, we need only consider what it would be like for an experience to lack this element of content. This experience of seeing “this green frog” is not free-floating, as if it were experienced by anyone or no one. Rather, I see “this green frog”. And I am, in a particular way, aware of myself seeing it, through the reflexive modal content , whence . The I or ego or self or subject has been a point of controversy for ages. In Husserl’s phenomenology, readers have found a “transcendental” ego, an embodied “empirical” ego, a naturalistic “psychological” ego, a social “personal” ego, and more. Since we are dealing with phenomenology, however, we should understand these as aspects of one being, aspects prescribed by different contents that appeal to different aspects of, well, myself. Phenomenologically, I experience myself in many different ways, and our focus here is on one particular way. Thus, what Husserl called the “pure” or “transcendental” I (Ich, ego) is not some strange and purely mental entity that is distinguished from myself as an embodied and social being. Rather, I submit, the “transcendental” I is I as experienced through the reflexive egocentric content in the modality of presentation in the experience at hand. (On different senses and aspects of the I, see Smith 1995 and 2007.) Can this reflexive modal content be absent in some of our experiences? Well, in Buddhist meditation where the sense of self is to be dissolved, either there is no egocentric character or, if you will, the slot is empty — “emptying” consciousness of the ego is precisely the goal of meditation (of certain kinds), so, yes, the slot remains but is emptied. Remember that we are talking about normal adult human experience. There may also be pathological mental states where one’s reflexive sense of self dissolves, and some of these have been studied in cognitive neuroscience. (Compare Ford and Smith 2006.) Still, some will ask, why must there be a slot for the content in the structure of such an experience? The “I” slot may seem redundant since, per hypothesis, an intentional experience is always directed as if from a center or origin of consciousness. Why isn’t the structure of intentionality then like that of an arrow that points toward some object distinct

59 from consciousness, but is merely an arrow with a tail and fletchings and no archer who aims it? That is the idea behind the ancient no-self view of consciousness. (Perhaps it is echoed in Anscombe’s argument that the firstperson “I” is not a referring expression, but merely a grammatical form as in the Latin “cogito” in no need of “ego cogito”. See Anscombe 1975.) The answer is that the content may fill different positions in the structure of an experience, and in any occurrence is tied to the person having that experience, so there must be an content in egocentric position indicating the subject performing the act of consciousness. For example, if I see myself in the mirror, the structure of my experience is, say, < … I see that I am standing near a portrait>. The first occurrence of is in the modality of presentation, reflexively indicating myself as subject of consciousness. But the second occurrence of is in the mode of presentation, indexically prescribing myself as object of consciousness, as part of what I see. Clearly, the content functions differently in the two occurrences. Moreover, the second occurrence depends on the first. Without the egocentric occurrence of in the modality, the indexical does not designate me, I myself, subject of the experience presenting myself in the mirror. Part of my self-consciousness, then, is my understanding, carried in the indexical content , that the “I” intended is the “I” who is now seeing that “I am standing near a portrait”. Many philosophers have argued, famously, that there is no proper “subject” of consciousness, whence nothing there to be indicated by the first-person content . (Compare Buddha, Hume, Sartre, Anscombe, Metzinger, among others). These lines of argument tend to conflate the phenomenological with the ontological features of the subject. However, if we keep our focus clearly on the content of consciousness, as in the phenomenological framework here, then we should allow that the content in an experience is: (a) a content rather than an object of experience, (b) a content that prescribes the subject both indexically and reflexively, (c) a content that normally occurs in egocentric position in the modality of presentation, (d) a content that may sometimes occur also in the mode of presentation.

60 Importantly, the individual prescribed by transcends the role of subject of the present experience. Here we enter ontology. Thus, on various ontologies, the individual is a Cartesian mental substance, or a Merleau-Pontian embodied subject, or a neuro-biological organism, or whatever kind of being you or I really am. (Compare, e.g., Baker 2006 where “I” am a being whose “constitution” is formed through a first-person perspective on a fundamentally material organism. That is a different sense of “constitution”, however, than we find in Husserlian transcendental phenomenology: compare Smith 2007 thereon.) That said, can the egocentricity slot be filled by anything other than the first-person indexical content < I >? At first thought, no, not in normal adult experience. Not by or , not by . Where the sense of self is part of inner awareness, as in the typical case described, then only < I > can fill the slot. That’s part of the “logic” of self-consciousness, or inner awareness: the content that occurs in the egocentric slot must work reflexively, as does the content in that postion. (However, there is one other case, that of collective consciousness, where the proper form of consciousness is < we intend to … >. See Searle 1995.) As Husserl and Merleau-Ponty showed in detail, self-awareness or egocentricity typically involves a sense of one’s embodiment. Here is yet another reason why the egocentric character must be granted its proper position in the structure of consciousness. What indicates, in the egocentric position, is the active subject, embodied and enminded and enacting consciousness in space and time. 6.

Temporal and Spatial Characters: < … now here … >.

A typical intentional experience is not only experienced as if directed from a subject, “I”, and toward an object, “this green frog …”. It is also experienced as being enacted “here” and “now” by the subject, and so as situated in time and space. Thus, in perception I see an object from my situated spatiotemporal perspective: my perceptual experience itself has a certain spatiotemporal “centricity”. And I am aware of that situatedness or centeredness as “now” and “here”, i.e., through the contents and . These elements of content and indicate the temporal and spatial location of the experience — my-seeing-the-object — as I

61 experience the perception. In this role the contents and take their place within the modality of presentation in the experience. For they modify the subject’s seeing such-and-such, as opposed to the perceived object. To be sure, in the case of perception, the object is presented as, say, “this frog here before me now”. (Compare the details of analysis in Smith 1989.) Thus, the indexical contents and in the mode of presentation serve to characterize the object’s spatiotemporal location, relative to the subject. These contents may occur explicitly in the mode of presentation, or implicitly in the horizon of background meaning characterizing “this green frog … (now here before me)”. (Compare Smith 2007 and Smith and McIntyre 1982 on the Husserlian notion of horizon.) By contrast, where and indicate the spatiotemporal location of the act of perception as experienced, as “I now here see this frog …”, the contents and occur in the modality of presentation. There they play the role of modifying the subject’s seeing “this green frog …”. They indicate the act’s spatiotemporal location reflexively, indicating a phenomenological feature of the subject’s performing the act of perception. That said, these characters also help to situate the object I see in relation to my seeing it, that is, as I see it. The elements and , in the modality of presentation, come into play in characterizing what Husserl called my “living” or “lived” body (Leib), as distinguished from my “physical” body (Körper). At issue are not two distinct bodies, but two aspects of my body, aspects indicated by two different forms of awareness, reflected in the contents and . Though we cannot go into the issue here, I would explicate the sense of embodiment as Leib, central to the Lebenswelt, in terms of the modal content . These indexical, reflexive contents are the door opening onto our experience of embodiment, of temporality, and of spatiality — phenomenological structures on which Husserl wrote many hundreds of pages. The complex content marks the spatiotemporal perspective in an experience in relation to its egocentric character. This complex indexical content ties into the structure of consciousness wherein I am aware of myself as an embodied, acting subject, performing acts of consciousness, including volitional bodily actions, in time and space. That phenomenological form of awareness of time and space modifies my

62 awareness of myself — “I” — as an embodied subject or Leib, situated spatiotemporally and so acting and experiencing things as within my Umwelt or Lebenswelt. Descartes famously argued that bodies have extension in space and time, whereas thoughts or experiences do not. But Descartes was pressing a point of ontology. Our task here is phenomenology. We may or may not accept the Cartesian claim that mental acts are nonspatiotemporal (most philosophers today do not). But in any event that is a claim of ontology that transcends the phenomenological observation that in the perceptual experience at hand “I now here see this green frog …”. We are currently pressing the point for the case of perception. What about an act of abstract thought? Suppose I think that 17 is a prime number. Can I think this arithmetic proposition without my cogitation carrying the modal content ? Not normally. I am here in my study. My so thinking is part of my stream of unified consciousness in which I also see the ocean beyond the window, as I sit in this chair while thinking “17 is a prime number”. Accordingly, the full structure of my so thinking follows the form we have discussed, articulated thus: . Accordingly, even in thinking about a numerical fact, the modality of presentation in my so thinking includes the characters of egocentricity and spatiotemporal situatedness, reflected in the modal content . Thus, the phenomenology of everyday experience, even where I am thinking about mathematics, finds me living in my role in the Lebenswelt. “In this very experience I now here see/think/wish …”: it is ever thus. (Compare Zahavi 2006 on temporality and embodiment in (self-) consciousness.) 7.

Reflexive Character: < … in this very experience … >.

The central form of self-consciousness in the sense of consciousness-ofconsciousness is the reflexive awareness of an experience that is achieved through the reflexive content in the modality of

63 presentation in the given experience. It is this form of awareness of our passing experience that some philosophers would analyze in terms of a concurrent monitoring of experience (either higher-order or same-order: see Kriegel and Williford eds 2006). On the modal analysis, however, the work of “monitoring” consciousness is done within the given experience itself. It is achieved through the specifically reflexive content in the indicated slot within the modality of presentation. Importantly, that content does not normally occur in the mode of presentation of the object of consciousness. Such would be the case where I think that “this very thought is false” (enacting the judgment of the Liar’s Paradox). In that case, my experience intends or is about itself, as the content , occurring in the mode of presentation, prescribes the very act of thinking in which the content occurs, predicating falsehood of the act. However, in a case like my seeing “this green frog …”, my experience itself is not presented to itself, and is no part of the object of consciousness. Still, on the present model, the content occurs in the modality of presentation. In that role it reflexively indicates the very experience in which the content occurs, albeit without presenting the experience as part of the object of consciousness. Because the content is part of the modality, not the mode, of presentation, then, the experience does not “make an object” of itself: it does not present itself as part of the object of consciousness. (Drawing on Husserl and Sartre, Zahavi 1999 criticizes higher-order and same-order monitoring models on grounds that they make an object of the mental state that is conscious. Drummond 2006 worries that my model tends in that direction. However, as I am urging now, it does not.) The content can, of course, play roles in both the mode and the modality of presentation, as in the case where: . In the structure of such an experience, the content plays two completely different roles. In the first occurrence it reflexively indicates the experience as it transpires, (partly) rendering the experience self-conscious. In the second occurrence, however, it indexically indicates

64 part of the object of thought, viz., the thought’s being false. Distinguishing these two roles underscores the role of in the modality as opposed to the mode of presentation. A variant phrasing of “in this very experience” is “herein”. Again, the reflexive form is clear. If a legal document declares, “Herein the first party promises to pay to the second party a sum of $500,” the reflexive “herein” indicates the document in which the promise is made. In spoken language we find a similar reflexive that is implicit where I say, “I [hereby] promise to pay you $500.” Again, where the priest says, “I hereby declare you husband and wife,” the reflexive “hereby” indicates the speech act itself, evoking the performative force of the speech act: by virtue of the act itself the marriage is effected, as expressly registered by “hereby”. Similarly, on the present model of self-consciousness, the content effects a reflexive awareness of the experience itself. That content reflects part of the way the subject performs the act of consciousness: with a reflexive awareness that is part of the modality, but not the mode, of presentation. Observing the reflexive character in experience, we have to address one further form of character that plays a role in making an experience conscious: what we call its phenomenality. 8.

Phenomenal Character: < phenomenally … >.

The phenomenal character of an act of consciousness is the “light” of consciousness. The difference between conscious sight and blind sight, for example, is that conscious sight has a phenomenal character (there is light!) that blind sight lacks. This character, reflected in the content , takes its place in the modality of presentation. For it is part of the way the act is experienced but no part of the characterization of the object in the mode of presentation. In recent philosophy of mind, phenomenality has often been identified with the sensuous character, the “qualia”, in sensory perception. However, conscious thought, as when I consciously think that “17 is a prime number”, has a phenomenal character but no sensory qualia. The light of consciousness is present, but the experience is not sensory. There has been much debate about the role of qualia in consciousness. Here, however, I

65 want to join with Kant and Husserl, as against pure empiricism, in recognizing that sensory perception itself does not divide into two experiences, sensation and conception. Rather, a typical visual experience is already a fusion of sensory and conceptual elements (hyletic and noetic elements, in Husserl’s idiom). And other types of experience are conscious, bearing a phenomenal character, without being sensory (without a sensory or hyletic component). Accordingly, we assume here that phenomenality is a distinctive character of consciousness, and a character distinct from that of sensous “qualia”. In the recent literature, phenomenality has often been characterized as the quality of “what it is like” to have a given type of experience. That characterization is all right as an opening salvo, but the point of our discussion is to factor out the several features of experience that are characteristic of consciousness. Either what-it-is-like is the phenomenal character or content , or it is the whole structure of consciousness that we are analyzing, the complex character or content (in our running example). But is the character redundant? Is it just the character that we have been analyzing throughout our discussion? I think not. For, as we factor the structure of consciousness, we find the several distinguishable elements we have addressed. And if it is possible for a mental act of blindsight to have an act species character, an egocentric, a spatiotemoral-centric character, even a reflexive character, and yet remain unconscious, then phenomenality is a distinguishable character that modifies these other characters, all within the full modality of presentation in an experience. 9.

The Character of Inner Awareness: .

The complex content embodies the fully articulated character of inner awareness of experience, or self-consciousness in the sense of consciousness-of-consciousness. Here lies the form of “monitoring” that other philosophers have sought in either higher-order or same-order monitoring (as appraised in Kriegel and Williford eds. 2006).

66 Could the phenomenal and reflexive characters occur separately, one without the other? Well, think about it. In blindsight, the mind could reflexively indicate the mental event of seeing a frog, but without any phenomenal character. Conversely, in a particular state of meditation, perhaps, one could phenomenally hear a steady tone of middle C without a reflexive awareness of “this very experience” of hearing. Again, it would seem that the brain could produce a phenomenal state without a reflexive monitoring. These “logical”, or phenomenological, possibilities need to be corroborated by meditative practice and experimental neuroscience, but we can here distinguish the formal structures of phenomenal and reflexive contents. Could an experience with the phenomenal-and-reflexive character lack the characters of egocentricity and spatiotemporal-centricity? Even if such forms of experience are possible (say) in altered states achieved through meditation or drugs, nonetheless in the normal course of everyday experience our acts of consciousness include the modal content . So in a typical conscious experience, on our analysis, the content modifies the full structure of experience, as in our example: < Phenomenally in this very experience I now here see this green frog I’m jumping over >. Our task has been to factor out the several, distinguishable elements of content that work together in this way to define the structure of consciousness, in such a form of experience. In a typical experience, however, my awareness of my passing experience is structured by not only the phenomenal-reflexive content , but also the centering or localizing content . Accordingly, my experience intimates itself not as a free-floating experience, performed some time and somewhere and by someone, but as “this very experience” centered or located in “I” and “now here”.

67 10.

Indexicality and the Stream of Consciousness.

On the present model, indexical elements of content are central to the phenomenological structure of everyday human consciousness. As suggested along our way, these indexical contents depend on, and draw upon, rich forms and activities of consciousness: in the sense of temporality, spatiality, subjectivity, embodiment, and the flow of consciousness — all characteristic structures of the stream of consciousness. Husserl analyzed these rich structures in great detail. Often he used indexical terms (“I”, ”now”, “here”) in his phenomenological analyses, but he did not organize his analyses in an explicit formal phenomenological structure like that in the present model. My emphasis here on indexical elements of content is meant to tie into richer structures like those Husserl appraised (even as we might modify the analyses). Exactly how the several indexical contents — , , , — tie into and depend on the stream of consciousness (and embodiment) is a longer story for another day. However, I want to clarify an important feature of how I understand the function of these indexical contents. It might be thought that a demonstrative or indexical content is a simple “pointer”. (Levine 2001 raises this worry, p. 109.) But that is not how indexical contents work. To stay with language for a moment, the word ‘that’ uttered in a particular context refers to the object then being “demonstrated” in some concrete way, perhaps by the speaker’s pointing his or her index finger in the direction of the object. According to Kaplan’s logic of demonstratives, the demonstration is an essential part of the expression ‘that [demonstration]’. (See Kaplan 1989.) Now, Husserl sketched a similar account of demonstratives, but explicitly tied the use of ‘that’ into the speaker’s perception of the object referred to. (See Husserl 1900/2001 on “essentially occasional expressions”, and Smith 1982 on Husserl’s account in light of Kaplan’s, and Smith 1989 on how I think indexical contents or senses work.) As we turn from the logic to the phenomenology, a much richer story comes into view. The indexical element in the mode of presentation in my seeing “this green frog …” may be codified in the demonstrative visual content . The particular object demonstratively intended in perception as “this …” is presented in a context from which the object itself is

68 abstracted. The function of attention draws the object out from the background context in which it is presented: as the Gestalt model puts it, the object is presented as a figure or Gestalt against a background. So demonstrative intentionality in perception is much richer than a simple “pointing”. And that form of intentionality, extracting the object from its background context, surely does not reduce to an external causal relation. (See details in Smith 1989 and in Searle 1983.) The indexical or reflexive elements in the modality of presentation also function in this way, where the indexical content draws the indicated feature of the experience out from, or against, a context or background. The relevant context for these features of consciousness, however, is the structured stream of consciousness itself. The content indicates the subject who is having the given experience, performing that act of consciousness as it transpires in the ongoing stream of consciousness “lived” by that subject. The content indicates the time of the act in the flowing stream. The content indicates the subject’s embodied location at that time. And indicates the act itself, extracted from the stream of consciousness.

CONCLUSION The modal model of (self-) consciousness articulates several interrelated forms of content in the rich flow of experience, defining respectively phenomenality, reflexivity, egocentricity, and spatiotemporal-centricity. My inner awareness of an experience as it transpires, as I perform the act of consciousness, is defined by these quite different types of content, each defining appropriate formally distinct elements in my consciousness, in the modality of presentation in the experience. Consciousness, as we experience it in the fullness of everyday experience, is thus a richly structured phenomenon.

NOTE: I want to thank the audience at the 2006 Lauener Symposium on Analytical Philosophy. Their questions and comments following my oral presentation of the ideas in the present essay were illuminating, and I have

69 tried to respond to these in drafting the essay for publication. In particular, I benefited from remarks by Dagfinn Føllesdal, Patrick Suppes, John Perry, Michael Friedman, Alex Burri, Christian Beyer, Øystein Linnebo, Iso Kern.

REFERENCES Anscombe, G. E. M. 1975. “The First Person”. In S. Guttenplan, ed., Mind and Language. Oxford: Clarendon Press. Pp. 45-65. Baker, Lynne Rudder. 2000. Persons and Bodies: A Constitution View. Cambridge and New York: Cambridge University Press. Beyer, Christian. 2012. “Noema and Reference”. In the present volume. Dreyfus, Hubert L. Editor. 1982. In collaboration with Harrison Hall. Husserl, Intentionality and Cognitive Science. Cambridge, Massachusetts: MIT Press. Drummond, John J. 1990. Husserlian Intentionality and Non-Foundational Realism: Noema and Object. Dordrecht and Boston: Kluwer Academic Publishers. Now Springer. —— . 2006. “The Case(s) of (Self-) Awareness”. In Kriegel and Williford eds. 2006. Føllesdal, Dagfinn. 1969/1982. “Husserl’s Notion of Noema”. In Dreyfus ed.1982. Pages 73-80. Reprinted from The Journal of Philosophy, 1969. Ford, Jason, and David Woodruff Smith. 2006. “Consciousness, Self, and Attention”. In Kriegel and Williford eds. 2006. Pp. 353-377. Hintikka, Jaakko. 1962. Knowledge and Belief. Ithaca, New York: Cornell University Press. —— . 1969. “On the Logic of Perception”. In his Models for Modalities. Dordrecht and Boston: D. Reidel Publishing Company. Now Springer. Hintikka, Jaakko, et al. 2006. The Philosophy of Jaakko Hintikka. Edited by Randall E. Auxier and Lewis Edwin Hahn. The Library of Living Philosophers, Volume XXX. Chicago and LaSalle, Illinois: Open Court.

70 Kriegel, Uriah, and Kenneth Williford. Editors. 2006. Self-Representational Approaches to Consciousness. Cambridge, Massachusetts: MIT Press. Kaplan, David. 1989. “Demonstratives: An Essay on the Semantics, Logic, Metaphysics, and Epistemology of Demonstratives and Other Indexicals.” In Almog, Joseph, John Perry, and Howard Wettstein. Editors. Themes from Kaplan. Oxford and New York: Oxford University Press. Pages 481-614. Levine, Joseph. 2001. Purple Haze: the Puzzle of Consciousness. Oxford and New York: Oxford University Press. Rosenthal, David M. 2005. Consciousness and Mind. Oxford and New York: Oxford University Press. Essays dating from 1986 forward. Dretske, Fred. 1995. Naturalizing the Mind. Cambridge, Massachusetts: MIT Press. Searle, John R. 1983. Intentionality. Cambridge and New York: Cambridge University Press. —— . 1995. The Construction of Social Reality. New York: The Free Press. Siewert, Charles. 1998. The Significance of Consciousness. Princeton, New Jersey: Princeton University Press. Sokolowski, Robert. 2000. Introduction to Phenomenology. Cambridge and New York: Cambridge University Press. Smith, Barry, and David Woodruff Smith. Editors. 1995. The Cambridge Companion to Husserl. Cambridge and New York: Cambridge University Press. Smith, David Woodruff, and Ronald McIntyre. 1982. Husserl and Intentionality: A Study of Mind, Meaning, and Language. Dordrecht and Boston: D. Reidel Publishing Company. Now Springer. Smith, David Woodruff. 1970. Intentionality, Noemata, and Individuation: The Role of Individuation in Husserl’s Theory of Intentionality. Stanford, California: Stanford University, doctoral dissertation; UMI, Ann Arbor, 1970. —— . 1982. “Husserl on Demonstrative Reference and Perception”. In Dreyfus ed. 1982. Pp. 193-213.

71 —— . 1986. “The Structure of (Self-) Consciousness”. Topoi, Vol. 5, No. 2. Pp. 149156. —— . 1989. The Circle of Acquaintance: Perception, Consciousness, and Empathy. Dordrecht and Boston: Kluwer Academic Publishers. Now Springer. —— . 1995. “Mind and Body”. In Smith and Smith eds 1995. —— . 2004. Mind World: Essays in Phenomenology and Ontology. Cambridge and New York: Cambridge University Press. —— . 2004a. “Return to Consciousness”. In Smith 2004. Pp. 76-121. —— . 2005. “Consciousness with Reflexive Content”. In David Woodruff Smith and Amie L. Thomasson. Editors. Phenomenology and Philosophy of Mind. Oxford and New York: Oxford University Press. Pp. 93-114. —— . 2007. Husserl. London and New York: Routledge. Thomasson, Amie L. 2000. “After Brentano: A One-Level Theory of Consciousness”. The European Journal of Philosophy, 8/2: 190-209. Zahavi, Dan. 1999. Self-Awareness and Alterity: A Phenomenological Investigation. Evanston, Illinois: Northwestern University Press. —— . 2006. “Thinking about (Self-) Consciousness: Phenomenological Perspectives”. In Kriegel and Williford eds. 2006.

Noema and Reference Christian Beyer*

Abstract Føllesdal's much-discussed "Fregean" reading of Husserl's notion of noema has recently been criticized on the grounds that, on this reading, Husserl would be committed to either (1) a radical form of internalism about intentional content and/or (2) descriptivism about the meaning of singular terms and/or (3) the claim that there can be no direct reference, since intentional representation is always mediated by noematic Sinn. I argue against these objections.

There can be no doubt that Dagfinn Føllesdal's paper on "Husserl's notion of noema", which appeared in The Journal of Philosophy in 1969, is the most-quoted paper on Husserl's phenomenology, probably of all time. I think it is safe to say that today it is still cited as often as 30 years ago. It is doubtful, however, if Føllesdal is happy with the way his interpretation is received in many cases. For, it seems to me that more often than not his interpretation gets distorted and his critics argue against a strawman. Føllesdal's reading is usually referred to as "the Fregean interpretation"; which is perfectly fine, to my mind. For, most, if not all, of the theses about the noema he ascribes to Husserl in his 1969 paper can be looked upon as generalizations of Fregean semantic principles to the field of all intentional states or "acts". Thus, given that Frege considers sentential meanings (called "thoughts" in the case of questions and judgements) both to be compositional (cf. Frege 1892, p. 32) and to be potential contents of propositional attitudes (cf. Frege 1918, p. 62), a Fregean could subscribe to all of the following twelve theses of Føllesdal's Husserl: 1. The noema is an intensional entity, a generalization of the notion of meaning (Sinn, Bedeutung) ... 2. A noema has two components: (1) one which is common to all acts that have *

Georg-August-Universität Göttingen

74

the same object, with exactly the same properties, oriented in the same way, etc., regardless of the 'thetic' character of the act – that is, whether it be perceiving, remembering, imagining, etc., and (2) one which is different in acts with different thetic characters ... 3. The noematic Sinn is that in virtue of which consciousness relates to the object ... 4. The noema of an act is not the object of the act (i.e., the object toward which the act is directed) ... 5. To one and the same [n]oema, there corresponds only one object ... 6. To one and the same object there may correspond several different noemata ... 7. Each act has one and only one noema ... 8. Noemata are abstract entities ... 9. Noemata are not perceived through our senses ... [10*. Noematic Sinne can be referred to.]1 [11*. Noematic Sinne are objects corresponding to higher-level noematic Sinne.]2 12. [The noema is a complex system of ... determinations ... which [in the case of perception; CB] make a multitude of ... data be appearances of one object ... [according to a] more or less vaguely predelineated pattern ...] This pattern of determinations, together with the 'Gegebenheitsweise,' [which includes the 'thetic' character; CB] is the noema. (Cf. Føllesdal 1982, pp. 73-80)

Recently, critics such as Dan Zahavi and, in a review of a book based on 1

This follows from Føllesdal's thesis 10: "Noemata are known through a special reflection, the phenomenological reflection" (p. 78). 2 This follows from Føllesdal's theses 10 and 11: "The phenomenological reflection can be iterated" (p. 79).

75

my PhD thesis, Matteo Bianchin have complained that on this reading Husserl would be committed to either •

internalism about intentional content (cf. Zahavi 2004, pp. 47, 54) and/or

•

strict descriptivism about the meaning of singular terms (cf. Zahavi 2004, p. 46) and/or

•

the claim that there can be no direct reference, since intentional representation is said to be mediated by an abstract object,

all of which are found to be implausible. What I like about these objections is that they are systematic, and related to relevant debates in analytic philosophy – though they also echoe criticism advanced by earlier adherents of the traditional interpretation. Let me address these objections in turn. Internalism Internalism equates intentional content with "narrow content" or, to put it another way, with (something like) context-independent linguistic meaning – what is called "semantic role" by John Perry and "general meaningfunction" by Husserl.3 This notion of content often helps us explain people's behaviour psychologically, as in the case of Perry's shopper with the torn sugar-bag who suddenly comes to believe that he himself is making a mess (cf. Perry 1993a, p. 33). All things being equal, the shopper's counterpart on Twin Earth would display the same kind of 3

The semantic distinction between an expression's general meaning-function, on the one hand, and the propositional, or sub-propositional, content – the "respective meaning" – expressed in a given context of utterance, on the other hand, is introduced in the first of Husserl's Logical Investigations, § 26 (Hua XIX/1/I, pp. 78 ff). To illustrate, if you and I claim "I am the author of this paper", then we both express the same general meaning-function but different respective meanings (which differ in their truth-value). According to Husserl, it is the respective meaning, rather than the general meaning-function, that uniquely determines the expression's referent, in the sense that two expressions sharing that meaning are thus bound to refer to the same object(s), if any.

76

behaviour as his psycho-physical twin on Earth when realizing that he is making a mess. In its most sophisticated form, a version of which has for instance been developed by David Smith (cf. Smith 1989), internalism denies that intentional content thus construed meets the condition specified in thesis 5 above, i.e. the claim that this content alone – that is, independently of the relevant context – uniquely determines the object referred to or represented. It rather amounts to a context-sensitive internalism, one that is incompatible with the claim that content uniquely determines object. Given thesis 5, then, Husserl must not be interpreted as subscribing to context-sensitive internalism. What about radical versions of internalism, according to which narrow content does uniquely determine the object, quite independently of context? Take another look at thesis 2, condition # (1), describing what Husserl calls noematic Sinn. Far from mentioning an intensional entity that is independent of the respective (say:) perceptual context, it explicitly mentions both the object and its orientation towards a given subject, which according to Husserl depends on the subject's living body and its relation to the environment. Thesis 2 thus leaves room for an externalist reading of the notion of noema, especially in connection with thesis 12. I shall come back to this point shortly. The only thesis that may at first glance suggest an internalist interpretation is thesis 1. However, thanks to a recently published volume of Husserl's collected works (Hua XX/1), we now have decisive textual evidence showing that while Husserl does identify the noema or, to be more precise, the noematic Sinn with a meaning-like entity, just as Føllesdal's interpretation (as manifested by thesis 1) would have it, he does not equate it with general meaning-function but rather with respective meaning, i.e. contextually determined (sub-)propositional content. This definitely rules out the ascription of a radically internalist position to Husserl while still speaking in favour of Føllesdal's reading; it makes it clear that Husserl is rather committed to subscribe to a version of externalism about intentional content. After all, the respective meaning expressed depends on objects located in, or constituting, the relevant context of utterance, according to Husserl, with these objects usually belonging to the external environment, e.g. to the perceptual surrounding of the speaker. And externalism can be looked upon as the view that the

77

environment helps determine the intentional content. In the text I have in mind, a section of chapter 1 of Husserl's revised version of the sixth of his Logical Investigations from 1913, he discusses indexicality and explicitly relates the "noema"-terminology to his distinction between general meaning-function and respective meaning. In § 5 Husserl says: Noematically speaking: the wording as signum has an intentional (noematic) feature belonging to and founded in it, a feature in virtue of which the word 'means' something ... Thus significance here amounts to this feature that is attached to the wording in conscious 'understanding' (or conscious assertion), whilst the meaning itself is the meant as such [das Vermeinte als solches; i.e., the noematic Sinn; CB] ... Let us begin by examining a primitive case, where a mere 'that' or 'this' is taking the place of the subject, as in the case of 'That is a blackbird'. 'That' is an occasional expression, an expression whose meaning depends on the respective context ... The meaning of 'this' [sic] differs according to circumstances, but in such a way that a common element is preserved across all those changes, an element that distinguishes this type of ambiguity from contingent equivocation. In all of these cases we are dealing with demonstrative reference [Es ist immer auf etwas hingewiesen] ... The proper meaning does not reside in the mere perception; rather, a new act builds itself on the basis of it, an act which orients itself by the perception: the act of this-meaning. Obviously, it is this latter act that serves the meaning-bestowing function proper ... The act of demonstrative reference remains essentially the same, regardless of which perception from the manifold of perceptions belonging together by presenting what is – recognizably – one and the same object throughout happens to underlie it. (Hua XX/1, text # 2, § 5, pp. 74-78; my translation)4 4

The German original runs: "Noematisch gesprochen: Zum Wortlaut als signum gehört ein in ihm fundierter intentionaler (noematischer) Charakter, durch ihn 'bedeutet' das Wort etwas ... Die Bedeutsamkeit bedeutet hier also diesen im 'verstehenden' (bzw. aussagenden) Bewusstsein dem Wortlaut anhaftenden Charakter, während die Bedeutung selbst das Vermeinte als solches ist ... Wir fassen zunächst einen primitiven Fall ins Auge, den des bloßen 'das' oder 'dies' an Subjektstelle, z.B. 'Das ist eine Amsel'. 'Das' ist ein okkasioneller Ausdruck, dessen Bedeutung vom jeweiligen Kontext abhängig ist ... Von Fall zu Fall hat das 'dies' [sic] eine andere Bedeutung, jedoch so, dass in allem Wechsel ein Gemeinsames erhalten bleibt, welches diese Vieldeutigkeit von derjenigen zufälliger Äquivokationen unterscheidet.

78

Besides supporting the claim that noematic Sinn is contextually determined respective meaning rather than general meaning-function, which rules out any internalist reading, this passage also suggests what a Husserlian externalism about intentional content might look like. For, just like thesis 12 from "Husserl's notion of noema" would lead us to expect, Husserl does not relate noematic Sinn to an isolated perceptual experience here but rather to a whole "manifold of perceptions", such as a continuous observation, i.e. to a holistic, transtemporal structure of consciousness. Here as elsewhere, Husserl employs a research strategy in the theory of intentional content and reference that could be called his dynamic method: Dynamic method: Intentional states and experiences are looked upon as momentary components of certain transtemporal cognitive structures – dynamic intentional structures – in which one and the same object or state of affairs is represented throughout a period of time during which the subject's cognitive perspective upon that object or state of affairs is constantly changing. This method has us look upon noematic Sinn under the "functional aspect" (Husserl) of how it enables us to keep the intentional object "in mind (im Sinn)" (cf. Hua III/1, pp. 196 ff.), instead of viewing it merely statically as a psychological type to be instantiated by isolated moments of consciousness. It makes us regard any content of the latter sort, particularly "static perceptual content", as a mere "abstraction from dynamic content", as Kevin Mulligan writes in his contribution for The Cambridge Companion to Husserl (Mulligan 1995, pp. 195, 197). If we individuate the noematic Sinn of a conscious demonstrative reference, or identification, according to the dynamic method, then that Es ist immer auf etwas hingewiesen ... Nicht liegt ... die eigentliche Bedeutung in der bloßen Wahrnehmung, sondern aufgrund der Wahrnehmung baut sich ein neuer, sich nach ihr richtender, in seiner Differenz von ihr abhängiger Akt auf, der Akt des DiesMeinens. Dieser ist es offenbar, der im eigentlichen Sinne die Bedeutungsgebung leistet ... [D]as hinweisende Meinen ist dasselbe, welche Wahrnehmung aus der Mannigfaltigkeit zusammengehöriger Wahrnehmungen zugrundeliegen mag, in denen immer derselbe, und erkennbar derselbe Gegenstand erscheint."

79

content indeed turns out to represent "a complex system of ... determinations ... which make a multitude of ... data be appearances of one object", as Føllesdal puts it in his formulation of thesis 12. However, to vary a question raised by Wittgenstein: which element of noematic Sinn makes my conscious demonstrative reference to this object a reference to this object? How does the Sinn manage to determine the intentional object? It is here that Husserl brings in a pair of notions the importance of which for an adequate understanding of Husserl's view Føllesdal has stressed repeatedly; I mean the concepts of the determinable X and the intentional horizon, respectively. Husserl sees quite clearly that conscious demonstrative acts of reference (just as intentional experiences given voice to by means of other indexicals or by genuine proper names) are characterized, among other things, by their singularity: they represent a particular object, or set of objects, x, such that x is to be regarded as the intentional object of the respective experience in all relevant possible worlds (i.e. in all actual or counterfactual circumstances relative to which we are determining the object represented by that experience). Thus, for instance, in § 47 of Ideas, he describes what an experiencing subject, at a given time, in the light of his (or her) current experiences, considers to be "the actual world" as a "special case" of a whole manifold of "possible worlds" each of which corresponds to a possible future course of experience (possible, that is, relative to the current experiences in question) (cf. Hua III/1, p. 100). These (actual or potential) future experiences can be said to be (more or less) anticipated by the experiencing subject at the respective time, and, as David Smith and Ronald McIntyre have explained in detail, they constitute what Husserl calls the "intentional horizon" of the experience in the light of whose noematic Sinn they are anticipated (cf. Smith and McIntyre 1982, sec. V-VIII). For example, if you consciously see something as a table, you will expect it to appear to you in certain ways if you go around and observe it. What binds together the intentional horizon of a given experience? According to Husserl, all of the (actual or potential) experiences constituting that horizon share a sense of identity through time, which sense he labels as the determinable X they belong to. As a first approximation, two experiences of a given subject belong to the same

80

determinable X if and only if they are believed to represent the same object, thus displaying, on the noetic side, what Perry calls "internal identity" (cf. Perry 1993b) – provided that we restrict ourselves to a single subject of experience.5 But we also need a criterion of intersubjective identity of the determinable X, and Husserl was well aware of this (for details see Beyer 2000, § 7). The determinable X a given experience belongs to, with respect to certain other experiences, helps us answer the question of what determines the referent of that experience, if not its general meaning-function alone. In order to take the role played by the determinable X into account properly, we ought to employ the dynamic method. Thus conceived, the determinable X is apt to lead us back through time towards the original situation where the referent of the relevant unified series of successive intentional horizons was fixed, like for instance the occasion of the subject's first perceptual encounter with a particular object: the corresponding perceptual experience will belong to the same determinable X as all of the (remaining) experiences belonging to the relevant series. In a more recent terminology, introduced by Perry, one may say that in this perceptual situation the subject has opened a mental file about a particular object (cf. Perry 1993b). In a recently published research manuscript from 1913 Husserl refers to mental files associated with proper names as "Eigenbegriffe (individual notions)": I see an object without a 'historic' horizon, and now it gets one. I have experienced the object multifariously, I have made 'multifarious' judgements about it and have gained multifarious [pieces of] knowledge about it, at various times, all of which I have connected. Thanks to this connection I now possess a 5

I was encouraged to work out my own reconstruction of the notion of noematic Sinn when I learnt about the close connection between Husserl's concept of the determinable X and Quine's notion of "full reification" that Føllesdal has drawn, among other things, in his paper on "Gödel and Husserl", a draft of which we discussed when I was happy enough to spend a year as a visiting scholar at Stanford (cf. Føllesdal 1999, pp. 390 ff.). By "full reification", Quine means "the sophisticated stage where the identity of a body from one time to another can be queried and affirmed or conjectured or denied independently of exact resemblance" (quoted after ibid., p. 391).

81

'notion' of the object, an individual notion ... [W]hat is posited in memory under a certain sense gains an epistemic enrichment of sense, i.e. the x of the sense is determined further empirically [erfahrungsmäßig]. (Hua XX/2, supplement XLI, p. 358; my translation)6

The identification of an individual notion such as this with a mental file à la Perry is also supported by the fact that Husserl characterizes such notions as being infinitely "open" and "in flux"7. Now it is the "referent" of the relevant mental file, or individual notion, that will normally count as the common intentional object of the experiences bound together in a unified series of successive intentional horizons in which the object "constitutes itself" empirically. In cases where the "referent" of a mental file changes across time – i.e. is unnoticedly replaced by another object – the situation becomes more complicated, to be sure. The same goes for cases of perceptual judgements leading to, or taken by the respective subject to be confirming, entries into an already existing file. (See Beyer, ibid.) On this reading, or rational reconstruction, of Husserl's conception of the determinable X, there is a close link, at least in the case of proper names and, arguably, in the ubiquitous indexical case, between intentional content (including determinable X) on the one hand, and extra-mental reality on the other, such that intentional content so conceived determines reference in much the same way more recent externalist theories of content would have it, i.e. in such a way that the referent can in turn be said to help determine the intentional content (see Beyer 2000; 2001). In a recent paper on "Indexicals and the mind (Indikatoren und der Geist)", Føllesdal agrees 6

The German original runs: "Ich sehe einen Gegenstand ohne einen 'historischen' Horizont [fn.: Der Bekanntheitshorizont und Wissenshorizont eines Gegenstands], und nun bekommt er ihn. Ich habe den Gegenstand vielfältig erfahren, 'vielfältige' Urteile habe ich über ihn gefällt, vielfältige Kenntnis von ihm in verschiedenen Zeiten gewonnen und habe sie verknüpft. Nun habe ich durch diese Verknüpfung einen 'Begriff' von dem Gegenstand, einen Eigenbegriff ... [D]as in [der Erinnerung] mit einem gewissen Sinn Gesetzte erfährt eine erkenntnismäßige Sinnbereicherung, das heißt, das x des Sinnes bestimmt sich näher erfahrungsmäßig." 7 Ibid., p. 359.

82

that Husserl has "a wide notion of meaning", according to which the reference determined by meaning "not only depends on the meaning of our words", which I consider to be their general meaning-function, "but also", he says, on the relation of "our living body" to "objects in the environment", e.g. in the relevant perceptual surrounding (cf. Føllesdal 2006, pp. 231 f.). One might aptly call this a two-factored view on respective meaning or noematic Sinn (Beyer 2003 ff., sec. 3). The following passage from a 1911 research manuscript on demonstrative reference supports this externalist reading; note that Husserl here basically anticipates the Twin Earth thought-experiment made famous many decades later by Putnam: [I]magine two people on two distinct celestial bodies who have exactly the same kind of appearances of their respective surrounding and who are both representing 'the same' object and making, on the basis of these representations, 'the same' assertion. Hasn't the 'this' got a different meaning in each of these cases? (Hua XXVI, supplement XIX, p. 212; my translation)

Husserl's answer is affirmative. An internalist would of course disagree. What about empty singular terms and the conscious acts of reference – based, for instance, on hallucination – that can be given voice to by means of such terms? As Zahavi partly acknowledges, the Fregean interpretation, as manifested by theses 3 and 4 in particular, can easily account for Husserl's remarks about these cases (in § 88 of Ideas and elsewhere), whilst its rival, the so-called East Coast interpretation, has a hard time explaining them, as it basically equates the noematic Sinn with the intentional object, albeit the object "as represented" by the experience, whatever exactly that may mean (cf. Zahavi 2004, pp. 53 ff.).8 Notice that theses 3 and 4 preclude a "neo-Russellian" conception of noematic Sinn: they do not admit of a conception of the noematic Sinn entertained in 8

I say "partly acknowledges" because it is somewhat unclear whether Zahavi wants to present an exegetical argument (concerning what Husserl actually says about the noema) or rather a phenomenological argument (regarding the question if phenomenology ought to be able to account for cases of hallucination).

83

conscious singular reference as the constituent of singular propositions à la Russell and Kaplan. One of the advantages of an externalist conception less radical than a neo-Russellian one (and, for that matter, a neo-Fregean one à la Evans) is that it allows us to specify the noematic Sinn of both veridical perceptions and hallucinations so as to bring out their singularity. Already in his 1894 essay titled "Intentional Objects" Husserl stressed that objectless representations such as hallucinations can in a sense be chararacterized as "representing an intentional object", provided that this characterization is understood to be made "under an existential assumption", as follows: "If the act of hallucination were veridical, it would represent such-and-such an object (under such-and-such aspects)" (cf. Hua XXII, esp. pp. 316 f.). Something similar goes with regard to the singularity of a hallucinatory experience's noema: if such an experience were veridical, it would, in virtue of its noema, represent a particular perceptual object in all relevant possible worlds (see above). Thus, we can provide an existentially neutral specification of the noema of a (veridical, illusory or hallucinatory) perceptual experience and still bring out the singular character of its content. The specification might run as follows: The noematic Sinn of a perceptual experience i is such that either (1) there is an object x that i represents in virtue of its noematic Sinn, where x is to be regarded as the referent of i in all relevant possible worlds, or (2) there would be an object meeting condition (1) if i were veridical. Condition # (2) enables us to make sense of the behaviour of a speaker/thinker making counterfactual assumptions about an object which he, unknowingly, merely hallucinates or quantifying into modal statements about that alleged object (cf. Beyer 2000, pp. 26-31). Notice that on the above-proposed externalist reading, the noematic Sinn will differ depending on whether condition # (1) or (2) is satisfied. Nevertheless, our noematic specification does not rely on the existence of a perceptual object: if there is no perceptual object, condition # (2) will be satisfied – provided that we are dealing with a perceptual experience. The rationale behind condition # (2) is that even in the non-veridical case an individual notion (a mental file) and consequently a unified series of intentional horizons gets activated, on the basis of the same sensory material, or hýle,

84

as in the veridical case. Let me now turn to the remaining two objections recently raised against the Fregean reading. Here I can be more brief. Descriptivism Descriptivism, in the strict sense of the term that Føllesdal's critics seem to have in mind, holds that the referent of any given singular term, or, since we are dealing with Husserl, the intentional object of a given mental act of reference, is uniquely determined by the general meaning-function of a non-indexical definite description (expressing what Husserl would call a purely ideal meaning; cf., e.g., Hua XX/1, p. 279). This view entails a radically internalist view of conscious reference, and we have already seen that Føllesdal's Husserl does not subscribe to a view such as this. To be sure, there is a famous footnote in Frege's "On sense and reference" that may be interpreted along strictly descriptivist lines (cf. Frege 1892, p. 27, fn. 2), but, quite apart from the fact that Føllesdal has never claimed that Husserl and Frege agree on everything, Gareth Evans has demonstrated that being a Fregean (or "neo-Fregean") about singular reference certainly does not commit one to being a descriptivist (see Evans 1985, 1982). Furthermore, there is independent textual evidence making it clear that Husserl rejects strict descriptivism, which evidence comes from a source that we have already found to support Føllesdal's reading: Close relatives of 'that' when functioning substantively are 'this', 'that' when referring back towards something that has already been thought or named before as well as 'the same thing' when resuming something in an identifying manner and, quite generally, every nominal expression [jedes Nominal] belonging to the class of expressions which do not conceptually represent any substantial part of the objectualities they refer to ... (Hua XX/1, text # 2, § 5, 78; translation and emphasis mine)9 9

In German: "Mit dem selbständig substantivisch fungierenden 'das' nah verwandt ist das auf ein vordem Gedachtes und Genanntes zurückweisende 'dieses', 'jenes', ferner das identifizierende anknüpfende 'dasselbe' und so überhaupt jedes Nominale aus der Klasse derjenigen, welche nichts vom Sachgehalt der Gegenständlichkeiten, auf welche sie sich beziehen, begrifflich fassen ..." – See also Hua XX/1, p. 82: "Mit dem

85

Further textual evidence against the ascription of a strictly descriptivist view to Husserl is to be found, among other things, in a 1908 research manuscript where Husserl anticipates Keith Donnellan's distinction between referential and attributive use of a definite description and clearly defends the view that there are acts of direct, i.e. singular, reference (cf. Hua XXVI, suppl. XII, pp. 170 f.; see Beyer 2000, pp. 52-70; Beyer 2001).10 It is to the claim that the Fregean reading does not allow for an according rational reconstruction of Husserl's position that I now finally turn. No direct reference In his discussion of my Føllesdal-inspired reconstruction, Bianchin agrees that there is direct reference, and that Husserl appeals to the dynamic method in order both to specify its content and to determine its referent. However, he goes on to argue that these observations refute the "Fregean interpretation", on the ground that noematic Sinne à la Føllesdal are abstract, atemporal objects rather than dynamic cognitive structures of the required sort (cf. Biancin 2004, p. 221; see also Zahavi 2004, p. 48). My reply is that while Bianchin is, of course, quite right that Føllesdal interprets noematic Sinne as abstract objects (see his thesis 8), the problem referred to, concerning ontological status, arises for Frege's specific conception of Sinn as well, so that the parallel between Frege and Husserl is not destroyed at this – admittedly crucial – point. For, as Evans has explained in detail, Frege's considerations on indexicals in "The thought" lead to the conception of a "dynamic Fregean thought" (cf. Frege 1918, esp. p. 64), whose ontological status is as much in need of Eigennamen kommt der Gegenstand selbst in seiner Eigenheit, aber es kommt keine Eigenartigkeit, kein Wiebeschaffensein des genannten Gegenstands zum Ausdruck." 10 An act of reference is direct iff it is not indirect. A given act of reference i is indirect iff i displays "referential multiplicity" (Jaakko Hintikka), i.e. if there are at least two circumstances of evaluation w, w' and a property F-ness involved in the content of i such that (1) in w there is an object x which uniquely has F-ness in w, (2) in w' there is an object y different from x which uniquely has F-ness in w', (3) due to (1) x is to be regarded as the referent of i in w and (4) due to (2) y is to be regarded as the referent of i in w'. Cf. Beyer 2000, pp. 35 ff.

86

clarification as the status of noematic Sinn qua (sub-)propositional content and especially of the determinable X as a meaning-like entity. This problem needs to be solved, for otherwise we run the risk of relapsing into an unacceptable form of psychologism about (sub-)propositional content. Another objection in connection with direct reference concerns thesis 3, according to which conscious acts of reference refer "in virtue of" noematic Sinn. Here Zahavi quotes an unpublished manuscript where Husserl denies the representationalist view that mental acts indirectly refer to their "transcendent" objects "through" intra-mental ("immanent") noematic representations (Ms B III 12 IV, 82a; cf. Zahavi 2004, p. 49; see also Cobb-Stevens 1990, pp. 172 f.). However, the claim that something refers "by means" of its Sinn certainly does not entail such a representionalist view on reference. According to Frege, expressions refer to their referent "by means of", or "mediated by", sense in that the respective language user entertains the relevant sense in his (or her) thought when referring to that referent in using, or understanding, the respective expression; and entertaining a sense is by no means tantamount to referring to it. The same goes for Husserl's general conception of conscious reference if we interpret it in accordance with Føllesdal's theses 9 through 11*. I have already pointed out that Husserl treats, e.g., conscious demonstrative reference as singular and that this view fits in well with Føllesdal's thesis 12 (among other things) without leading to special problems that would not arise for Frege's conception of dynamic thought, too. All things considered, I thus conclude that while the critics rightly contend that on a proper reading Husserl's conception of noema does not commit him to either internalism, strict descriptivism or the rejection of the view that there is direct reference, they are wrong in claiming that these requirements speak against the Fregean reading.

87

References Beyer, Christian: Intentionalität und Referenz, Paderborn: mentis 2000 Beyer, Christian: "A neo-Husserlian theory of speaker's reference", in: Erkenntnis 54 (2001), pp. 277-297 Beyer, Christian: "Edmund Husserl", in: Edward N. Zalta, ed., Stanford Encyclopedia of Philosophy, http://plato.stanford.edu/entries/husserl/, Stanford 2003 ff. Biancin, Matteo: "Book review: Christian Beyer, Intentionalität und Referenz", in: Husserl Studies 19 (2003), pp. 217-224 Cobb-Stevens, Richard: Husserl and Analytic Philosophy, Dordrecht: Kluwer, 1990 Evans, Gareth: The Varieties of Reference, Oxford: Oxford University Press 1982 Evans, Gareth: "Understanding demonstratives", reprinted in: Collected Papers, Oxford: Clarendon 1985, pp. 291-321 Føllesdal, Dagfinn: "Husserl's notion of noema", reprinted in: Hubert L. Dreyfus (ed.), Husserl, Intentionality and Cognitive Science, Cambridge/Mass.: MIT Press 1982, pp. 73-80 Føllesdal, Dagfinn: "Gödel and Husserl", in: Jean Petitot, Francisco J. Varela, Bernard Pachoud and Jean-Michel Roy (eds.), Naturalizing Phenomenology, Stanford: Stanford University Press 1999, pp. 385-400 Føllesdal, Dagfinn: "Indikatoren und der Geist", in: Geert Keil and Udo Tietz (eds.), Phänomenologie und Sprachanalyse, Paderborn: mentis 2006, pp. 227-232 Frege, Gottlob: "Über Sinn und Bedeutung", in: Zeitschrift für Philosophie und philosophische Kritik NF 100 (1892), pp. 25-50 Frege, Gottlob: "Der Gedanke", in: Beiträge zur Philosophie des deutschen Idealismus 2 (1918), pp. 58-77 Husserl, Edmund: Husserliana, Edmund Husserl – Gesammelte Werke (= Hua), Den Haag/Dordrecht: Nijhoff/Kluwer/Springer 1959 ff. Mulligan, Kevin: "Perception", in: Barry Smith and David W. Smith (eds.), The

88

Cambridge Companion to Husserl, Cambridge: Cambridge University Press 1995, pp. 168-238 Perry, John: "The problem of the essential indexical", reprinted in: The Problem of the Essential Indexical, New York: Oxford University Press 1993, pp. 33-50 (= Perry 1993a) Perry, John: "A problem about continued belief", reprinted in: The Problem of the Essential Indexical, New York: Oxford University Press 1993, pp. 69-89 (= Perry 1993b) Smith, David W.: The Circle of Acquaintance, Dordrecht: Kluwer 1989 Smith, David W. and Ronald McIntyre: Husserl and Intentionality, Dordrecht: Reidel 1982 Zahavi, Dan: "Husserl's noema and the internalism-externalism debate", in: Inquiry 47 (2004), pp. 42-66

Transcendental Philosophy and Modern Physics: Neo-Kantianism, Logical Empiricism, and Phenomenology Michael Friedman*

Abstract In the heyday of logical empiricism it was conventional wisdom that modern physics—represented, at the time, by Einstein’s general theory of relativity—is incompatible with all forms of transcendental philosophy. More recently, however, scholarship on the origins of logical empiricism has revealed a significant debt to Kantian and neo-Kantian ideas in some of the early attempts to assimilate Einstein’s discoveries by the logical empiricists; and scholarship on the history of general relativity has emphasized the substantial influence of Husserlian transcendental phenomenology on Hermann Weyl’s Raum-ZeitMaterie (1918). This paper discusses the background to the various philosophical responses to Einstein’s work—neo-Kantian, logical empiricist, and phenomenological—in Kant’s original attempt to comprehend Newtonian physics; and, in light of this, it suggests a way in which the insights of all of these twentieth-century approaches may be fruitfully combined.

In the heyday of logical empiricism it was conventional wisdom that modern physics—represented, at the time, by Einstein’s general theory of relativity—was entirely incompatible with all forms of transcendental philosophy. After all, general relativity uses a non-Euclidean geometry to describe the space of the physical world, a circumstance that is clearly incompatible with the original Kantian idea that we know the Euclidean geometry of space a priori as the pure form of our outer intuition. Moreover, this use of non-Euclidean geometry in physics can easily suggest, as well, that all attempts to find a priori necessary structures in our experience of the physical world—not just the original Kantian attempt—are forever doomed to failure. The logical empiricists drew precisely this inference, *

Stanford University

90 and they concluded, accordingly, that it is impossible in principle to articulate a transcendental analysis of experience—to find something like the synthetic a priori in experience. On this basis, Moritz Schlick, the founder of the Vienna circle, famously polemicized against both contemporary neoKantianism and Husserlian phenomenology in his Allgemeine Erkenntnislehre (1918).1 More recently, however, scholarship on the origins of logical empiricism has revealed a significant debt to Kantian and neo-Kantian ideas in some of the earliest attempts to assimilate Einstein’s discoveries by the logical empiricists themselves. Perhaps the most striking such case is Hans Reichenbach’s first book, Relativitätstheorie und Erkenntnis Apriori (1920)—which, as its title suggests, explicitly attempts to preserve an important role for the Kantian a priori after all. Yet scientific a priori principles, for Reichenbach, are not, as in Kant himself, necessary and rigidly fixed for all time. Rather they vary from theory to theory as our science evolves (from Newtonian to relativistic physics) and are therefore a priori only in a relativized sense.2 I shall briefly return to Reichenbach and Schlick below. But I first want to note that recent work on the history of general relativity has also emphasized the substantial influence of Husserlian transcendental phenomenology on Hermann Weyl’s Raum-Zeit-Materie (1918) (first published in the same years as Schlick’s Allgemeine Erkenntnislehre). In particular, Ryckman (2005) explains how Weyl’s extensive immersion in Husserlian phenomenology gave rise to a deep investigation into the mathematical and philosophical foundations of geometry under the rubric of what Weyl called “purely infinitesimal geometry [rein infinitesimale Geometrie]”— which, in turn, was intended to reflect the philosophical primacy of that which is directly and immediately presented to the transcendental ego in the phenomenological here-and-now. On this basis, Weyl developed what he himself conceived as a constructive mathematical “essential analysis 1

This work, in turn, is importantly based on Schlick’s earlier treatment of Einstein’s general theory of relativity in Raum und Zeit in der gegenwärtigen Physik (1917). For discussion of the relationship between the two works see Friedman (1999, pp. 34-43), and compare also Friedman (1997). 2 For discussion of this work, and of Reichenbach’s relationship with Schlick, see Friedman (1999, chapter 3).

91 [Wesensanalyse]” of the nature of space (and of space-time), and he put this analysis to physical use in articulating, for the first time, the idea of a gauge transformation. So there is no doubt, in any case, that Weyl’s ambitious combination of mathematical, physical, and philosophical thinking provides a striking illustration of the wide-ranging intellectual fruitfulness of transcendental phenomenology. This paper discusses the background to the various philosophical responses to Einstein’s work—neo-Kantian, logical empiricist, and phenomenological—in Kant’s original attempt to comprehend Newtonian physics; and, in light of this, I suggest a way in which the insights of all of these twentieth-century approaches may be fruitfully combined. I suggest, in particular, that Husserlian transcendental phenomenology needs supplementation by ideas characteristic of the neo-Kantian tradition in order adequately to ground the notion of an “essential analysis.” This suggestion, of course, is very controversial, and so I look forward especially to Dagfinn Føllesdal’s reaction. The problem of a transcendental foundation for modern physics begins, therefore, with the work of Kant—and Husserl himself, in Die Krisis der europäischen Wissenschaften und die transzendentale Phänomenologie (1954), originally completed around 1936, rightly takes Kant to be the father of transcendental philosophy. For Kant, this project took the form of what he called a metaphysical foundation for Newtonian physics, in which the Newtonian concepts of absolute space, time, and motion were replaced by the schematized application of the pure concepts of the understanding to the pure intuitions of space and time. In particular, the categories of substance, causality, and community resulted in what Kant understood as the fundamental laws of Newtonian mechanics—the conservation of the total quantity of matter, the law of inertia, and the equality of action and reaction—which, in turn, provided an a priori basis for a procedure of “timedetermination” resulting in an empirical interpretation of the concepts of absolute space, time, and motion within our experience of the physical world. In modern terms, the Newtonian laws of motion define a class of privileged frames of reference—the so-called inertial frames of reference—which then give well-defined empirical meaning to the Newtonian

92 concept of absolute motion.3 I shall return to this point later, but I first want to step back from the details of mathematical physics to discuss the most general features of Kant’s own approach to transcendental philosophy. What is most distinctive of Kant’s approach is his sharp and fundamental distinction between two—at first entirely independent—faculties of the mind: the passive faculty of sensibility and the active faculty of understanding. Both faculties, for Kant, are sources of a priori knowledge, and both, in Kant’s sense, perform a transcendental function in a proper philosophical account of our empirical knowledge. Whereas the pure forms of sensibility, space and time, explain the possibility of pure mathematics and its application to the empirical or phenomenal world, the pure concepts or categories of the understanding (substance, causality, community, and so on) are initially derived from the logical forms of judgement (of traditional Aristotelian syllogistic) and thus reflect the a priori structure of the intellect. However, such purely intellectual concepts, for Kant, only have cognitive sense and meaning when they are applied to (schematized in terms of) the pure intuitions of space and time—and are thereby applied, in turn, to the empirical or phenomenal world of objects of experience (appearance) presented to our senses within space and time. Explaining this process of application in detail occupies the most central and difficult sections of the Kritik der reinen Vernunft—the transcendental deduction of the categories and the schematism—and it finally results, in the following sections, in the very general a priori “principles of pure understanding” governing all objects of possible experience in space and time. These general principles are then further specified or realized (in the Metaphysische Anfangsgründe, compare note 3 above) by what Kant calls the principles of 3

See the Introduction to my (2004) edition of Kant’s Metaphysische Anfangsgründe der Naturwissenschaft (1786) for more details. Briefly, Kant views the Newtonian laws of motion as rules for successively approximating, in experience, a privileged frame of reference he calls “absolute space.” This process begins from our parochial perspective here on earth, then moves to the center of mass of the solar system, then to the center of mass of the Milky Way galaxy, then to the center of mass of a system of such galaxies, and so on ad infinitum. Each step of this procedure thereby results, from a modern point of view, in a better approximation to an inertial frame of reference—one in which Newton’s laws of motion are exactly satisfied.

93 pure natural science, so that Kant’s transcendental explanation of the conditions of the possibility of experience is, at the same time, an answer to the question how pure natural science is possible. As is well known, however, virtually all post-Kantian transcendental philosophers of the nineteenth century, including the post-Kantian German idealists, took their starting point from an explicit rejection of Kant’s dualistic picture of the faculties of mind, and they explicitly rejected, in particular, Kant’s central philosophical problematic of showing how a purely intellectual faculty of understanding—which, by itself, is “empty” and without any “relation to an object”—can then be somehow related to an independent faculty of sensibility.4 The most important such philosophers, for our purposes, were the Marburg neo-Kantians centered around the initial work of Hermann Cohen in the late nineteenth century; for these philosophers, like Kant, took modern mathematical physics as a central object of transcendental investigation, and they, again like Kant, understood the explanation of how pure natural science is possible as paradigmatic of transcendental argumentation. In rejecting any independent faculty of intuition, however, they developed what they called a genetic [erzeugende] conception of knowledge, according to which the historical progress of the mathematical natural sciences is depicted as a never completed infinite sequence that is converging, as it were, on a never to be attained ideal limit. The actual empirical object of mathematical natural science is in no way passively given prior to or independently of the sequence in question; it is instead conceived as an infinitely distant ideal limit or X towards which the historical sequence of successive mathematical theories is converging— only then will the actual empirical object be constituted, not as “given [gegeben]” but rather “set as a task [aufgegeben].”5 Ernst Cassirer, in the early years of the twentieth century, developed an especially important version of this view, which he then applied, in particular, to the theory of general relativity. For Cassirer, late nineteenthand early twentieth-century developments in the foundations of mathematics and logic had led to a definitive replacement of the traditional Aristote4

For the post-Kantian German idealists on this point see Guyer (2000), and compare also Beiser (1987) (2002). 5 For Cohen and the Marburg School see Friedman (2000, chapter 5), and compare also Richardson (2006).

94 lian “abstractionist” theory of concept formation by a new mathematically informed “functional” theory. Cassirer begins from the modern mathematical conceptions of function and order, and, in particular, from the fundamental idea of an ordered mathematical sequence (such as the sequence of natural numbers). Such an ordered sequence (not the Aristotelian logical forms of judgement) represents the fundamental “functional form” of the intellect; and this kind of functional form (as again opposed to the logical forms of Aristotle) requires no schematization of the intellect in terms of a passive faculty of sensibility in order to constitute the objects of experience. On the contrary, we can depict the indefinitely extended historical sequence by which the empirical object of natural science is constituted as precisely such an ordered sequence, with no need to “coordinate [zuordnen]” this sequence to anything “given” from outside it. All coordination [Zuordnung] rather consists in a purely mathematical functional relation or mapping in virtue of which one stage of the historically progressing sequence of mathematical structures is subsumed by or embedded within a later stage.6 The latest such stage, circa 1920, was the new mathematical structure for physical space-time articulated in Einstein’s general theory of relativity; and Cassirer applied these ideas to Einstein’s theory in his Zur Einsteinschen Relativitätstheorie. Erkenntnistheoretische Betrachtungen (1921). This work brought Cassirer into close proximity with the work of the early logical empiricists, especially Moritz Schlick and Hans Reichenbach, who were simultaneously placing this new theory of Einstein’s at the very center of their philosophical projects. Schlick (1917) was a particularly influential discussion (compare note 1 above), and Reichenbach (1920) defended a relativized version of Kant’s original physical a priori. Reichenbach and Cassirer had seen pre-publication copies of one another’s books, and they both mention one another in their published texts. Schlick and Cassirer were familiar with one another’s work as well, and Schlick

6

Cassirer’s version of the genetic conception of knowledge is developed most fully in Substanzbegriff und Funktionsbegriff: Untersuchungen über die Grundfragen der Erkenntniskritik (1910). For further discussion see again Friedman (2000, chapter 5), and compare also Ryckman (1991).

95 published a critical study of Cassirer’s book in Kantstudien in 1921.7 Although Reichenbach was far more sympathetic (in 1920) to Cassirer’s neoKantian project,8 both Reichenbach and Schlick developed general conceptions of mathematical-physical knowledge which, in important respects, are fundamentally opposed to the version of the genetic conception of knowledge (and thus the version of transcendental idealism) articulated by Cassirer. In particular, the notion of coordination [Zuordnung] is basic to the conceptions of both Schlick and Reichenbach, just as it is for Cassirer, but the two logical empiricists have a completely different notion in mind. For Schlick, a theory in mathematical physics consists of a purely abstract mathematical structure—defined, in the manner of David Hilbert’s axiomatization of geometry, by a system of purely formal “implicit definitions”—which must then be somehow related to or coordinated with an independently existing empirical reality. This comes about, for Schlick, when an abstract mathematical structure is related to or coordinated with a sequence of subjective experiences consisting of directly given objects of “acquaintance [kennen]”—as opposed to objects of “knowledge [erkennen].” Indeed, as mere given objects of acquaintance, the sequence of experiences in question is wholly unconceptualized, consisting merely of fleeting and subjective “images [Vorstellungen]” as opposed to stable and objective “concepts [Begriffe].” What Schlick calls coordination therefore consists, in the end, of a purely formal mapping between an abstract mathematical structure, on the one side, and something wholly unconceptualized and ineffable, on the other.9 Reichenbach, too, begins from the Hilbertian picture of purely formal implicit definitions, and he also defines empirical knowledge (the application of such a purely formal system to existing physical reality) in terms of a procedure of coordination—a coordination or mapping which, in Reichenbach’s view, relates a precisely defined but purely abstract mathematical structure to something else, empiri7

See “Kritizistische oder empiristische Deutung der neuen Physik?” (1921). For further discussion see Friedman (2000, pp. 114-117). 8 Friedman (1999, chapter 3) explains how Schlick eventually persuaded Reichenbach to give up his neo-Kantian perspective. 9 For further details see again Friedman (1999, pp. 34-43), (1997).

96 cal reality, which is itself simply undefined prior to the coordination in question.10 For both Schlick and Reichenbach, then, knowledge of concrete empirical reality essentially involves a coordination with something undefined and ineffable: a something we know not what, a Ding an sich. Such a conception is of course entirely opposed to Cassirer’s version of transcendental idealism—and, I would argue, to any coherent version of transcendental philosophy. By contrast, however, it would appear that Husserlian phenomenology is in a much better position than either early logical empiricism or Cassirer’s version of the genetic conception of knowledge to do justice to the empirical application of mathematical physics, since Husserl takes his starting point (most explicitly, of course, in the Krisis) from the concrete immediately and perceptually given life-world constituting the transcendental basis of all human experience and all positive science. In particular, there is no fundamental dualism between sensibility and intellect in Husserl’s picture,11 and, unlike in Cassirer (at least in his earlier, properly scientific works),12 the sensible and intuitive dimensions of experience are placed at the center rather than the periphery.13 10

This view is developed in the fourth chapter of Reichenbach (1920), entitled “Knowledge as Coordination [Erkenntnis als Zuordnung].” 11 It is fundamental to Husserl’s conception of “essential insight [Wesenserschauung]” that all of our knowledge, both a priori and empirical, both universal and particular, is fundamentally intuitive. And it is for this reason, above all, that Schlick (1918, § 17) includes a polemical dismissal of just this Husserlian conception: because of Schlick’s stark opposition between conceptual thought and intuitive experience, the notion of “intuitive knowledge” or “knowledge by acquaintance” is simply a “contradictio in adjecto” (1918, § 11). This polemical dismissal was removed, as Schlick explains, from the corresponding discussion in the second (1925) edition—now § 18. 12 Cassirer’s later “philosophy of symbolic forms” (which he began to develop around the same time as his 1921 work on general relativity) takes the genetic conception of scientific knowledge characteristic of the Marburg School as representing only one (particularly abstract and non-intuitive) form of symbolic meaning, which, in turn, can only be erected on the basis of the more intuitive forms embodied in ordinary language and sense perception. For further discussion see again Friedman (2000, chapter 6), and compare also Krois (1987). 13 An Appendix to the second chapter of Ryckman (2005) considers my recent revival of Reichenbach’s (1920) version of the “relativized a priori,” and Ryckman (2010) rightly criticizes me for relying too uncritically on the Schlickean and Reichenbachian

97 Now, as I noted at the beginning, Hermann Weyl implements Husserlian phenomenology in the foundations of general relativity by means of a sophisticated constructive mathematical development of what he calls purely infinitesimal geometry. This constructive “essential analysis” of the concept of space (and of space-time) is mathematically quite deep; but I do wonder whether Weyl’s purely mathematical investigation can provide a complete philosophical answer to the problem addressed, in rather different ways, by Schlick, Reichenbach, and Cassirer (and, in an earlier and less complicated time, by the original transcendental idealism of Kant)—the problem, namely, of how modern mathematical physics achieves a concrete empirical meaning in terms of the sensibly given objects of our perceptual experience. What I would like to do, in the remainder of this paper, is to consider an alternative way of approaching this question and to consider the relationship, in particular, between this approach and transcendental phenomenology. The approach in question has recently been suggested by Scott Tanona (in a context that was not explicitly engaged with transcendental phenomenology), and the basic idea is to emphasize the role of concrete and perceptually defined frames of reference in the empirical application of our abstract mathematical theories of space, time, and motion.14 Indeed, we saw at the beginning that such a consideration of concrete empirical reference frames was already necessary in the context of Newton’s original theory, since Newtonian absolute space, on Newton’s own account, is by no means presented to our senses. Kant’s version of transcendental idealism was fundamentally engaged with this question; and, as I already pointed out, the modern (late nineteenth-century) version of Kant’s solution is notion of “coordination.” The present discussion of the application of mathematics in modern physics involves a response to Ryckman’s work, and, accordingly, it relies on material originally appearing in Friedman (2010). I am indebted to André Carus, the Hegeler Institute, and Open Court Publishing Company, for permission to reprint this material here. 14 See Tanona (2010)—which, among other things, also rightly criticizes some of my own earlier work for relying too uncritically on the Schlickean and Reichenbachian notion of “coordination” (compare note 13 above). As Tanona points out, his approach is closely related to important features of Niels Bohr’s approach to the foundations of quantum mechanics, as discussed in Tanona (2004).

98 that the Newtonian laws of motion serve to characterize or define a class of privileged frames of reference, the so-called inertial frames of reference, which can then be concretely given (or at least concretely approximated) in our actual perceptual experience of the empirical world. Thus, the familiar “laboratory frames” attached to the surface of the earth provide us with good approximations to an inertial frame, at least for the phenomena standardly considered therein; but, for the purposes of astronomical science, it is necessary to move to a more exact frame of reference defined by the center of mass of the solar system (for it is here, and here alone, that the Newtonian description of this system by the theory of universal gravitation is valid). This astronomical frame of reference—which, in particular, is a better approximation to an inertial frame than our initial laboratory frames—is certainly quite distant from our sensible perceptual experience here on earth (and thus more distant, as it were, from the Husserlian life-world), but it is nonetheless reachable from our initial experience here on earth by a well-defined series of perceptual and conceptual operations (including, of course, mathematical operations).15 Now Einstein’s development of the general theory of relativity involved a series of natural extensions of the concept of an inertial frame of reference—a concept originally developed to account for the empirical meaning of the Newtonian theory of space, time, and motion in the late nineteenth century.16 The special theory of relativity, developed in 1905, gave a central place to this concept, and it then proceeded to adapt this concept to the new empirical discovery of the invariance of the velocity of light—a discovery that stood in contradiction, in particular, to the way in which inertial frames having different velocities were assumed to be relat15

Thus, for example, we might take the operations in question to comprise the transition from geocentric to heliocentric astronomy initiated by Copernicus, Galileo’s development of a non-Aristotelian terrestrial physics based, in effect, on the consideration of a “laboratory frame” attached to the surface of the earth, and Newton’s synthesis of heliocentric (now Keplerian) astronomy and Galilean terrestrial physics. For Kant’s conception of how such a sequence of operations may then be conceived within Newtonian physics see again note 3 above, together with the paragraph to which it is appended. 16 For discussion of the late nineteenth-century development of the concept of an inertial frame see Torretti (1983, § 1.5), DiSalle (1991), (2002).

99 ed to one another in the context of Newtonian theory (by the Galilean transformations).17 Einstein resolved this contradiction by articulating a new theoretical structure for describing space, time, and motion—the structure of what we now call Minkowski space-time—wherein the inertial frames of Newtonian theory (compare note 17 above) are now necessarily related to one another in a fundamentally different way (by the Lorentz transformations). Nevertheless, this new space-time structure is still related to the empirically given phenomena observable in inertial frames of reference, and these frames, in turn, are related to our sensible perceptual experience of the life-world in a way precisely analogous to the situation in Newtonian theory.18 Yet the well-known empirical phenomenon of gravitation—as described by the Newtonian theory of universal gravitation—could not be easily reconciled, in turn, with the new spatio-temporal theoretical structure (what we now call the structure of Minkowski space-time). For Newtonian universal gravitation acts instantaneously at a distance across arbitrarily large spatial distances (across the solar system, for example), and it thus involves, at the level of space-time structure, a relation of absolute simultaneity—precisely the relation now rejected by Einstein’s new theory. Einstein therefore embarked on the project of developing a field theory of gravitation compatible with the new space-time structure, one in which gravitational interaction would propagate with the velocity of light. The 17

Einstein implicitly appeals to the late nineteenth-century notion of an inertial frame in 1905 by beginning with “a system of coordinates in which the equations of Newtonian mechanics are valid.” In a later (1913) reprinting, Einstein adds a footnote with the qualification that “the equations of Newtonian mechanics” are here assumed to be valid only approximately—since, of course, the laws of Newtonian mechanics are no longer precisely valid in special relativity. Einstein now, in effect, is defining the inertial frames as those in which light has the invariant velocity c. 18 Thus, the sequence of operations of note 15 above is now extended one step further, where the Newtonian inertial frames, in particular, are now seen to be valid only approximately. As Tanona (2010) emphasizes, the famous Michelson-Morley experiment (which showed, in effect, that light has the same velocity c in all inertial frames) was carried out in a classical “laboratory frame” attached to the surface of the earth (but capable of being rotated into different orientations), and it thus presupposes that the Newtonian concept of an inertial frame is suitable to a sufficient degree of accuracy for the phenomena considered therein.

100 theory of general relativity, published in 1915-16, represents Einstein’s solution; and, in this theory, the classical law of inertia—that free particles affected by no external forces traverse Euclidean straight lines with constant or uniform speed (a law which remains equally valid in the special theory of relativity)—is no longer valid. It is replaced by a new principle, the principle of equivalence, according to which freely falling particles acted on only by gravitation traverse straightest possible four-dimensional trajectories or geodesics in a variably curved perturbation of Minkowski space-time. As a consequence, inertial reference frames (which are supposed to be themselves moving uniformly and rectilinearly with respect to one another) no longer exist in the new space-time structure of the general theory. Nevertheless, there are still local inertial frames, which approximate, in small regions and appropriate conditions, the global inertial frames of our previous theories.19 So in this way, in particular, there is still a well-defined perceptual and conceptual route leading from our actual experience of the life-world here on earth to the empirical application (to such phenomena as light, gravitation, and so on) of our ever more abstract mathematical theoretical structures. This sketch of how the abstract mathematical structure of general relativity successively acquires its empirical application and meaning is analogous, as I have already begun to suggest, to Husserl’s well-known account, in the Krisis, of the origins of the mathematical science of geometry from the life-world. For Husserl there portrays the development of geometry in terms of a twofold sequence of successive idealizations: the idealization or precisification of the inexact spatial forms given in ordinary perceptual experience by the ideal and exact figures studied in pure geometry, and a parallel idealization or precisification of the empirical practice of spatial measurement (as in surveying, for example). Indeed, all exact science, for Husserl, develops out of the life-world by an indefinitely extended sequence of idealizations or successive approximations, a process that results in successive layers of “sedimentation” which then become covered over or forgotten by the later practitioners of the sciences in question. Phenomenologically guided historical reconstruction of this process then allows us 19

For the concept of a local inertial frame in general relativity see Torretti (1983, pp. 136-7).

101 to recover the original intentional meaning of these sciences in terms of perceptually and ostensively given life-world structures—and, in particular, to thereby comprehend and reaffirm the necessary origin of all positive science in our original transcendental subjectivity. What I have just suggested, in the above sketch of the development of general relativity, is that the question of the empirical application and meaning of our most abstract mathematical scientific theories can be profitably addressed in essentially the same way.20 But, if this is correct, it seems to me that there is more room for the kind of historical reconstruction of the development of the exact sciences practiced by Cassirer (and the Marburg School) in a phenomenological account of the practice of modern mathematical physics than may have at first been supposed. Of course such an account, in contrast to Cassirer’s, would give much more weight to immediate perceptual experience, and it would give much more weight, in particular, to ostensively and indexically available perceptual structures (such as reference frames, centered on a here-and-now). Nevertheless, historical reconstruction, aimed at uncovering successive layers of “sedimentation,” also operates, necessarily, with non-ostensively available transcendental structures; and it would appear thereby to involve regressive transcendental argumentation of precisely the kind favored by the Marburg School—where we begin with a “fact of science” and then seek for the necessary transcendental preconditions of this fact. Since Husserl, in the Krisis, explicitly rejects such “regressive” transcendental argumentation as merely hypothetical and insufficiently grounded intuitively, the problem of integrating such reconstructive historical accounts with the practice of Husserlian phenomenology is delicate indeed.21 Perhaps the kind of historical reconstruction envisioned here, which integrates the development of our most abstract mathematical scientific theories with a parallel tracing out of their successive empirical appli20

For the sequence of idealizations or precisifications in question see notes 15, 17, 18, and 19 above. 21 See Husserl (1954, §§ 28-32) for the rejection of (what he takes to be) Kant’s own regressive method. Husserl (1954, § 25) portrays this method as one that attempts to explain how mathematics and the objective sciences are necessarily applicable to the world of ordinary experience by depicting this world as the outcome of a construction produced by the two underlying faculties of “pure intuition” and “pure reason.”

102 cations all the way back to the originally given life-world, can contribute to a solution of this problem—and perhaps we can thereby envision an integration of Marburg neo-Kantianism and Husserlian phenomenology into a new type of transcendental philosophy combining the strengths of them both.22 This, at any rate, is my suggestion, and I will conclude by simply stating, as clearly as I can, why I think that this kind of supplementation of Husserlian phenomenology is needed. Why, more specifically, can we not conceive transcendental phenomenological analysis as purely (and directly) descriptive in Husserl’s sense, and thereby dispense entirely with all regressive transcendental argumentation? The answer, as I see it, is that phenomenological analysis is essential analysis—it is supposed to capture the essence or a priori formal structure of various features of consciousness, not simply to record the particular individualities of which our consciousness is full. Our target, to put it another way, is pure rather than merely empirically given consciousness. But how do we know when we have got hold of a pure, essential, and universal element of consciousness? The only answer that makes sense to me is that we can do this by starting with something that already embodies pure, universal, and a priori knowledge—pure mathematical knowledge.23 Thus, in the case of a phe22

Thus, as we have seen, the view that mathematical physics involves an indefinitely extended historically given sequence of successive idealizations is characteristic of the Marburg genetic conception of knowledge. What I am now envisioning, then, is that we can erect such a sequence from a starting point in the originally given life-world, just as Husserl himself portrays the development of mathematical geometry. Our consideration of the relationship between abstract mathematical theory and concrete empirically given frames of reference is the means for implementing this idea. 23 From this point of view, therefore, it is by no means surprising that the early sections of Husserl (1913) explicitly rely on geometrical and mechanical examples when first introducing the notions of “essential insight [Wesenserschauung]” and “eidetic universal validity [eidetischer Allgemeingültigkeit]”—and also present a conception of the relationship between essential or “eidetic” truths and empirical truths explicitly modelled on Kant (via the Kantian examples “all bodies are heavy” and “all material things are extended” discussed in § 7). Empirical knowledge is thereby seen to be framed by both material eidetic sciences (such as geometry and rational mechanics) and formal eidetic sciences (including the theory of manifolds), and it is in precisely this way that we find both universality and necessity within our intuitive experience.

103 nomenological analysis of modern physics, we need to start with clear examples of the application of pure mathematical knowledge to actual physical and empirical phenomena. Such mathematical knowledge, as Husserl emphasizes in the Krisis, is itself transcendentally grounded in the nonmathematical structures of the life-world, and so a phenomenological analysis of the life-world, in Husserl’s technical terminology, is descriptive rather than exact.24 But it is only the circumstance that the life-world, in turn, necessarily gives rise to structures of consciousness that are exact (i.e., mathematical structures) which then enables us to give a pure, essential, and universal description of the life-world. What makes a pure, essential, and universal description (that is, a transcendental description) of the life-world possible, in other words, is precisely the circumstance that pure and applied mathematical knowledge is thereby transcendentally grounded.25 Or so at least it seems to me. 24

This distinction is developed in Husserl (1913, §§ 71-75). Basically, exact eidetic sciences are mathematical (and are characterized, in particular, by mathematical precision), while descriptive eidetic sciences—such as transcendental phenomenology—are not. In terms of the terminology developed in the Krisis, the latter sciences describe the immediately given life-world itself, while the former describe the positive or objective sciences that then (necessarily) develop out of the life-world by an indefinitely extended sequence of idealizations and precisifications. 25 Why, for example, can we say that the originally given life-world is necessarily and essentially characterized by intuitive “subjective-relative” spatiality and temporality (the structure of the I-here-now) and a system of subjective “anticipations” (of what I may expect in the immediate future under certain conditions)? Of course it is “evident” to me that my life-world is so structured—and I assume it is “evident” to others as well. But how do I know that this structure is necessarily and essentially present in every possible human subject? The answer, as I see it, is that objective mathematicalphysical space and time necessarily develop out of subjective-relative spatiality and temporality, objective mathematical-physical laws out of subjective-relative sets of “anticipations.” And objective mathematical-physical space and time (together with the fundamental laws of physics) are paradigmatic of structures having genuine universal validity for all. Note, finally, that this point of view need not conflict with Husserl’s rejection of Kant’s regressive method as Husserl himself understands it (compare note 21 above), for we do not view the originally given life-world as the product of a “hypothetical construction” by our rational faculties, but rather as the necessary originally given starting point for the eventual (and complementary) construction of objective scientific knowledge (compare note 22 above).

104 References Beiser, F. C. (1987). The Fate of Reason: German Philosophy from Kant to Fichte. Cambridge, Mass.: Harvard University Press. Beiser, F. C. (2002). German Idealism: The Struggle against Subjectivism, 1781-1801. Cambridge, Mass.: Harvard University Press. Cassirer, E. (1910). Substanzbegriff und Funktionsbegriff: Untersuchungen über die Grundfragen der Erkenntniskritik. Berlin: Bruno Cassirer. Translated by W. & M. Swabey (1923) as Substance and Function. Chicago: Open Court. Cassirer, E. (1921). Zur Einsteinschen Relativitätstheorie: Erkenntnistheoretische Betrachtungen. Berlin: Bruno Cassirer. Translated by W. Swabey and M. Swabey (1923) as Einstein’s Theory of Relativity. Chicago: Open Court. DiSalle, R. (1991). “Conventionalism and the Origins of the Inertial Frame Concept.” PSA 1990 2: 139-147. DiSalle, R. (2002). Reconsidering Ernst Mach on Space, Time, and Motion. In D. Malament, ed. Reading Natural Philosophy: Essays in the History and Philosophy of Science and Mathematics. Chicago: Open Court, pp. 167-191. Friedman, M. (1997). “Helmholtz’s Zeichentheorie and Schlick’s Allgemeine Erkenntnislehre: Early Logical Empiricism and its Nineteenth-Century Background.” Philosophical Topics 25: 19-50. Friedman, M. (1999). Reconsidering Logical Positivism. Cambridge: Cambridge University Press. Friedman, M. (2000). A Parting of the Ways: Carnap, Cassirer, and Heidegger. Chicago: Open Court. Friedman, M., ed. (2004). Immanuel Kant: Metaphysical Foundations of Natural Science. Cambridge: Cambridge University Press. Friedman, M. (2010). “Synthetic History Reconsidered.” In M. Domski and M. Dickson, eds. Synthesis and the Growth of Knowledge. Essays at the Intersection of History, Philosophy, Science, and Mathematics. Chicago: Open Court, pp. 571-813.

105 Guyer, P. (2000). “Absolute Idealism and the Rejection of Kantian Dualism.” In K. Ameriks, ed. The Cambridge Companion to German Idealism. Cambridge: Cambridge University Press, pp. 37-56. Husserl, E. (1913). Ideen zu einer reinen Phänomenologie und phänomenologischen Philosophie. Halle: Max Niemeyer. Translated (from the augmented 1950 edition) by F. Kersten (1983) as Ideas Pertaining to a Pure Phenomenology and Phenomenological Philosophy. First Book. The Hague: Martinus Nijhoff. Husserl, E. (1954). Die Krisis der europäischen Wissenschaften und die transzendentale Phänomenologie: Eine Einleitung in die phänomenologische Philosophie. Ed. W. Biemel. The Hague: Martinus Nijhoff. Translated by D. Carr (1970) as The Crisis of European Sciences and Transcendental Phenomenology. An Introduction to Phenomenological Philosophy. Evanston: Northwestern University Press. Krois, J. M. (1987). Cassirer: Symbolic Forms and History. New Haven: Yale University Press. Reichenbach, H. (1920). Relativitätstheorie und Erkenntnis Apriori. Berlin: Springer. Translated by M. Reichenbach (1965) as The Theory of Relativity and A Priori Knowledge. Berkeley and Los Angeles: University of California Press. Richardson, A. (2006). “‘The Fact of Science’ and Critique of Knowledge: Exact Science as Problem and Resource in Marburg Neo-Kantianism.” In Friedman and Nordmann, eds. The Kantian Legacy in Nineteenth-Century Science. Cambridge, Mass.: MIT Press, pp. 211-226. Ryckman, T. A. (1991). “Conditio Sine Qua Non: Zuordnung in the Early Epistemologies of Cassirer and Schlick.” Synthese 88: 57-95. Ryckman, T. A. (2005). The Reign of Relativity: Philosophy in Physics 1915-1925. Oxford: Oxford University Press. Ryckman, T. A. (2010). “The ‘Relativized A Priori’: An Appreciation and a Critique.” In M. Domski and M. Dickson, eds. Synthesis and the Growth of Knowledge. Essays at the Intersection of History, Philosophy, Science, and Mathematics. Chicago: Open Court, pp. 455-470. Schlick, M. (1917). Raum und Zeit in der gegenwärtigen Physik. Berlin: Springer. Translated by H. L. Brose (supplemented by P. Heath) as Space and Time in Contem-

106 porary Physics. In H. Mulder and B. van de Velde Schlick, eds. (1978). Moritz Schlick: Philosophical Papers. Vol. 1. Dordrecht: Reidel, pp. 207-269. Schlick, M. (1918). Allgemeine Erkenntnislehre. Berlin: Springer. Translated (from the 1925 2nd ed.) by A. Blumberg (1985) as General Theory of Knowledge. La Salle: Open Court. Schlick, M. (1921). “Kritizistische oder empiristische Deutung der neuen Physik?” Kant-Studien 26: 96-111. Translated by P. Heath as “Critical or Empiricist Interpretation of Modern Physics?” In H. Mulder and B. van de Velde-Schlick , eds. (1978). Moritz Schlick: Philosophical Papers. Vol. 1. Dordrecht: Reidel, pp. 322-334. Tanona, S. (2004). “Uncertainty in Bohr’s Response to the Heisenberg Microscope.” Studies in History and Philosophy of Modern Physics 35: 483-507. Tanona, S. (2010). “Theory, Coordination, and Empirical Meaning in Modern Physics.” In M. Domski and M. Dickson, eds. Synthesis and the Growth of Knowledge. Essays at the Intersection of History, Philosophy, Science, and Mathematics. Chicago: Open Court, pp. 423-454. Torretti, R. (1983). Relativity and Geometry. New York: Pergamon. Weyl, H. (1918). Raum-Zeit-Materie. Vorlesungen über allgemeine Relativitätstheorie. Berlin: Springer. Translated (from the 1921 4th ed.) by H. L. Brose (1923) as Space-time-matter. London: Methuen.

Hume’s Phenomenological Conception of Space, Time and Mathematics Graciela De Pierris*

Abstract Hume’s Treatise, Book One, Part Two, contains perplexing arguments against the infinite divisibility of space (and time) and the exactitude of geometry. We can attain a charitable reading of these difficult arguments by appreciating that they are guided by Hume’s phenomenological model of apprehension, and that this model (unlike Descartes’s, for example) is intended to be sensible as opposed to intellectual. From the point of view of pure mathematics, Hume’s assumptions about the infinite may appear as crude misunderstandings of the continuum. However, if we appreciate that phenomenological apprehension, for Hume, must always be of bounded, discrete, sensible particulars, we can explain not only Hume’s arguments against infinite divisibility but also the privileged role he assigns to arithmetic (the mathematics of discrete quantity) and to one-to-one correspondence between discrete units as the ultimate standard of exactitude.

At T 1.2,1 Hume argues that neither space nor time can be infinitely divisible – and, indeed, that to suppose that space and time are infinitely divisible is absurd, impossible, and contradictory. The most notorious such argument, presented at T 1.2.2, traces back to Zeno’s metrical paradox of extension. If any finite interval of space (or time) is infinitely divisible, then it must consist of an infinite number of ultimate parts. However, these ultimate parts, when added together, must then result in an extension that is infinitely great, contrary to the supposition that it was a finite interval with which we started. From the point of view of pure mathematics, there is an *

Stanford University All citations from David Hume’s A Treatise of Human Nature (abbreviated Treatise or T) are from Hume (2000), and thus include the book, part, section, and paragraph numbers. 1

108 obvious objection to this argument: for, if we divide a finite interval into an infinite sequence of decreasing finite parts, their sum when added together can still be finite (as in the sum 1/2 + 1/4 + 1/8 + . . . = 1).2 Antony Flew3 calls attention to the crucial role for Hume of two principles regarding infinity. The first concerns the limited capacity of the human mind to apprehend the idea of the infinite; the second, which leads to the paradox just mentioned, states that whatever could be capable of being divided ad infinitum would have to consist of an infinite number of parts – which, in turn, would add up to something infinite. Flew offers a correct interpretation of Hume’s first principle, but ignores Hume’s distinctive epistemological approach in his uncharitable criticism of Hume’s second principle. In my view, Hume’s two principles about infinity reveal the constraints of Hume’s sensible phenomenological model of apprehension. When we place Hume’s doctrines within this model, we can appreciate that obvious and correct criticisms issuing from a purely mathematical conception of the infinite do not directly affect Hume’s discussion. Concerning the first principle, Hume writes: ’Tis universally allow’d, that the capacity of the mind is limited, and can never attain a full and adequate conception of infinity: And tho’ it were not allow’d, ’t wou’d be sufficiently evident from the plainest observation and experience. (T 1.2.1.2)4 2

Aside from mathematical objections, many interpreters have regarded Hume’s arguments about space, time and mathematics at Treatise, Book One, Part Two as generally confused. Norman Kemp Smith (1960), Chapter XIV and Appendices to Chapter XIV, offers a helpful discussion and for the most part attempts to understand Hume’s views about space and time sympathetically. However, Kemp Smith thinks that there are serious defects in this discussion due to Hume’s “crudely mechanistic psychology”: that Hume’s positive teaching that space and time consist of ultimate indivisible minima is “one of the least satisfactory parts of his philosophy” (p. 287). As I shall argue, on the contrary, Hume’s views on space, time, and mathematics are plausible when we acknowledge that they derive from his phenomenological model of apprehension of items before the mind. 3 Antony Flew (1976). Flew’s classic article has influenced most commentators on the topic. 4 In my view, contrary to Flew, Hume’s claim that this is “universally allowed” is not arbitrary. It is likely that Hume here refers to the view (shared by Descartes and Locke) that our faculties cannot fully apprehend the actual (or completed) infinite. As

109 Flew rightly points out that to be able to form “a full and adequate conception” of anything is, for Hume, to be able to imagine it in the appropriate way, and that to imagine in this way is to form the appropriate mental image. This imagist notion of “conception” constitutes a central aspect of what I take to be Hume’s sensible phenomenological model of apprehension. The other aspect of this model, which should be added to Flew’s analysis, is that images are fully particular and discrete. That they are discrete is dramatically emphasized in the Treatise’s discussion of our idea of personal identity: our mental life ultimately consists of a series of distinct, separable, interrupted, fleeting perceptions.5 That ideas are fully determined particulars is the view developed in the Treatise’s section on abstract ideas (T 1.1.7). In the sensible phenomenological model of apprehension, there is no notion of the universality of general concepts as in Leibniz, for example, nor is there an acquaintance with universals (“true and immutable natures”) as in Descartes. At most there are the particular relations, for example resemblances, which we are able to pick out when we acquire the habit of associating a particular word with resembling particular images in the formation of abstract ideas.6 Moreover, in Hume’s sensible phenomenological model of time, each component (or “moment”) of time is filled in with a discrete phenomenological presentation, and any progression in a series of discrete phenomenological presentations must be a progression occurring in a finite and discrete number of steps or moments of time, each of which is also finite and discrete.7 we shall see, Hume goes a step further than his predecessors; for, his radicalized sensible phenomenological model of apprehension commits him to the rejection of the possibility of attaining a “full and adequate conception” of both the actual and the potential infinite. 5 In my view, this is also a central tenet of Hume’s skeptical arguments concerning causation, external objects, and a priori reasoning. 6 Thus, our cognitive activities ultimately reduce to, or are based on, the phenomenological apprehension of fully determined particular and discrete items (impressions, ideas or objects) and their relations, which are or have been present before the mind. 7 There is an interesting connection between this Humean conception of time and Descartes’s view that the divisions of time are separable (independent) from each other since there is no necessary connection between the moments of time. Thus, Descartes

110 Therefore, we can add the following more specific consequences of Hume’s first principle about infinity. First, our finite understanding cannot attain a particular and discrete phenomenological presentation of something actually infinite. Second, it cannot attain a particular and discrete image of each of an infinite number of parts supposedly composing a given extension. Third, and most importantly, it cannot attain particular and discrete images corresponding to even the potential infinity supposedly resulting from step by step additions or subtractions of particular and discrete parts of any given particular image. For, not only are we unable to traverse an infinite number of steps, but there is a phenomenological limit, for any given perceiver at any given time in any given context, to how much we can augment by adding, or diminish by dividing, such an image. Thus, no image corresponding to a potential infinity can be formed.8 These consequences of the first principle greatly help to clarify Hume’s second principle about infinity. Hume’s second fundamental principle reads: ’Tis also obvious, that whatever is capable of being divided in infinitum, must consist of an infinite number of parts, and that ’tis impossible to set any bounds to the number of parts, without setting bounds at the same time to the division. It requires scarce any induction to conclude from hence, that the idea, which we form of any finite quantity, is not infinitely divisible. (T 1.2.1.2)

According to Hume, it is absurd to suppose that any finite extension, such as a segment of a certain finite length, could be infinitely divisible. For, as we have seen, if it were thus divisible, it would consist of an infinite number of parts, and an infinite number of parts must add up to an inwrites, in the Fifth Set of Replies (to Gassendi’s objections to the Third Meditation): “[W]e are considering the time or duration of the thing that endures, and here you would not deny that the individual moments can be separated from those immediately preceding and succeeding them, which implies that the thing that endures may cease to be at any given moment” Descartes (1985), vol. 2, p. 255. Descartes makes the same point, for example, in article 21 of the Principles of Philosophy, Part One. I discuss the relationship between Hume’s and Descartes’s views on this issue in its application to demonstrative inference in De Pierris (2005). 8 In particular, Hume has no room for the apprehension of even the potential infinity of a series of discrete quantities, such as the number series in arithmetic.

111 finite extension. As I pointed out, from the point of view of pure mathematics, this is, of course, wrong, for there can be a sum of an infinite series of diminishing quantities that converges to a finite quantity, not to infinity. However, Hume’s claim that it is absurd or “contradictory” that a finite quantity be infinitely divisible follows from Hume’s sensible phenomenological notions of “part” or “consisting of parts,” and of “divisible” or “divisible into parts.” For the purpose of clarifying these Humean notions, let us consider Flew’s criticisms of Hume’s second principle. According to Flew, the second principle is flatly false for two reasons: First, and less importantly, to say that something is divisible into so many parts is not to say that it consist of – that it is, so to speak, already divided into – that number of parts. A cake may be divisible into many different numbers of equal slices without its thereby consisting in, through already having been divided into, any particular number of such slices. Second, and absolutely crucially, to say that something may be divided in infinitum is not to say that it can be divided into an infinite number of parts. It is rather to say that it can be divided, and sub-divided, and sub-sub-divided as often as anyone wishes: infinitely without limit. That this is so is part of what is meant by the saying: ‘Infinity is not a number!’ (Op. cit., pp. 259-60)

If we ignore the fact that Flew uses the example of a physical object, such as a cake, which is not infinitely divisible, we can concede that, from the point of view of pure geometry, the “parts” into which a whole region of space or the whole of space can be divided, are not conceived as actual, discrete, finite parts of equal size pre-existing the whole. We can also concede that the notion of infinite division can certainly refer, as in the tradition since Aristotle, to a potentially infinite division rather than to an actually completed division. Flew does not acknowledge, however, that Hume’s notions of “part” and “divisible” are not those of pure geometry. In considering Hume’s second principle about infinity, Flew should have taken the clue provided by his own correct interpretation of Hume’s first principle of infinity, namely, that to form “a full and adequate conception” means for Hume to form an image. More precisely, as I pointed out above, to form “a full and adequate conception” is to form a perceptible particular and discrete image copied

112 from a particular, discrete impression. The parts, wholes, divisions, and additions to which Hume refers are those which we can phenomenologically apprehend as sensory impressions or images of the imagination. In particular, Hume’s first positive conclusion after the two principles is that, if we undertake perceptible divisions of our ideas (images), the imagination always reaches minimum (simple) ideas which cannot be subdivided without annihilation; and this clearly shows that Hume interprets “divisible” and “consisting of parts” in terms of his sensible phenomenological model. Ideas or impressions consist of parts if and only if we can sensibly distinguish them as such through sensibly apprehended divisions – “part” means “perceptible part” and “division” means “perceptible separation of a perceptible whole into perceptible parts.” Hume’s discussion of our ideas of space and time reveals the sensible phenomenological character of his earlier distinction between simple and complex perceptions,9 namely, that whatever impressions, ideas or objects are different are distinguishable, and thus separable. This is Hume’s criterion of separability and thus, ultimately, the criterion of simplicity of ideas, impressions or objects. Any presentation must be regarded as simple and indivisible, if by attempting further to divide it – by attempting to distinguish and thus separate sensibly perceptible parts in it – we are left with no perceptible image before the mind: What consists of parts is distinguishable into them, and what is distinguishable is separable. But whatever we may imagine of the thing, the idea of a grain of sand is not distinguishable, nor separable into twenty, much less into a thousand, ten thousand, or an infinite number of different ideas. (T 1.2.1.3)

In the next paragraph Hume makes the same point now concerning impressions of sensation: ’Tis the same case with the impressions of the senses as with the ideas of the imagination. Put a spot of ink upon paper, fix your eye upon that spot, and retire to such a distance, that at last you lose sight of it; ’tis plain, that the moment before it vanish’d the image or impression was perfectly indivisible. (T 1.2.1.4)

9

Hume introduces this distinction at T 1.1.1.2.

113 Hume’s explanation of the modification of one’s visual field by the use of a microscope or a telescope might be read as making a realist point about the existence of minima waiting, so to speak, to be discovered: A microscope or telescope, which renders [the minima] visible, produces not any new rays of light, but only spreads those, which always flow’d from them; and by that means both gives parts to impressions, which to the naked eye appear simple and uncompounded, and advances to a minimum, what was formerly imperceptible. (T 1.2.1.4)

However, I believe Hume is here mainly interested, not in the question whether there are minima independently of what an observer can perceive, but in what, in a given context and under specific circumstances, a perceiver apprehends after a series of divisions, just before the impression or image is annihilated. In the last quoted text, Hume writes, “gives parts to impressions,” thus he refers to sensory impressions and their content or intentional object, not to independently existing external objects (whatever external objects might be). In the spot of ink example, the minima or simples are sensory impressions, which are the perceptible parts into which the complex sensory impression cannot be further subdivided without causing the perception in question simply to vanish. Similarly, in the example of the grain of sand, the minima or simples are ideas, which are the indivisible parts of a complex idea of an aggregate of grains of sand. Hume is indicating that the further empirical advance of the study of nature might always give us new minima – not because the minima are metaphysically simple and indivisible real entities, independent of the observer, but because we might always increase our capacity to perceive smaller and smaller entities. This is Hume’s contention after the paragraph where he has referred to the microscope and the telescope: We may hence discover the error of the common opinion, that the capacity of the mind is limited on both sides, and that ’tis impossible for the imagination to form an adequate idea, of what goes beyond a certain degree of minuteness as well as greatness. Nothing can be more minute, than some ideas, which we form in the fancy; and images, which appear to the senses; since these are ideas and images perfectly simple and indivisible. … This is however certain, that we can form ideas, which shall be no greater than the smallest atom of the animal

114 spirits of an insect a thousand times less than a mite. We ought rather to conclude, that the difficulty lies in enlarging our conceptions so much as to form a just notion of a mite, or even of an insect a thousand times less than a mite. For in order to form a just notion of these animals, we must have a distinct idea representing every part of them; which according to the system of infinite divisibility, is utterly impossible, and according to that of indivisible parts or atoms, is extremely difficult, by reason of the vast number and multiplicity of these parts. (T 1.2.1.5)

Hume is not here contradicting himself by denying that we must always reach an indivisible minimum. He is saying that some ideas are as small as can be imagined, simply because there is nothing smaller than minima sensibilia, but what counts as a sensible minimum is relative to what a given perceiver can actually reach at a given time and in a given context. Yet, despite this relativity, all minima, in a sense, have exactly the same “size” – the size of one indivisible single unit. Three theses follow from Hume’s identification of minima with phenomenological presentations that would disappear from one’s sensible phenomenological field if further divisions were undertaken. First, Hume’s notion of divisible into parts depends on an observer’s actual sensible apprehension of items before the mind, and on acts of separation or division performed by the observer in his/her own phenomenologically given field of images. Second, Hume’s characterization of minima reveals that he takes the minimum components of our complex ideas and impressions of space and time to be perceived as discrete parts or ultimate units, but only after the observer has performed the operation of dividing the sensory image or the idea of the imagination, and can go no further without annihilating the images. Third, the impression of a spot of ink or the idea copied from it is indivisible just exactly at the distance in one’s visual field before it disappears; but that exact distance varies in relation to different observers. Concerning a single observer, the threshold before annihilation (the minimum image) varies with the observer’s sensible capacities, the use of instruments such as a microscope or a telescope, internal features of the impressions and so on. Thus, Hume’s notion of divisibility into parts is perceiver-dependent and context-dependent in a way that explicitly leaves open the possibility of empirical advance and refinement.

115 In my view, “perceptible part,” for Hume, can mean either a minimum indivisible discrete atom or a complex sub-whole into which a given whole extension can be finitely divided without reaching the threshold of a minimum part. Nonetheless, the parts are always phenomenologically apprehensible images or impressions into which an extended or temporal compound whole can be finitely divided. If a whole can be divided at a given time, in a given context, into more than two simple, indivisible, discrete minima, the phenomenologically given sub-wholes are themselves composed of ultimate minima, and could be perceived as such if the observer proceeded with the divisions to reach the minima. However, Hume also suggests, in particular for space, that we are not separately aware of the minima that enter into the composition of, for example, a uniformly colored extension, before we have undertaken divisions. In this way, Hume acknowledges the appearance of continuity in our sensible apprehension of homogeneously colored or textured spatial extensions. The ultimate indivisible simple parts can be discovered to be such, only when a particular perceiver, in a particular context, actually undertakes distinctions, separations, and thus divisions. This now allows us to explain the precise sense, for Hume, in which the parts of space and time are pre-determined and pre-existing parts of equal size. In one of Hume’s senses of the notion of perceptible part, the latter are not pre-determined with respect to their size and do not pre-exist the whole. Progressive divisions of an extended or temporal whole allow us to find particular perceptible parts at each step in the subdivision; at certain stage in the progression, one is presented with sub-wholes, and nothing in Hume’s account precludes that the divisions into sub-wholes might result in sub-wholes of different sizes. However, with respect to the simple and indivisible minimal parts with which, according to Hume, such divisions must ultimately end, the situation is different. For, despite the fact that the minima are relative to a perceiver and his/her context and means of perception, there is, nonetheless, a sense in which Hume takes the minima to pre-exist the whole. For, the complex ideas of space and time, for Hume, themselves emerge solely from the perceived manner or order of composition of simple, indivisible minima. Thus, the minima can be taken to be determinate particular parts pre-existing the whole, because the whole can only be phenomenologically apprehended when we sensibly apprehend the

116 manner in which the minima appear. In the case of space, the manner of appearance is co-existence, in the case of time, the manner is succession: As ’tis from the disposition of visible and tangible objects we receive the idea of space, so from the succession of ideas and impressions we form the idea of time, nor is it possible for time alone ever to make its appearance, or be taken notice of by the mind. (T 1.2.3.7)

But now these ultimate minima, as we have seen, all have the same “size” – namely, that of one indivisible single unit. If an infinite number of these were to be added together, we would thus have an infinite sum of units – precisely as Hume’s main argument against infinite divisibility, deriving from Zeno’s metrical paradox of extension, contends. Indeed, Hume himself makes this very point in a footnote to his argument: It has been objected to me, that infinite divisibility supposes only an infinite number of proportional not of aliquot parts [i.e. parts of the same size], and that an infinite number of proportional parts [of different sizes] does not form an infinite extension. But this distinction is entirely frivolous. Whether these parts be call’d aliquot or proportional, they cannot be inferior to those minute parts we conceive [i.e. the minima]; and therefore cannot form a less extension by their conjunction. (T 1.2.2.2 n6)

There is thus no mathematical blunder here. Hume devotes a large portion of this part of the Treatise to answer the objection that demonstrations in geometry show the infinite divisibility of space. The outcome of this discussion is that geometry, unlike arithmetic, is not an exact science, because the demonstrations in geometry are not exact: [N]one of these demonstrations can have sufficient weight to establish such a principle, as this of infinite divisibility; and that because with regard to such minute objects, they are not properly demonstrations, being built on ideas, which are not exact, and maxims, which are not precisely true. When geometry decides any thing concerning the proportions of quantity, we ought not to look for the utmost precision and exactness. None of its proofs extend so far. It takes the dimensions and proportions of figures justly, but roughly, and with some liberty. Its errors are never considerable, nor wou’d it err at all, did it not aspire to such an absolute perfection. (T 1.2.4.17)

117 This is because geometers could only attain such an absolute perfection – an ideal exactness concerning proofs of dimensions and proportions of figures – if the exact number of points (minima) in each figure were known; but because of precisely the relativity and context dependence which, for Hume, is responsible for the appearance of continuity, this is impossible. Proofs of equality based on congruence fail for the same reason. When placing one figure upon the other, the supposition that we can determine whether all their parts correspond to and touch one another is fictitious. For we do not have a distinct notion of all the minute parts of the lines or figures compared. We could only know that congruence obtains, if we knew the exact number of minima in each of the objects compared, but we do not have a separate and distinct awareness of each of the minima composing extension. In the phenomenological appearance of extension, the minima are confused one with another, as Hume already claimed, for example, in the text at T 1.2.1.5 concerning the difficulties involved in enlarging our image so as to form a “just notion” of a mite. Later, in another part of the Treatise, where Hume is not discussing space and time, Hume summarizes his skeptical arguments concerning the exactitude of geometry and its demonstrations thus: I have already observ’d, that geometry, or the art, by which we fix the proportions of figures, tho’ it much excels both in universality and exactness, the loose judgments of the senses and imagination, yet never attains a perfect precision and exactness. Its first principles are still drawn from the general appearances of the objects; and that appearance can never afford us any security, when we examine the prodigious minuteness of which nature is susceptible. (T 1.3.1.4)

In the following paragraph, Hume explains that arithmetic provides the model of the method of enumeration of indivisible parts, namely, a one-to-one correspondence of discrete units. Indeed, precisely because it deals with discrete quantities, arithmetic, unlike geometry, can successfully apply this method in its demonstrations of equality and proportions of quantity: There remains, therefore, algebra and arithmetic as the sole sciences, in which we can carry on a chain of reasoning to any degree of intricacy, and yet pre-

118 serve a perfect exactness and certainty. We are possest of a precise standard, by which we can judge of the equality and proportion of numbers, and according as they correspond or not to that standard, we determine their relations, without any possibility of error. When two numbers are so combin’d, as that the one has always an unite answering to every unite of the other, we pronounce them equal; and ’tis for want of such a standard of equality in extension, that geometry can scarce be esteem’d a perfect and infallible science. (T 1.3.1.5)

The unextended minima composing extension, just like arithmetical units, are indivisible and have no parts. However, in the entire appearance they are presented confused or “confounded” with one another, and this results in the appearance of continuity. As we have seen, we can have an image of each of the minima constituting space only by embarking on successive acts of division of the complex extension: the minimum is the unit that appears just before another attempted division would annihilate it. Determining the total number of minima comprising a given whole of extension in this way is, for Hume, impossible. Hume’s sensible phenomenological model of apprehension therefore results in an internally coherent conception of geometry, algebra, and arithmetic, which reverses the more familiar conception of the early modern tradition in pure mathematics. According to the latter, geometry – the mathematics of continuous quantity – has an exact standard of congruence or equality of its own, which is necessarily more exact than any corresponding standard supplied by algebra and arithmetic – that is, the mathematics of number or discrete quantity. For many finite intervals, we can only approximate their length relative to a given unit by a never-ending sequence of numbers. For Hume, by contrast, there is no exact length of a continuous magnitude independent of discrete quantity. According to Hume’s model, phenomenologically presented (continuous) extension consists of a finite number of discrete indivisible units, and the sum of these units gives the exact magnitude of the extension in question. Yet, because of what Hume calls the “confounding” of these units in any given phenomenological field, this finite number is, for us, inaccessible; and the best we can do is use algebra and arithmetic to form more and more precise (but never completely precise) approximations. Only the sciences of discrete number can attain the ideal of complete exactness which geometry can only successively approximate but never actually reach.

119 References De Pierris, Graciela (2005). “Hume and Descartes on Skepticism with Regard to Demonstrative Reasoning,” Análisis Filosófico 25, No.2: 101-119. Descartes, Rene (1985). The Philosophical Writings of Descartes. John Cottingham, Robert Stoothoff, Dugald Murdoch, eds. Cambridge: Cambridge University Press. Flew, Antony (1976). “Infinite Divisibility in Hume’s Treatise.” In Donald Livingston and James King eds. Hume. A Re-evaluation. New York: Fordham University Press, pp. 257-269. Hume, David (2000). A Treatise of Human Nature. David Fate Norton and Mary J. Norton eds. New York: Oxford University Press. Kemp Smith, Norman (1960). The Philosophy of David Hume. New York: St. Martin’s Press.

Essays Part II Science

On Solidity and Rigidity‡ Some Notes on a Paper of Dagfinn Føllesdal Wilhelm K. Essler∗

Abstract In his paper of 2001, Dagfinn Føllesdal investigated, among other items, the relation between the views of Riemann and Helmholtz concerning the philosophical foundation of geometry. This paper continues that research: It is shown here that Riemann’s way of developing and justifying geometries – in contrast to Helmholtz’s procedure – is in accordance with Kant’s view concerning relative – or empirical – geometries as well as with the way in which geometries are handled since Einstein's theories on relativity.

In his very important paper “Relativity, Rotation and Rigidity”, published at 2001 in “Erkenntnis”, Dagfinn Føllesdal analyzed, among other items, the contribution of Riemann as well as that of Helmholtz to the philosophy of physics. The following essay of mine is – strictly speaking – nothing but a series of holistically interconnected footnotes to his paper. Riemann was, according to my historical knowledge, the first one to develop the intellectual means to discern what parts of geometry may be regarded as general attributes of spaces and therefore a priori1 truths of ‡

For drawing my attention to linguistic and content-related inadequacies in an earlier draft of this essay, I am deeply grateful to Michael Frauchiger. I am also very much obliged to my colleague and friend Dagfinn Føllesdal for tracking the extant shortcomings of the essay down, fixing them, as well as for turning my Continental English into more readable American English. Of course, any lingering inadequacies (due to my stubbornness) are solely my responsibility. ∗ Johann-Wolfgang-Goethe-Universität Frankfurt a. M. 1 In contrast to Kant's usage of “a priori”, I am using this expression without his associated meaning of non-empirical provability and therefore not contextindependently. In this way, in contexts of theoretical physics, this term is connected with the conditions being necessary for determining the measuring concepts; but in

124 geometry, and what parts of it are to be regarded as its empirical contents. The 3-dimensional physical space is thereby regarded as a special case of an n-dimensional magnitude, i.e. of a function depending on n independent variables. The description of such an abstract space – of a Riemann space, the predecessor of a Hilbert space – is then not derived from experience, but laid down in some a priori manner by axioms that describe what is to be regarded as the content of this abstract space,2 of this system of extended n-fold magnitudes. Riemann wrote: Für den gegenwärtigen Zweck genügt es, aus diesem allgemeinen Theile der Lehre von den ausgedehnten Grössen, wo weiter nichts vorausgesetzt wird, als was in dem Begriff derselben schon enthalten ist, zwei Punkte hervorzuheben, wovon der erste die Erzeugung des Begriffs einer mehrfach ausgedehnten Mannigfaltigkeit, der zweite die Zurückführung der Ortsbestimmungen auf Quantitätsbestimmungen betrifft und das wesentliche Kennzeichen einer nfachen Ausdehnung deutlich machen wird.3

These rules are therefore – anticipating Hilbert – regarded by Riemann as a priori statements, which establish the concept of a general extended space, i.e. the structure of every concrete as well as abstract space. They therefore do not – and must not – determine the degree of curvature at any point of that space. But then, as a consequence, being unlimited does not coincide with infinity: Bei der Ausdehnung der Raumconstructionen in's Unmessbargrosse ist Unbegrenztheit und Unendlichkeit zu scheiden; jene gehört zu den Ausdehnungsverhältnissen, diese zu den Massverhältnissen.4

Of course, this degree of curvature cannot be measured directly. In order to arrive at a decision, the instrument of simplicity of empirical theories has to be applied: contexts of experimental physics, much more is required a priori to receive a posteriori results. 2 Cf. Hilbert’s letters to Frege, see Frege [1969], interpreted in Essler [1970]. 3 Riemann [1868], 274. 4 Riemann [1868], 284.

125

... es ist also sehr wohl denkbar, dass die Massverhältnisse des Raumes im Unendlichkleinen den Voraussetzungen der Geometrie [mit durchgehendem Krümmungsmass 0] nicht gemäss sind, und dies würde man in der That annehmen müssen, sobald sich dadurch die Erscheinungen auf einfachere Weise erklären liessen.5

According to Riemann, the physicist has to start using classical mechanics; but he slightly and carefully modifies his theory with regard to those empirical results which do not conform with classical mechanics: Die Entscheidung dieser Fragen kann nur gefunden werden, indem man von der bisherigen durch die Erfahrung bewährten Auffassung der Erscheinungen, wozu Newton den Grund gelegt, ausgeht und diese durch Thatsachen, die sich aus ihr nicht erklären lassen, getrieben allmählich umarbeitet; solche Untersuchungen, welche, wie die hier geführte, von allgemeinen Begriffen ausgehen, können dazu dienen, dass diese Arbeit nicht durch die Beschränktheit der Begriffe gehindert und der Fortschritt im Erkennen des Zusammenhangs der Dinge nicht durch überlieferte Vorurtheile gehemmt wird.6

Using some semi-Kantian methodological concepts, Riemann’s point of view may be reformulated as follows: “In formulating as well as in using classical mechanics, the concepts of its geometry, which is simpler than alternative geometries, are to be taken and are to be used a priori for classical mechanics in order to formulate this theory and use its laws in applications. But it might turn out that in developing alternative empirical theories concerning the Unmessbarkleine – or even the Unmessbargrosse – we have to alter parts of that formerly a priori accepted geometrical presuppositions for reasons of simplicity. Even the concept of rigid body, where ‘rigid body’ is to be understood as ‘body whose Massgrösse does not depend on place within space’, is not a rigid concept.” The intentions of Riemann, guiding his investigations which were published posthumously in his paper “Über die Hypothesen, welche der Geometrie zu Grunde liegen” (Riemann [1868]), were directed mainly to 5 6

Riemann [1868], 285. Riemann [1868], 286.

126 items of mathematics and physics. For example, he wanted to determine the conditions when distances within such Riemann spaces are to be measured by real numbers and when they are to be measured by complex numbers. Helmholtz, on the other side, as he is indicating in his response “Über die Thatsachen, die der Geometrie zu Grunde liegen” (Helmholtz [1868]), was mainly interested in physics and in epistemology, especially in the theory of physical measurement and in an empirical theory of perception. He is describing this himself: “ich selbst, als Anhänger der empiristischen Wahrnehmung”.7 By the way, Helmholtz’s response arouses the impression of a rapid shot, written to be published – at least almost – simultaneously with that of Riemann. For it contains a serious gap, discovered and corrected by the Norwegian mathematician Lie;8 and its philosophical content is remarkably poor – not pure, but poor, compared to Kant’s theories of perceiving and measuring. We therefore have, in addition, to rely on his paper „Über den Ursprung und die Bedeutung der geometrischen Axiome“ (Helmholtz [1868/69]) as well as on „Über den Ursprung und Sinn der geometrischen Sätze – Antwort gegen Herrn Professor Land“ (Helmholtz [1878]). These two papers contain, besides serious philosophical errors, also interesting insights. Helmholtz’s intention was to establish an empirical foundation for Euclidian geometry by substituting Kant’s non-convincing a priori arguments for this geometry9 by empirical ones. He is doing this by using some statements of the background physical theory of his period, and – seen from my point of view – by using them in that context in an a priori manner, like: (1) “Within physical space, straight lines are nothing but light rays within some homogenous medium”; (2) “Within physical space, measuring is nothing but moving bodies within this space”; 7

Helmholtz [1878], 74. Cf. Føllesdal [2001]. 9 This, too, is an error of Helmholtz. For Kant nowhere argued that the axiom of parallelity was a priori true! 8

127

(3) “Within physical space, perfect measuring is the result of moving rigid bodies, i.e. bodies which do not change their inner and outer form and shape, esp. not by moving them”; (4) “Within physical space, there exist such rigid bodies, at least in the sense of the limit of increasingly solid bodies, starting at the body of oneself and ending at the standard meter in Paris and its official copies, for example in Göttingen”; (5) “Within physical space there exists throughout free movability of such rigid bodies within every part of this space, so that every distance in space is measurable”.10 In fact, Helmholtz regarded these and related conditions not only as being sufficient for establishing the existence of the physical space but also as being necessary for that purpose, although he did not conclude expressis verbis: “If these conditions were not satisfied, then there were no physical space at all, and we then would not live in some physical space”. And furthermore, he obviously insisted that all the axioms of physical geometry are – in Kant’s sense of the expression – “empirical.” To establish this he argued that the degree of curvature at different points of a physical space, which is determined by conditions like (1) to (5), can be measured only by empirical means,11 i.e. by means within physical space and thus within empirical reality. On the other hand, Kant – to whose philosophy he implicitly is referring all the time without mentioning Kant’s name until his concluding remarks – did not avoid assertions about the outer frontiers of the empirical reality, i.e. to the Ding an sich and to Transzendentale Apperzeption. Both are mere Begriffe without Anschauungen, i.e. without designing anything we can look at by using these concepts. But they are, according to him, erkenntnisleitende Ideen, i. e. knowledge-establishing ideas: If there were no Ding an sich by which the senses are affected, then everything which 10 11

Helmholtz [1868], 35-36, 60, [1868/69], 22-29, [1878], 69-70. Helmholtz [1878], 76.

128 was experienced would be nothing but dreams; and if there were no transzendentale Apperzeption – manifesting itself as the idea "Ich bin" without any recurring to the flow of time, therefore being not in time and thus unchangeable through time – then there would be no unchangeable space and no unchangeable laws of nature of objects within space. But this approach of Kant is, of course, all but an empirical kind of developing some justification for the stability of the laws of nature and the rigidity of some of its objects. Helmholtz does not take into account whether there exists, besides a rigid part of his own body in order to measure distances, also a solid part of his mind in order to receive some firm and unchangeable knowledge of the results of such performances of measurement. But Kant’s Ding an sich was that one thing which he was interested in. He called it “das Reale”, in addition to “das Objective” which, in his sense of the word, consisted in perceptions and measuring results, in short: in empirical results. He wrote: Nun finden wir als Thatsache des Bewusstseins, dass wir wahrzunehmen glauben Objecte, die sich an bestimmten Orten im Raume befinden; dass ein Object an einem bestimmten besonderen Orte erscheint und nicht an einem anderen, wird abhängen müssen von der Art der realen Bedingungen, welche die Vorstellungen hervorrufen.12 Wir müssen schliessen, dass andere reale Bedingungen hätten vorhanden sein müssen, um zu wirken, dass die Wahrnehmung eines anderen Ortes des gleichen Objects eintrete. Es müssen also in dem Realen irgend welche Verhältnisse oder Complexe von Verhältnissen bestehen, welche bestimmen, an welchem Ort im Raume uns ein Object erscheint. Ich will diese, um sie kurz zu bezeichnen, topogene Momente nennen. Von ihrer Natur wissen wir nichts; wir wissen13 nur, dass das Zustandekommen räumlich verschiedener Wahrnehmungen eine Verschiedenheit der topogenen Momente voraussetzt. Daneben muss es im Gebiete des Realen andere Ursachen geben, welche bewirken, dass wir zu verschiedenen Zeiten am gleichen Orte verschiedene stoffliche Dinge von verschiedenen Eigenschaften wahrzunehmen glauben. Ich will mir erlauben, diese mit dem Namen der hylogenen Momente zu bezeichnen.14 12

In other words: The conditions of the Dinge an sich (= das Reale) establish – via the affected sense organs – the conceptions. 13 Instead of “wir wissen” he should have had to write: “wir bilden uns ein”. 14 Helmholtz [1878], 77-78.

129 In the sense of mathematics, this attempt of Helmholtz resembles trying to get a solution for one equation containing two unknown variables, i.e. to conclude from the empirically determined objects via some unknown isomorphism to some unknown reals. His point of view is, in the sense of Kant, a kind of transcendental realism, which was already refuted by Kant.15 As I noted earlier, in Kant’s sense there exist two limits wherein empirical knowledge is available, namely: the Ding an sich at the outer side of a person, and the Transzendentale Apperzeption at the inner side of this same person. But these two limits of empirical knowledge are not themselves within this realm of empirical knowledge. Of course, they are erkenntnisleitende ideas. But nevertheless, they are mere ideas and no empirical concepts; for they are without any Anschauung, i.e. without anything to look at. What may be angeschaut, what may be looked at, are the appearances containing objects in their different temporal and partially – w.r.t. the outer objects – also spatial forms. In analysing hints to be found in Kant’s “Kritik der reinen Vernunft” [KrV], Scholz pointed out that Kant distinguished epistemic resp. transcendental time and relative resp. empirical time, the former being the base for the category of causality, whereas the latter is based on causal laws which are obtained empirically according to the principle of causality which itself is obtained by the concept of causality.16 And in Kant’s “Metaphysische Anfangsgründe der Naturwissenschaft” [MAN], Kant drew this distinction concerning space by distinguishing epistemic resp. transcendental space and relative resp. empirical space, the former being the base for the category of movement, whereas the latter is based on spatial laws of distance, by which movements may be described in accordance with suitable empirical laws of movement: In aller Erfahrung muß etwas empfunden werden, und das ist das Reale17 der sinnlichen Anschauung, folglich muss auch der Raum, in welchem wir über die 15

See the respective arguments in Kant [KrV], A 367-380. Scholz [1932]. 17 Note: Kant’s use of “das Reale” does not coincide with Helmholtz’s use of this expression! 16

130 Bewegungen Erfahrung anstellen sollen, empfindbar, d.i. durch das, was empfunden werden kann, bezeichnet sein, und dieser, als der Inbegriff aller Gegenstände der Erfahrung und selbst ein Object derselben, heißt der empirische Raum. Dieser aber, als materiell, ist selbst beweglich.18

In Kant’s view, there does not exist any rigidity or solidity within the realm of appearance: Everything there is subject to change and cannot therefore be regarded as being substantial, but is to be regarded as mere accidental, as the changing states of the one unchangeable substance, namely the totality of power: Unter den verschiedenen Arten von Einheit nach Begriffen des Verstandes gehört auch die der Kausalität einer Substanz, welche Kraft genannt wird. Die verschiedenen Erscheinungen ebenderselben Substanz zeigen beim ersten Anblicke so viel Ungleichartigkeit, daß man daher anfänglich beinahe so vielerlei Kräfte derselben annehmen muß, als Wirkungen sich hervortun ... .19

The amount of power resp. energy is the substance of the appearances; and the accidental moments are the kinds of its appearing: Grundsatz der Beharrlichkeit: Alle Erscheinungen enthalten das Beharrliche (Substanz) selbst als den Gegenstand und das Wandelbare, als dessen blosse Bestimmung, d.i. eine Art, wie der Gegenstand existiert.20 Es ist aber das Substrat alles Realen, d.i. zur Existenz der Dinge Gehörigen, die Substanz, an welcher alles was zum Dasein gehört, nur als Bestimmung kann gedacht werden. Folglich ist das Beharrliche, womit in Verhältnis alle Zeitverhältnisse der Erscheinungen allein bestimmt werden können, die Substanz in den Erscheinungen, d.i. das Reale derselben, was als Substrat alles Wechsels immer dasselbe bleibt. Da diese also im Dasein nicht wechseln kann, so kann ihr Quantum in der Natur auch weder vermehrt noch vermindert werden.21

18

Kant [MAN], 481. Kant [KrV], A 648. 20 Kant [KrV], A 182. 21 Kant [KrV], B 225. 19

131 Therefore, Kant’s view may be brought in accordance with Riemann’s theory, but not with Helmholtz’s epistemology and philosophy of science; for the latter one appears, in the view of Kant’s philosophy, as a preKantian theory. Kant did not investigate whether transcendental time and empirical time may diverge. This is, in my view, a pardonable neglect; for he never proclaimed to be a representative of some empirical theory of perception. But to be such a representative, that is what Helmholtz proclaimed to be. Also, Kant did not investigate the relational dependence of the empirical space, where the local characteristics are to be determined by empirical means according to the method of simplicity. To do this job decades before Einstein established his Special Theory of Relativity as well as his General Theory of Relativity, this would have been a fruitful job to be done by philosophically interested mathematicians, instead of stating, as Helmholtz did: ... Dagegen ist die Annahme einer Kenntnis der Axiome aus transcendentaler Anschauung: 1) eine unerwiesene Hypothese; 2) eine unnötige Hypothese ...; 3) eine für die Erklärung unserer gänzlich unbrauchbare Hypothese ... .

Kentniss

der

wirklichen

Welt

Hier ist Kant in seiner Kritik nicht kritisch genug gewesen; aber freilich handelt es sich dabei um Lehrsätze aus der Mathematik, und dieses Stück kritischer Arbeit musste durch die Mathematiker erledigt werden.22

With regard to the question whether Helmholtz was a great mathematician, I am unable to answer it; and I therefore forward it to Lie and to Hilbert as well as to famous mathematicians of our days. With regard to the question whether he was a great philosopher who is able to critizise Kant, we have to investigate whether he really understood the essential points of Kant's 22

Helmholtz [1878], 80-81. – This kind of arrogance with regard to philosophy in general and to great philosophers in particular has not died out totally, up to now at least.

132 epistemology and philosophy of science; and here I am able to answer that question, namely: Helmholtz completely misunderstood the central parts of Kant's philosophy; and in publishing this misunderstanding, Helmholtz condemned himself. And still unbroken is the statement: „Große Philosophen sind selbst in ihren Irrungen noch größer als die kleinen in ihren Einsichten, seien es nun ihre wirklichen oder gar nur ihre vermeintlichen.“23 Helmholtz failed to investigate the solidity of the inner limit of the realm of experiences, namely the Transzendentale Apperzeption, the Transzendentales Ich. From an empiristic point of view, this was done later on by Mach, who rejected this idea of an independent and stable I completely.24 Husserl, however, is also one of the great philosophers; and he, together with his disciple Fink, studied this limiting concept and investigated its significance in detail.25 But I myself feel unable to elaborate their investigations, at least in the presence of Dagfinn Føllesdal; for he is the competent person to perform this analysis. According to Kant, within the realm of experience there is nothing to be perceived outside the person which might be regarded to be rigid; and there is nothing to be perceived inside the person – i.e. within the Empirisches Ich – which might be regarded to be solid. But this is in accordance with all the empirical theories of the last decades. And, furthermore, this is in accordance with some philosophies of the prePlatonic times in West and East, like that of Heraclitus and of Buddha Shakyamuni. The latter, like Mach, rejected also those limiting concepts of experience, refuting in this way Kant's presuppositions.

23

After the publishable version of this paper had been finished, I was informed by Michael Frauchiger that something similar was formulated by F. Nietzsche: „Die Irrthümer großer Männer sind verehrungswürdig weil sie fruchtbarer sind als die Wahrheiten der kleinen“. See Nietzsche, F.: Werke. Kritische Gesamtausgabe (KGW) I/4, Berlin, New York 1999, p. 420. 24 Mach [1883] and [1905]. 25 Fink & Husserl [1932/33/34].

133 Postscriptum After this paper was presented, one of the participants argued that Kant was using an epistemic – or primary, or subjective – concept of space only. But this assertion is false: Within the KrV of course, Kant was focussing his view on this primary concept of space, which may be used as a basis for developing – in relation to it – some physical, or secondary, or objective, concept of space. This latter is therefore a dependent one, a relative one. Kant‘s view in its specific aspects is explained by him in MAN as follows:26 Wenn ich den Begriff der Materie nicht durch ein Prädicat, was ihr selbst als Object zukommt, sondern nur durch das Verhältniß zum Erkenntnißvermögen, in welchem mir die Vorstellung allererst gegeben werden kann, erklären soll, so ist Materie ein jeder Gegenstand äußerer Sinne, und dieses wäre die bloß metaphysische Erklärung derselben. Der Raum aber wäre bloß die Form aller äußeren sinnlichen Anschauung (ob eben dieselbe auch dem äußeren Object, das wir Materie nennen, an sich selbst zukomme, oder nur in der Beschaffenheit unseres Sinnes bleibe, davon ist hier gar nicht die Frage). Die Materie wäre im Gegensatz zu der Form das, was in der äußeren Anschauung ein Gegenstand der Empfindung ist, folglich das EigentlichEmpirische der sinnlichen und äußeren Anschauung, weil es gar nicht a priori gegeben werden kann. In aller Erfahrung muß etwas empfunden werden, und das ist das Reale der sinnlichen Anschauung, folglich muß auch der Raum, in welchem wir über die Bewegung Erfahrung anstellen sollen, empfindbar, d. i. durch das, was empfunden werden kann, bezeichnet sein, und dieser, als der Inbegriff aller Gegenstände der Erfahrung und selbst ein Object derselben, heißt der empirische Raum. Dieser aber, als materiell, ist selbst beweglich. Ein beweglicher Raum aber, wenn seine Bewegung soll wahrgenommen werden können, setzt wiederum einen anderen, erweiterten materiellen Raum voraus, in welchem er beweglich ist, dieser eben sowohl einen anderen und so forthin ins Unendliche. Also ist alle Bewegung, die ein Gegenstand der Erfahrung ist, bloß relativ; der Raum, in dem sie wahrgenommen wird, ist ein empirischer Raum, der selbst wiederum und vielleicht in entgegengesetzter Richtung in einem erweiterten Raume bewegt, mithin auch die in 26

See Kant [MAN], 481 f.

134 Beziehung auf den erstern bewegte Materie in Verhältniß auf den zweiten Raum ruhig genannt werden kann, und diese Abänderungen des Begriffs der Bewegungen gehen mit der Veränderung des relativen Raums so ins Unendliche fort. Einen absoluten Raum, d. i. einen solchen, der, weil er nicht materiell ist, auch kein Gegenstand der Erfahrung sein kann, als für sich gegeben annehmen, heißt etwas, das weder an sich, noch in seinen Folgen (der Bewegung im absoluten Raum) wahrgenommen werden kann, um der Möglichkeit der Erfahrung willen annehmen, die doch jederzeit ohne ihn angestellt werden muß. Der absolute Raum ist also an sich nichts und gar kein Object, sondern bedeutet nur einen jeden anderen relativen Raum, den ich mir außer dem gegebenen jederzeit denken kann, und den ich nur über jeden gegebenen ins Unendliche hinausrücke, als einen solchen, der diesen einschließt und in welchem ich den ersteren als bewegt annehmen kann. Weil ich den erweiterten, obgleich immer noch materiellen, Raum nur in Gedanken habe und mir von der Materie, die ihn bezeichnet, nichts bekannt ist, so abstrahire ich von dieser, und er wird daher wie ein reiner, nicht empirischer und absoluter Raum vorgestellt, mit dem ich jeden empirischen vergleichen und diesen in ihm als beweglich vorstellen kann, der also jederzeit als unbeweglich gilt. Ihn zum wirklichen Dinge zu machen, das heißt die logische Allgemeinheit irgend eines Raums, mit dem ich jeden empirischen als darin eingeschlossen vergleichen kann, in eine physische Allgemeinheit des wirklichen Umfangs verwechseln und die Vernunft in ihrer Idee mißverstehen. Schließlich merke ich noch an: daß, da die Beweglichkeit eines Gegenstandes im Raum a priori und ohne Belehrung durch Erfahrung nicht erkannt werden kann, sie von mir eben darum in der Kritik der r. V. auch nicht unter die reinen Verstandesbegriffe gezählt werden konnte, und daß dieser Begriff als empirisch nur in einer Naturwissenschaft als angewandter Metaphysik, welche sich mit einem durch Erfahrung gegebenem Begriffe, obwohl nach Principien a priori, beschäftigt, Platz finden könne.

Already Scholz27 noted that Kant, besides his epistemic – or primary, or subjective – concept of time regarded also a physical, or secondary, or objective, time, to be determined via applications of the category of causality.

27

See: Scholz [1932]

135 References Essler, W. K. [1970] “Ueber die Interpretation von Wissenschaftssprachen”, in: Philosophisches Jahrbuch 77 (1970), 117-130 Fink, E. and Husserl, E. [1932/33/34] VI. Cartesianische Meditation. I. Teil: Die Idee einer transzendentalen Methodenlehre: Texte aus dem Nachlass Eugen Finks (1932) mit Anmerkungen und Beilagen aus dem Nachlass Edmund Husserls (1933-34), ed. by H. Ebeling, J. Holl, and G. van Kerckhoven, HusserlianaDokumente II/1, Dordrecht, Boston, London 1988 Føllesdal, D. [2001] “Relativity, Rotation and Rigidity”, in: Erkenntnis 54 (2001), 3138 Frege, G. [1969] Nachgelassene Schriften und wissenschaftlicher Briefwechsel I, ed. by H. Hermes, F. Kambartel, and F. Kaulbach, Hamburg 1969 Helmholtz, H. v. [1868] “Über die Thatsachen, die der Geometrie zum Grunde liegen” (1868), repr. in: H. v. Helmholtz, Über Geometrie, Darmstadt 1968, 32-60 Helmholtz, H. v. [1868/69] “Über den Ursprung und die Bedeutung der geometrischen Axiome” (1868/69), repr. in: H. v. Helmholtz, Über Geometrie, Darmstadt 1968, 1-31 Helmholtz, H. v. [1878] “Über den Ursprung und Sinn der geometrischen Sätze – Antwort gegen Herrn Professor Land” (1878), repr. in: H. v. Helmholtz, Über Geometrie, Darmstadt 1968, 61–81 Kant, I. [MFP] De mundi sensibilis atque intelligibilis forma et principiis (1770), repr. in: W. Weischedel (ed.) Immanuel Kant: Werke in 10 Bänden V, Darmstadt 1968, 8-107 (Latin-German) Kant, I. [KrV] Kritik der reinen Vernunft A (= 1781), B (= 1787), repr. Leipzig 1924 Kant, I. [MAN] Metaphysische Anfangsgründe der Naturwissenschaft (1786), repr. in: Preuß. Akad. Wiss. (ed.) Kants gesammelte Schriften IV, Berlin 1903/11, 465565 Mach, E. [1883] Die Mechanik in ihrer Entwicklung historisch-kritisch dargestellt, Leipzig 1883.

136 Mach, E. [1905] Erkenntnis und Irrtum: Skizzen zur Psychologie der Forschung, Leipzig 1905. Riemann, B. [1868] “Über die Hypothesen, welche der Geometrie zu Grunde liegen” (1854, publ. posthum 1868), repr. in: H. Weber (ed.) Bernhard Riemann's gesammelte mathematische Werke und wissenschaftlicher Nachlass, 2nd ed. 1892, repr. Vaduz n.d., 272-287 Scholz, H. [1932] Kant (Vorlesungsnachschrift), Münster i. W. 1932 Stein, H. [1977] “Some Philosophical Prehistory of General Relativity”, in: J. S. Earman, C. N. Glymour, and J. J. Stachel (eds.) Foundations of Space-Time Theory, Minneapolis 1977, 3-49 Størmer, C. [1902] Verzeichnis über den wissenschaftlichen Nachlass von Sophus Lie, erste Mitteilung, Norweg. Akad. Wiss., Oslo 1902

Some Remarks on the Distinction between Basic (Theoretical) and Applied (Practical) Science and Its Importance in the Politics of Science‡ Nils Roll-Hansen ∗

Abstract The paper argues that the traditional conceptual distinction between basic and applied science is important for an appropriate institutional differentiation of scientific research. The difference in social function is reflected, e.g., in the criteria of success. In basic science (often called “pure science”) the primary criterion of successful research is contribution to the common fund of knowledge, while in applied research (often discussed as “technology”) it is the solution of specific practical problems. In other words applied research is subordinate to external (social) goals while basic research is primarily answerable to internal (epistemic) goals of science. The paper discusses Philip Kitcher’s argument for “constraints on free inquiry” based on historical claims that certain kinds of basic research have in fact had bad consequences and that further research in the same area is most likely to also have bad consequences, and is thus to be morally condemned. Kitcher’s main case, his historical interpretation of research on human heredity, is questioned. And contrary to Kitcher’s claim, that there is no politically important difference between basic and applied research, ‡

This paper has not been reworked since it was originally submitted in December 2006. The only change is the addition of an explanatory footnote on page 144. My interest in the topic was developed during many years working at the Nordic Institute for Studies in Innovation, Research and Education. A broader theoretical development of my view was published as a preprint in 2009: "Why the distinction between basic (theoretical) and applied (practical) research is important in the politics of science." Centre for Philosophy of Natural and Social Science. Technical report No. 04/09. London School of Economics. An underlying idea is that the adequacy of philosophical theories about the nature of science and its social roles can be tested on how well they are able to support tenable historical accounts of political-scientific controversies like those over spontaneous generation, eugenics, Lysenkoism, or acid rain. ∗ Universitetet i Oslo

138 the paper argues that the distinction is important for upholding the difference between science and politics, representing facts and values respectively, that a liberal democratic political system depends on.

In recent years there have been frequent demands for a new social contract for science, including a clarification of the moral and social responsibilities of scientists. The mutual dependence of science and democracy is also reemerging as a central topic in the philosophy of science, after having been widely neglected or derided by scholars in science studies since the 1960s. I will argue that a lack of appropriate distinctions is a serious obstacle to fruitful analysis and discussion in this field. “Science” is regularly taken in a wide sense of “science & technology” covering a diverse set of activities with very different social functions and effects. The heavy reliance on such a broad category reduces both insight and usefulness of science studies. It blunts theoretical understanding and undermines political and administrative decision-making. For instance the traditional distinction between “basic” and “applied” science is widely considered to be lacking in substance and justification. Some critics have even argued that it is harmful by preventing democratic control of science. The negative attitude to the distinction between basic and applied science can be curiously ambiguous, however. Quite frequently the same person or document that denies or doubts the importance of the distinction will go on to say that basic science needs more generous support. It appears that even for sceptics the traditional distinction catches something important. I will argue that its core reflects the political significance of the difference between descriptive and normative claims, between facts and values. While the purpose of (basic) science is to describe the world as it is, the ambition of politics is to make it what it ought to be. Applied science appears as an area of overlap between the two. The supreme criterion of good applied research is that it contributes to the solution of specific practical problems, and research projects are evaluated according to such expectations. In applied science there will always be a selective pressure for wishful thinking, i.e., there will be a tendency that thinking which adapts to

139 the way its patrons would like the world to be is preferred at the expense of more objective approaches. An important political function of basic science is to provide a platform from which such adaptation can be criticized with relative impunity. Basic science in this way guards the interests of society as a whole. This can be seen as the primary rationale for its relative institutional autonomy. Values in science It is a commonplace today that science is not value-free and merely descriptive. It clearly depends fundamentally on “internal” values, like the traditional epistemic values listed for instance by Thomas Kuhn: “accuracy, simplicity, fruitfulness and the like” (Kuhn 1970: 199). The problem is the role of “external” social and political values like social justice and equal opportunities. That such values are essential for deciding on practical applications and research projects in applied research is hardly in dispute. Controversy only begins when such values are invoked to argue that certain theoretical questions should not be pursued or that certain factual claims are illegitimate or untrue. In this paper I will take a closer look at two proposals for a radical epistemic role for external values: Philip Kitcher’s argument for “Constraints on Free Inquiry” (Kitcher 2001, 2002a, 2002b), and Helen Longino’s claim that scientific objectivity depends on the participation of conflicting external interests (Longino 2002, 2002a, 2002b). Philip Kitcher is well known as a champion of scientific rationalism. In the book Science, Truth, and Democracy (2001) he attempts to steer a middle course in the aftermath of the 1990s “science wars” between the “scientific faithful” and the “detractors” of science (pp. 4, 9). On the one hand are those who “proclaim the search for objective knowledge to be one of the crowning achievements of our species” on the other those who “deny the objectivity of the sciences, question our ability to attain truth and knowledge, and conclude that the sciences are instruments of oppression” (p. xi). The question is whether he ends up, not showing convincingly that both are “false images” of science (p.199), but rather by combining them in a way that undermines scientific rationality.

140 The myth of the ivory tower Social constructivist and relativist critics of traditional scientific rationalism have regularly presented their target in the garb of a straw man. They depict basic science as an isolated activity, absorbed in its own affairs, and without concern for its effects on the rest of society. This caricature of basic science, often called “pure science”, has been an important element in criticism of the distinction between basic and applied science. Some defenders of scientific rationalism have also gone along with this characterization as a reason for rejecting the distinction. It has been argued that if basic science is to be distinguished by epistemic significance, from technology or applied science, which are valued for their practical significance, then basic science must be conceived as completely cut off from human worldly concerns. For instance Philip Kitcher claims that such a distinction presupposes that “the notion of epistemic significance has nothing to do with us and our ephemeral practical concerns, and everything to do with the structure of the world” (Kitcher 2001:66). In this perspective the concept of basic, or “pure”, science becomes untenable because it seems obvious that most important theoretical science has some connection to practical affairs (Kitcher 2001:89). But is this “myth of purity” (p.85 ff.) representative of the rationalist tradition? Among those who argued most strongly, in the middle decades of the th 20 century, that science in the theoretical basic sense must be protected against too much control and steering from political and commercial interests, a much more sophisticated conception of the interaction and mutual dependence of basic and applied science was prevalent. Neither Karl Popper, Michael Polanyi, Robert Merton, nor for that matter Vannevar Bush, held that basic science should have no commerce with practical affairs. The legitimate concern underlying the caricature is that some scientists have attempted to avoid responsibility for the social effects of their work. Some decades ago Tom Lehrer sang: “Once the rockets are up, who cares where they come down? That is not my department says Werner von Braun.” (Kitcher 2001: 89.) But this attitude is hardly typical or dominant among modern scientists. For instance, the reaction of many physicists after the atomic bomb was quite the contrary.

141 Constraints on scientific research It is tempting to try to protect society from dangerous scientific and technological discoveries and inventions by curtailing the research that produce them. The world would be a safer place without nuclear weapons. So why not control the development of molecular biology to prevent similar evils? Since it is impossible to tell beforehand in which special field of research such possibilities will emerge, and since the different specialties of molecular biology are so interwoven, it is only by stopping research on a broad front that such control could work. This applies in particular to basic research. Once the chain reaction in uranium had been discovered nuclear weapons were an obvious possibility. In applied research aimed at specific practical purposes control and restriction makes much better sense. Certain kinds of innovations can be stopped or at least substantially postponed. For instance, the practical development of effective nuclear weapons demanded extensive technological research and great material resources. Even today when the general technology of nuclear weapons is well known it takes time and great resources for new countries to do the “research” necessary to produce them. But restrictions in more specific areas of applied research is also likely to have costs by preventing development of beneficial technology. Research on human embryos is a current case. Since the 1970s sociobiology has been pointed out as an area where research should be restricted. In this case it is ideological rather than technological effects that are feared. For instance, if certain groups of people are found to be genetically disposed for lower intelligence than others this can lead to social discrimination. Researchers in human behavioural genetics have defended their activity by appealing to a right of free scientific inquiry based on the principles of free expression which is generally recognized as essential to a liberal democratic society. Philip Kitcher has countered this defence and argued for “constraints on free inquiry”. He points to John Stuart Mill’s caveat to the justification of freedom of expression in On Liberty: Freedom is a good thing only “so long as we do not attempt to deprive others of theirs or impede their efforts to obtain it”. Through a complex argument Kitcher makes the claim that under present circumstances research on heritability of human intelligence is likely to stimulate discrimination of blacks, women etc. It should there-

142 fore not be pursued. He does not recommend prohibition but moral pressure, relying on the moral attitude of individual scientists and others to restrain such research (Kitcher 2001: 93-108). Kitcher’s platform is an overall belief in scientific rationality, including the right to free inquiry, and he argues only for local constraints. Indeed he presupposes a second order scientific argument to identify the areas where constraint is appropriate. But how can we reliably decide on the appropriate constraints for the local context in question? And in particular, how do we secure the ability to properly modify and correct such decisions as the situation changes or new knowledge is introduced? Recent and present difficulties with issues of environment and health, for instance acid rain, whale hunting, global warming, HIV-AIDS or mad cow disease, illustrate well how hard it can be to achieve an adequate balance between political decisiveness and flexibility in the face of uncertain and changing knowledge. Arguing from historical experience Kitcher’s approach depends on reliable knowledge about the historical development of a specific field of scientific research and its social effects. He accepts the widespread view in recent history and sociology of science that modern biological science has supported discrimination and social repression. This, according to Kitcher, is true in particular for genetics since its formation around the turn of the 19th century. He claims that “a wealth of historical studies hammers home the same moral” (Kitcher 2001: 209). In a thought experiment Kitcher (2001: 96-100) describes a socially underprivileged group, whose discrimination is shored up by public beliefs in its genetic inferiority. Due to a general “epistemic bias” among the public scientific results that indicate substantial genetic differences will reinforce this belief, while results that indicate small or zero differences are likely to have no impact. A problem with the presentation of this thought experiment is that it assumes a God’s eye point of view. The concluding claim of public bias assumes “true” scientific understanding of genetic facts as well as the social relations of science. Kitcher draws his understanding of the mechanism that mediates between science and politics from historical studies.

143 But how adequate is Kitcher’s selection of historical evidence and interpretation? I would argue that the selection is one-sided and that on a broader and more balanced use of history a more positive assessment of the role of biological science is likely to be true. Genetic investigations have gradually revealed that with respect to contested properties like IQ genetic differences between groups of humans is small or non-existent. This general development of knowledge in behavioural genetics must surely be a very important support for anti-discriminatory views in present social politics. Kitcher does not take this wider historical and social context properly into account. An interesting example of biased historiography is the strong tendency to identify the programs for sterilization in early and mid-20th century with eugenics. In Scandinavia, for instance, it was more motivated by social concerns over women’s emancipation and improved living conditions for children than by eugenic visions of a genetically better race (Broberg and Roll-Hansen 2005). A fundamental problem is the role of scientific experts, philosophers and historians as much as biologists and medics, in deciding on measures of constraint. We know from history how fragile and penetrable to interference are the institutions that guard the autonomy of scientific research and judgement. At present there are worries that we have entered a period with more pressure on scientific objectivity than we have had since the mid-20th century. To secure a sound balance and interaction between science and politics, between scientific research and the aim of general human wellbeing, Kitcher proposes a system for the governing of science which he calls “Well-Ordered Science”. This ideal implies democratic control of science policy. Decisions are to be taken by democratically elected bodies under the tutorship of scientific experts. A key problem, as Kitcher is well aware, is how the experts can get legitimate authority and influence. How can the interaction between science and the rest of society be organized to make the production of scientific knowledge both reliable, relevant and respected. On the one hand scientists are often, and sometimes rightly, accused of condescending, narrow-minded, and self-serving dogmatism. On the other hand populism and common prejudice often disregards scientific knowledge with disastrous consequences. Being a scientific expert one may have good reasons to become impatient with stubborn ignorance. But

144 it is also very easy to overstep the limits of well supported knowledge claims. As I indicated above Kitcher’s judgement about the repressive historical role of genetics is not convincing. Longino’s pluralism Some critics have found Kitcher’s model for democratic governance of science and technology to assume a too dogmatic belief in the objectivity of scientists’ knowledge and judgement (Longino 2002a, 2002b). On my own part I find it uncomfortable when he states: “I believe that a sober review of the history of research into racial and sexual differences supports the view recorded in the argument, and thus any attempts to read that history differently embody just that epistemic bias that the argument diagnoses” (Kitcher 2001: 106).1 As I have pointed out the historical basis for such a generalization is weak, and the argument appears lacking in critical self-reflection. Helen Longino and Philip Kitcher have exchanged views on how “social and cognitive aspects of the practice of the sciences might be integrated” (Kitcher 2002a: 549), or more specifically, how is epistemic pluralism motivated by liberal democracy to be meshed with scientific rationalism. It turns out they disagree on how radical pluralism can be and still respect science. In Longino’s view the direct involvement of diverse external non-epistemic values in scientific research guarantees maximum objectivity. Only by ensuring that a broad variety of legitimate social interests is represented through active researchers will science realize its full potential for true knowledge. Kitcher finds her view, especially as it is expressed in 1

The ”argument” referred to is Kitcher’s own argument that historical studies show that research into genetic differences between races has in fact caused social discrimination and it is therefore morally wrong to pursue such research. A crucial part of Kitcher’s argument is that due to ”political asymmetry”, leading to “epistemic bias”, empirical results contrary to existing beliefs in substantial differences will make little or no difference to these beliefs, while results confirming such differences are bound to strengthen the beliefs. He apparently lets concern for popular beliefs overrule concerns about truth. Kitcher explicitly rejects a ban or prohibition of the disputed kind of genetic research, but appeals to the moral sensibility of each scientist. The argument is explained in a chapter called “Constraints on free inquiry.” Kitcher 2001: 93-108. (Footnote added May 2012.)

145 her repeated criticism of a “pervasive Rational-Social dichotomy” in most of present science studies, to be “obsessive” and unclear (Kitcher 2002a: 549). In response Kitcher sets a limit to epistemic pluralism: “the true statements, the accurate maps” etc. should be “jointly consistent”. The history of science provides many examples of equally legitimate but inconsistent accounts of the same phenomena, he admits, but science typically works to achieve “a consistent account of the area of overlap”. As an example he mentions the Bohr atom which contradicted classical electromagnetic theory (Kitcher 2001: 570), and led to the development of quantum mechanics. However, Bohr himself suggested that for fundamental phenomena relating to matter, life and mind we may have to accept mutually inconsistent but complementary accounts as the best that human science can achieve. The idea of “incommensurability” as developed through historical case studies by Kuhn and Feyerabend similarly suggests a more radical pluralism than Kitcher’s. There are in my opinion good reasons “for wanting to go beyond” his “modest realism and ‘conservative’ epistemology” (Kitcher 2001: 571). A new era for the philosophy of science? After the confrontations of the science wars in the 1990s analytical philosophy of science is now engaged in a radical re-examination of the moral foundations of scientific epistemology, analyzing the complex relationships between different kinds of facts and values in research and other scientific activities. Feminist philosophy of science has been a main force insisting that the basic epistemic issues are unavoidably political and cannot be sincerely and truly answered except from an explicit political standpoint. Inspiration for a “New Program for Philosophy of Science” has been sought in radical socialist empiricism of the Vienna Circle (Cartwright et al. 1996, Kourany 2003). Critics have argued that this political project is incompatible with “the professional pursuit of the philosophy of science” (Giere 2003). I have used Philip Kitcher and Helen Longino as two representative examples. Both want to preserve scientific knowledge and rationality as an essential element in World civilization, though it is presently in acute need of revision. They agree that the ultimate criterion of a rationally acceptable

146 science is that it serves human welfare, and that science must be brought under democratic political control in order to minimize its bad social effects and maximize the good. But there is disagreement on the means and ways of democratization, related to different conceptions of scientific method. The feminist Longino embraces radical pluralism. A broad representation of external social interests among researchers is the best means to further objective conclusions. The rationalist Kitcher fears that lack of empirical accuracy and logical stringency can undermine the reliability of what passes as scientific knowledge. A vague pluralism will make adequate gate-keeping to the scientific community impossible and obscure the difference between science and pseudoscience in the public understanding. Longino finds Kitcher is blocking progress with a conservative epistemology. Kitcher sees Longino as a well-meaning but somewhat confused and irresponsible romantic. Kitcher’s choice of a line of defence may be too much inspired by logical empiricist belief in the virtues of a precise and meaning invariant scientific language. Kuhn and especially Feyerabend have demonstrated the historical existence of an incommensurablity of theories pointing toward a more radical pluralism than Kitcher is willing to accept. Feyerabend’s explanation of incommensurability as a consequence of realism (Feyerabend 1981: 201-2) indicates how a more robust and substantial scientific realism than Kitcher’s “modest realism” may compensate for the loss of formal methods. The common interpretation of Feyerabend as an anti-rationalist is superficial, it does not properly appreciate his fundamental loyalty to Mill’s ideas about “Liberty of Thought and Discussion” (Lloyd 2000). I do share Kitcher’s worry that increasing confusion concerning standards for scientific knowledge, inside the scientific community as well as outside, will reduce the possibility of a democratization that does not seriously undermine the ability of science to produce reliable knowledge. A key problem for the politics of science is how to build institutions capable of accepting the radical epistemic pluralism and still hold on to the elusive but essential standards for judging claims to knowledge.

147 A sound and useful distinction: basic vs. applied science A main weakness of present discussions about the moral basis of scientific epistemology is a lack of critical comparative empirical studies. Historical claims play an essential role in the arguments. But they are often based on problematic selection and interpretation of the historical material. For instance Kitcher’s argument for constraints on scientific inquiry is built on problematic claims about the historical effects of human genetics. A main reason for unsatisfactory handling of the historical evidence appears to be disregard of the differences between basic and applied science. Such a distinction is needed to clearly perceive what the morally relevant differences were, for instance, between pursuit of the chromosome theory and eugenically motivated investigations of heritability of mental retardation and illness. The former was relatively detached from specific social policy programs. It provided a platform for more objective judgement of such programs, and over time it produced general knowledge that indicated equality between sexes and races. The latter was closely linked to the promotion of specific policies in practical social work. It depended for funding on service to certain policies, including their implicit ideologies and values, and tended to support existing prejudices, at least in the short run. Eugenics to the extent that it can be characterized as research rather than policy promotion, was applied and not basic research (Roll-Hansen 2001). Longino seems to agree with Kitcher in rejecting the distinction between basic and applied science as politically important. She agrees with Kitcher that there is no “’morally significant’ distinction between science and technology” (Longino 2002a: 561). But is there really no significant moral difference between the Hahn-Meitner discovery of the chain reaction in uranium 236 and the Manhattan project of building the first atomic bomb? Or between the Watson-Crick discovery of the double helix, and the development of new drugs, or biological weapons using molecular genetic technology? Kitcher is uncomfortably dogmatic in his rejection of autonomy for (basic) science. Recently the journal Philosophy Today published a special issue promoting the development of a “philosophy of science policy” (Mitcham 2004). The content indicated growing interest in this field of study, but also limited interaction between theoretical analysis and empiri-

148 cal description. One of the contributors is Philip Kitcher who concluded that: “The idea of the autonomy of the sciences is an unfortunate hang-up from our past.” (Kitcher 2004: 57). He takes autonomy to imply an absolutely unconstrained search for truth. But the interesting and historically realistic idea is relative autonomy, through institutions that safeguard a higher degree of autonomy for some scientific activities than for others. There is an element of dogmatic dichotomizing in Kitcher’s argument: Either science has to be completely autonomous, or there is no important difference between basic and applied. Contrary to Kitcher’s conclusion it appears plausible that existing institutional differentiations between scientific academies, universities, research councils of different kinds, special institutes serving industry or political management, etc., express important differences in moral responsibility as well as political roles. A relative autonomy for basic science as compared to applied science may be an important condition for sound democratic governance of science. It protects basic science from too much interference of short term political and economic goals and ideologies, and thus makes it possible for basic science to act as a platform for effective criticism and independent advice into current politics. The political insistence on no important difference between basic and applied research, the “unity of theory and practice” was an important factor in precipitating the Lysenko scandal in Soviet science (Roll-Hansen 2005). Through this conflation the government lost the extra grip on reality that modern science has provided.

References Broberg, Gunnar, and Nils Roll-Hansen (eds.) (2005) Eugenics and the Welfare State. Sterilization Policy in Denmark, Sweden, Norway, and Finland. 2nd edition. East Lansing: Michigan State University Press. Feyerabend, Paul (1981) Realism, rationalism and scientific method. Philosophical papers. Vol. 1. Cambridge etc.: Cambridge University Press. Giere, Ronald (2003) “A new Program for Philosophy of Science?” Philosophy of Science, 70: 15-21.

149 Kitcher, Philip (2001) Science, Truth, and Democracy. Oxford: Oxford University Press. Kitcher, Philip (2002a) “The Third Way: Reflections on Helen Longino’s The Fate of Knowledge”. Philosophy of Science, 69: 549-559. Kitcher, Philip (2002b) “Reply to Helen Longino”. Philosophy of Science, 69: 569572. Kitcher, Philip (2004) “On the autonomy of the sciences”, Philosophy Today, Supplement 2004, pp. 51-57. Kouranyi, Janet A. (2003) “A Philosophy of Science for the Twenty-First Century”. Philosophy of Science, 70: 1-14. Kuhn, Thomas S. (1970) The Structure of Scientific Revolutions. 2nd ed., with “Postscript”. Chicago: University of Chicago Press. Lloyd, Elisabeth (1997) “Feyerabend, Mill, and Pluralism”, Philosophy of Science, Supplement to vol. 64, S396-408. Republished in John Preston et al. (eds) The Worst Enemy of Science? Essays in Memory of Paul Feyerabend. New York, Oxford: Oxford University Press, 2000. Longino, Helen E. (2002) The Fate of Knowledge. Princeton: Princeton University Press. Longino, Helen E. (2002a) “Science and the Common Good: Thoughts on Philip Kitcher’s Science, Truth, and Democracy”. Philosophy of Science, 69: 560-568. Longino, Helen E. (2002b) “Reply to Philip Kitcher”. Philosophy of Science, 69: 573577. Mitcham, Carl (2004) “New directions in the philosophy of Science. Toward a Philosophy of Science Policy”. Philosophy Today, supplement 2004, pp.3-15. Roll-Hansen, Nils (2001) “Eugenic Practice and Genetic Science in Scandinavia and Germany: Some Comments on Peter Weingart’s Comparison of Sweden and Germany”, Scandinavian Journal of History 26: 75-86.

150 Roll-Hansen, Nils (2005) The Lysenko Effect. The Politics of Science. Amherst NY: Humanity Books.

Essays Part III Logic and Rationality

Some Consequences of the Entanglement of Logic and Mathematics‡ Charles Parsons*

Introduction That logic and mathematics are intimately connected with one another is obvious and attested by the very existence of mathematical logic. Furthermore, claims of inclusion between the two are familiar philosophical theses: Logicism holds roughly that mathematics is included in logic, while L. E. J. Brouwer seems to have held that logic is included in mathematics. I don't plan to discuss either of these views in the present paper. Furthermore, by 'entanglement' I mean something more specific than these remarks suggest. The term suggests a relation that is in some way problematic. George Washington enjoined the early American republic against foreign entanglements. The American constitutional tradition seeks to avoid entanglement of church and state. That certain states of particles in quantum mechanics are entangled is a source of puzzlement. I will argue that in the ‡

This paper was presented, in unavoidably condensed form, at the Second International Lauener Symposium on Analytical Philosophy at the University of Bern, on the occasion of the presentation of the Lauener Prize to Dagfinn Føllesdal. It was a special pleasure to participate in an event honoring Dagfinn, a friend since our graduate student days from whom I have learned much over many years. This paper does not have a direct connection with his work, but the reader might detect the influence of our common mentor W. V. Quine. The paper results from seminars at Harvard University in the fall of 2004 and UCLA in the spring of 2005. I am indebted to the participants in these seminars. The material was also used for talks at the Hebrew University of Jerusalem, the University of California, San Diego, Stanford University, and the University of Pittsburgh. I owe special debts to D. A. Martin, Yiannis Moschovakis, Carl Posy, and Agustín Rayo. Section V was prompted by the discussion at Bern, in particular by Føllesdal's remarks. Some revisions (mostly additions) have been made in 2008. Thanks to Peter Koellner for comments at that point. * Harvard University

154 most basic cases the phenomenon I am concerned with is not problematic, but in other cases, particularly that of second-order logic, it probably is. But I mainly want to point to it as a pervasive feature of logic that is neglected in a lot of philosophical discussion and argue that the persuasiveness of some philosophical theses is influenced by this neglect. I To explain what I have in mind in the case of first-order logic, I will take up a few points in John Etchemendy's The Concept of Logical Consequence.1 This book has generated an extensive critical literature, and I do not want to add to that beyond some specific points that illustrate my theme. His work also sparked controversy about the interpretation of Tarski's work on logical consequence as well as his classic monograph on truth. I will not be entering into that at all. As a preliminary I will make some very elementary logical observations. First-order logic is looked at both syntactically and semantically, in the sense that there are proof procedures on the one hand and wellestablished model-theoretic definitions of validity (logical truth), satisfiability, and logical consequence. Validity is usually defined as truth under any interpretation in a non-empty domain, which provides a range for the quantifiers and interpretations for the predicates and names.2 This is most commonly understood in terms of the set-theoretic notion of model, where the domain is a set, the one-place predicates are interpreted by assigning subsets of the domain as extensions, and so on. This understanding is required for some of what we say but not all. If we look at proofs in a proof procedure for first-order logic, nothing needs to be asserted categorically that involves more than the minimal ontology embodied in the non-emptiness of the domain, and at the cost of slight complication we could avoid even that if the language does not contain names. So either way the logic seems to have only trivial ontological commitment. Some of the usefulness of logic depends on this: although few take seriously the idea that there are only finitely many objects abso1

Etchemendy (1990). For simplicity I assume the absence of function symbols, since nothing essential would be changed by taking them into account.

2

155 lutely, we want to be able to reason about finite domains.3 Even if it's not a genuine possibility, we can even reason about a totally finite world by a kind of pretense.4 The ontological minimalism of first-order logic is compromised when we think not in terms of proof and validity but in terms of their absence: nonvalidity or unprovability. As is well known, to witness nonvalidity (or satisfiability) models with domains of cardinality up to ℵ0 can be needed. And of course a statement of unprovability is a generalization with a domain that is potentially infinite. This is probably the most elementary expression of the entanglement of first-order logic with mathematics. From the fact that the definitions of validity, satisfiability, and consequence are typically stated in model-theoretic terms relying on settheoretic concepts, it might seem that already here there is entanglement with set theory. However, a more explicit version of the completeness theorem proved in Hilbert and Bernays (1939) (§4.2 in 2d ed.) shows that this is not so at least so long as we confine our attention to single formulae or finite lists of them. Let A be a quantificational schema, that is a first-order formula on whose predicates (other than '=') no interpretation has been placed. Then we can prove in first-order arithmetic PA a formula whose antecedent says (in terms of Gödelization) that A is not logically refutable (by some standard proof procedure) and whose consequent is an arithmetical instance of A, that is, results from A by substituting for its nonlogical predicates predicates in the language of first-order arithmetic. The predicates instantiated may be chosen to be provably ∆02 . In many cases, of course, we do not need the full resources this theorem allows and can choose the domain to be finite. Some find more satisfactory the version of the completeness theorem of which the Hilbert-Bernays version is roughly the contrapositive, that any valid formula is derivable. In work of the re3

In fact finite model theory is a branch of logic that has had a lot of development in recent years. But its interest is as a study of arbitrary finite models of some language or theory, and thus finite model theory itself is not compatible with there being absolutely only finitely many objects. 4 That it is a pretense is indicated by the fact that probably the number of symbols in the formulae involved will be larger than the number of objects in the domain.

156 verse mathematics school it is proved that this completeness theorem is provable in a very weak subsystem of second-order arithmetic, WKL0.5 This result yields the information of the Hilbert-Bernays version. Strong completeness, that is satisfiability of any consistent set of formulae, is also provable in WKL0. That means that there is a model in rather simple predicative classes of natural numbers. I think the general thesis of entanglement is widely held if not emphasized as such. I think it might well be credited to Paul Bernays, who points to a number of aspects of it in the early sections of his classic paper (1930) and states it more briefly in the later (1969).6 Let us now turn to Etchemendy. The discussion that interests me is the argument of chapters 7 and 8 of his book, where he takes aim at conceptions of logical truth according to which a sentence is logically true if a certain general statement related to it is true. The definition of logical truth directly modeled on Tarski's original 1936 definition of logical consequence would be of this form. He states a principle that such an account seems to be based on, what he calls the Reduction Principle: (iii) If a universally quantified sentence is true, then all of its instances are logically true.7

He writes that "it is important to see that Bolzano and Tarski both base their accounts on this rather unlikely principle, in some form or other" (p. 100). The qualification is important because (iii) is pretty obviously false, and Etchemendy does not attribute to Tarski a commitment to (iii) as it stands. But he has a criticism that seems independent of the details of the qualification to be made:

5

Simpson (1999), theorem IV.3.3. This theorem includes the statement that, over a weaker subtheory, the completeness theorem is equivalent to the axiom Weak König's Lemma. 6 Bernays (1930), I, §§2-3, and (1969), passim. 7 Etchemendy (1990), p. 99. In the remainder of this section, this work is cited by page number in the text.

157 When we equate the logical truth of a sentence with the ordinary truth of a universal generalization of which it is an instance, we risk an account whose output is influenced by facts of an entirely "extralogical" sort (p. 107).

He evidently thinks that this criticism applies to the version of the principle that he takes most seriously: (iii'') If a universally quantified expression is true, and the constant expressions appearing in its matrix are of a distinctively logical sort, then all of its instances are logically true (p. 110).

What is of interest to us is less (iii'') itself than the particular considerations that Etchemendy brings to bear against it. He considers first the sentences σn that, for each n ≠ 0, say that the domain has at least n elements. Whether by Etchemendy's lights or according to standard criteria, neither σn nor ¬σn is a logical truth for any n ≠ 0. Thinking of each as an "instance" of a sentence with a vacuous string of universal quantifiers, (iii'') declares whatever of these sentences turns out to be true as logically true. If we think of the universal quantifiers in (iii'') as absolute, that is as ranging over absolutely everything there is, then presumably all σn are true and thus logically true according to (iii''), contrary to the above. That is likely to be the result even without the contested absolute reading. What troubles Etchemendy at this point is not so much the verdict as its dependence on what there is: Clearly, some of these sentences are true. Exactly how many depends, of course, on the size of the universe — that is, on how many objects there happen, in fact, to be. If the universe is infinite, then all of the sentences are true, and so will be mistakenly judged logically true. If it is finite, then only a finite number of them will be judged logically true. But the important point is not how many of these sentences the definition gets wrong, but rather the fact that the assessments here are clearly dependent on a nonlogical state of affairs. This is exactly the defect that seemed so apparent when we first considered the unmodified principle (iii). As we noted then, whether a sentence is logically true should not depend on substantive, extralogical facts, whether historical or physical or mathematical (pp. 111-12, emphasis added).

158 At this point the obvious idea that logical truth involves generalization over the domains of the quantifiers has not been introduced. The usual reason why, say, σ2 is not a logical truth is that it is not true in all nonempty domains, since the domain might have just one element. Etchemendy does take this into account, and considers generalizations of the sentences ¬σn that are meant to express the falsity of σn with the domain taken as any subcollection of the universe. I assume he says "subcollection" because he doesn't mean to assume that the universe is a set. The idea is then that the logical truth of ¬σn would amount to the truth of the generalization. But if the universe is finite, say with m elements, then the result is that for n > m this idea implies that ¬σn is logically true. Etchemendy says Sentence (5) [the generalization of ¬σn] makes a perfectly ordinary claim about the world, one that has little, if anything, to do with logic (p. 114).

This is a rather strange statement, since (5) is false if the "universe" contains the natural numbers, or linguistic or formal expressions as ordinarily conceived. Etchemendy evidently resists the conclusion that something's not being logically true can depend on existential claims in mathematics. Curiously, he does not emphasize the fact that mathematics (even rather modest mathematics, as noted above) is enough to construct countermodels to the sentences σn and ¬σn and even to others that fail only in infinite models. He is unmoved by the thought that the mathematics involved is necessary. As he sees it, the problem is that assumptions like the settheoretic axiom of infinity are not logical truths. Something is fishy here. Etchemendy seems to be demanding that if a sentence is not logically true, this has to be by virtue of statements that are logical truths, so that even mathematics that goes beyond logical truth is out of bounds. I don't see how this demand can be satisfied, whatever one's criterion of logical truth. Given a sentence A, the statements 'A is logically true' and 'A is not logically true' are neither of them logical truths on the usual criteria, for a rather trivial reason: they depend on the existence of the sentence A and of the elements making up its structure, as well as on its truth-conditions. In some broad sense these are reasonably called matters of logic. But the concept of logical truth is narrower even by

159 Etchemendy's own lights. He endorses what we called above the ontological minimalism of logical truth. I don't see how one could show a sentence not logically true without appeal to extra-logical facts. Even such a statement as 'A is not a truth-functional tautology' is not a logical truth by the lights of most of us. Etchemendy does not say that he thinks otherwise. The point of the exercise I have engaged in is not to defend (iii'') or another principle of the same kind but to exhibit the resistance to the entanglement of logic even with arithmetic or elementary syntax that Etchemendy's argument exhibits. A similar resistance, I believe, underlies his tendency to read the quantifiers in (iii'') and similar principles in an "actualist" way, in spite of the fact that in the usage of proponents of criteria along these lines (for him primarily Tarski) understand them to include a rich domain of mathematical objects. Minimizing the mathematical commitment of a metatheory of logic makes sense. Eliminating it altogether does not. Although I don't think Etchemendy really disagrees with that statement, I don't see how the argument we have been discussing is consistent with it. There is much more to be said about the entanglement of first-order logic with mathematics. An important question is whether there is an "intuitive" conception of logical truth that differs in an important way from the most usual model-theoretic one even if they are extensionally equivalent. Since Kreisel raised this issue in the 1960s, a number of writers, including Etchemendy, have weighed in on it. But instead of pursuing it I propose to turn to second-order logic. II The entanglement of second-order logic with mathematics is more farreaching and problematic. I will again appeal to Etchemendy to illustrate the point, but it will be clear that the entanglement goes beyond what is unavoidable in a serious conception of logic. I will consider separately the standard semantics for second-order logic and axiomatizations, since for that semantics there is no completeness theorem. Since Etchemendy is concerned with semantic definitions of logical consequence and validity, his focus is almost entirely on the former.

160 A model structure for a second-order language is the same as for a first-order language: one has a non-empty set D as domain, subsets of Dn as extensions of n-place predicates, elements of D as referents of names, and appropriate functions as referents of functors if they are present. What makes the semantics standard is that in defining satisfaction, one treats the second-order variables (with n arguments) as ranging over all subsets of D n. Although I will concentrate on this semantics for the moment, one should not forget that an attraction of the language of second-order logic is the variety of interpretations it admits, depending to some extent on one's philosophy. The second-order variables may be taken to range over intensional entities (attributes), over concepts in Frege's sense, over classes (however understood) distinguishable in some way or other from sets. And then there is the reading of the monadic second-order language by the English plural, originally proposed by George Boolos. This last will occupy us later. A mathematical virtue of second-order logic interpreted by the standard semantics is its ability to characterize structures. One can credit already to Dedekind the proof that the natural numbers are characterized up to isomorphism by the second-order "Dedekind-Peano" axioms. The real numbers, Euclidean space of a fixed dimension, and even quite large initial segments of the cumulative hierarchy of sets can be similarly characterized. I want first to consider what Etchemendy says about a much discussed example, the continuum hypothesis. Let R be a second-order sentence that characterizes the real numbers up to isomorphism. Thus R has essentially only one standard model. Furthermore, the continuum hypothesis CH can be stated in the second-order theory of the reals; it simply says that any subset of the reals is either countable or of the cardinality of the reals. It follows that in the standard semantics, either R → CH or R → ¬CH is valid; equivalently, either CH or ¬CH is a consequence of R. Different writers have drawn different conclusions from this rather simple fact. Some have used it to argue that the continuum hypothesis must have an objective truth-value, contrary to the intuition of many, including some set theorists. Others have maintained that, intuitively, nei-

161 ther of the above sentences is a logical truth and have objected to the criterion given by standard semantics on that ground. Etchemendy's intuition is clearly in the second camp; he writes that "clearly neither it [CH] nor its negation is a logical truth" ((1990), p. 123). Though his formulation differs, it appears that Etchemendy has in mind, or at least would say the same about, sentences like our R → CH and R → ¬CH.8 This example gets special force from the fact that not only do we not know whether CH is true, it is known to be undecidable from the known set-theoretic axioms that have broad acceptance, including large cardinal axioms. Although there is research in progress aiming to decide CH, it is fair to say that any axioms proposed are controversial.9 In The Concept of Logical Consequence, Etchemendy gives an argument that does not depend on this special feature of CH.10 Mentioning the situation described above, he writes: This is often taken to be an objectionable feature of second-order logic: after all, why should the logical truth of a sentence depend on such highly abstract set-theoretic claims, claims that are not, intuitively, a matter of logic at all? (p. 123.)

However, he views the case as that where the logical truth of a secondorder sentence, say R → CH, is tied to the "ordinary" truth of a certain generalization, whose truth depends on that of CH as a set-theoretic statement. The same point could be made if R → CH is replaced by a secondorder sentence whose validity depends on the truth of a less problematic set-theoretic statement. For example we might let ZF2 be the conjunction 8

Cf. Etchemendy (2008), pp. 276-77, where he uses '¬CH' to mean something very close to our R → ¬CH, and 'CH' to mean something that would be true if the domain has cardinality less than that of the continuum. It appears that what he calls CH and ¬CH will both be true in a domain of individuals of cardinality less than that of the continuum. 9 See for example Woodin (2001). I note (May 2012) that Woodin has in recent years pursued a different approach to the continuum problem. 10 I am indebted to D. A. Martin for pointing out the different character of the arguments here and in "Reflections on consequence."

162 of the axioms of second-order Zermelo-Fraenkel set theory without Choice, and let A be some theorem of ZFC whose proof requires Choice. Assuming, as we normally would, AC in the metalanguage, ZF2 → A is true in any standard model, but with the right choice of a model of set theory refuting AC but including a strong inaccessible, there will be a "standard" model in which ZF2 → A fails. One might question this example because it depends on working within an eccentric model of set theory and talking of standard models within that. But the symmetry of a statement and its negation is not really needed for Etchemendy's point. Let B be the statement that there is a strongly inaccessible cardinal. Then ZF2 → B is not valid, but in order to witness this we need a standard model of ZF2. But the existence of such a structure is equivalent in ZFC to the existence of a strong inaccessible. The situation is quite parallel to that of the firstorder cases Etchemendy canvasses and that we have discussed, except that the entanglement is with higher set theory instead of arithmetic. Etchemendy evidently believes that neither CH nor ¬CH is a consequence of R, but it is not entirely clear how much he is willing to have his argument depend on that. In The Concept of Logical Consequence, evidently it does not. In the later paper (2008) he is somewhat tentative, but he says it is "reasonable" to conclude that neither R → CH nor R → ¬CH is a logical truth, so that the criterion derived from the standard semantics is not even extensionally correct (p. 277). One would think a more natural conclusion, already from the facts that Etchemendy advances, is that the entanglement with set theory that second-order logic exhibits when interpreted according to standard semantics disqualifies it as being logic in at least the basic sense. Although his reasons were quite different, Quine would be justified in saying that secondorder logic is "set theory in sheep's clothing."11 Etchemendy resists this conclusion, calling it "one of the more surprising and implausible conclusions of recent philosophy" (ibid.). He writes:

11

Quine (1970), p. 66 of 1st or 2d ed. For discussion of Quine's view of second-order logic see Parsons (2011), section II.

163 After all, second-order languages, like all languages, have a logical consequence relation. Some inferences employing the expressive devices of these languages are logically valid, and others are not (ibid.).12

He offers no suggestion as to what the logical consequence relation of the second-order language of set theory is. It would probably have to be intermediate between one that would make a standard axiomatic secondorder logic complete and the one based on standard models. My own guess is that beyond that, he hasn't a clue. The problem Etchemendy and others discern in the case of CH can be generalized, as is done by Peter Koellner in a recent paper.13 It has long been known that CH cannot be decided by any known large cardinal axioms. From the proof it follows, as Koellner observes, that validity for standard models in second-order logic can be altered by forcing even in the presence of very generous large cardinal assumptions. He contrasts it in this respect with other strong logics that have been considered, even Hugh Woodin's set-theoretically very complicated Ω-logic, which is robust under forcing given sufficiently strong large cardinal assumptions. III In recent years there has been what might be called a love affair between philosophers of mathematics and second-order logic. It is highly convenient for a number of programs. The outstanding example is probably the 12

A sentence later he writes, "But the idea that studying the logic of these languages is somehow not the business of logic is hardly a supportable conclusion." I'm not sure what he means by the "logic" of a language, given his own view that any choice of expressions as logical constants is arbitrary. However, he seems to want to align what is properly of interest to a logician with logical truth, consequence, and the like. What he says at the end of the paper about how logicians approach the study of certain languages is entirely reasonable. But that might just show that logicians are reasonably interested in more than logical truth, consequence, and closely related notions. 13 Koellner (2010). The author's focus is on strong logics, consequence relations that are stronger than that of first-order logic but have, at least given strong set-theoretic background assumptions, absoluteness properties. Ω-logic is identified as the strongest such first-order logic that has two such properties, generic invariance and faithfulness, if there is a proper class of Woodin cardinals. On Ω-logic see this paper or Woodin (2001).

164 neo-Fregean program of Crispin Wright and his associates, but other examples can be cited: the version of eliminative structuralism developed by Geoffrey Hellman, and the use by nominalists of George Boolos's interpretation of monadic second-order logic by the English plural. But in all these cases, what does the work is a more or less standard axiomatization of second-order logic (in Hellman's case augmented by modality). That fact suggests exploring the question of entanglement for axiomatic secondorder logic, without assuming the semantics discussed in the last section. The usual formulation of second-order logic adds to first-order logic rules for the second-order quantifiers parallel to those for first-order quantifiers, but universal instantiation and existential generalization are allowed only for variables. Then there is a comprehension schema ∃F∀x1 … xn[Fx1 … xn ↔ A(x1 … xn)], where F is an n-place second-order variable, and A must not contain F. Monadic second-order logic allows only one-place second-order variables, so that comprehension is allowed only for n = 1. But this is no limitation in the context of a theory that contains a pairing function. I have in mind what is called full comprehension, where A may be any formula, in particular containing bound second-order variables, so that the schema has impredicative instances. It is well known that this logic is incomplete for the standard semantics. In the literature choice principles are often added, but the resulting logic is still incomplete. It is remarkable that one does not find in the literature additional axioms or proposed axioms that would enable the derivation of additional formulae valid in the standard semantics, or in some other way intuitively logically true. But it is also not maintained that there are no such principles to be had. There is of course a semantics that yields a completeness theorem, the Henkin semantics. But that amounts to just putting into the domain of the second-order variables enough subsets of the domain to satisfy the axioms. The result is that it is very sensitive to what axioms are assumed and so does not offer a concept of validity against which axioms might be measured.

165 What would be an example of a valid sentence not derivable in axiomatic second-order logic? It's natural to turn to Gödel's incompleteness theorem. Let PA2 be the conjunction of the second-order Dedekind-Peano axioms (with just 0 and S as arithmetical primitives, taking the natural numbers to be the domain). In this language we can formulate the Gödel sentence of second-order arithmetic, call it G2. PA2 → G2 is valid in the standard semantics, but by Gödel's theorem it cannot be derivable in axiomatic second-order logic.14 However, it would be preferable not to rely on the standard semantics in asserting that it is valid. One way of seeing that is to observe that it does become provable by a modest detour into thirdorder logic; for example we could prove the consistency of second-order arithmetic by a truth-definition and deduce G2 from the consistency statement. That would suggest that there is a good reason why axioms for second-order logic other than choice principles are not proposed in the literature: I would conjecture that none has been found that is evident on logical grounds not going beyond second-order at least into third-order logic. Now consider third-order arithmetic. Its Gödel sentence G3 is arithmetical, so expressible in the language of second-order arithmetic. In the standard semantics, PA2 → G3 is a valid second-order sentence, not provable even in third-order logic. But again it can be proved by a detour into fourth-order logic to prove the consistency of third-order arithmetic. This process can evidently be iterated. If we now consider the full simple theory of types, i.e. ωth-order logic, all the sentences PA2 → Gn will be derivable, but there will of course be a new Gödel sentence Gω for ωth-order arithmetic, such that PA2 → Gω is not derivable but is valid in the standard semantics and plausibly logically valid in an intuitive way. The way of proving it that comes to mind involves introducing transfinite types, so that we have a logic of order ω+1. That in turn involves making the types cumulative, so that if our logic was not already set theory, it has become more like it. The iteration can be continued further. For long enough iterations, the question will arise: Where do the ordinals come 14

Second-order arithmetic does not include an axiom like the choice principles considered, but it is obviously provable that the domain of individuals is well-ordered, so that they are superfluous. Choice at the second-order level is another matter.

166 from? That would be another entanglement with set theory. There might be some iteration of the formation of types such that the resulting theory has a claim to be logic, while with a stronger iteration it no longer does. But I don't know of a proposal of this form advanced for this purpose.15 However, already beyond the recursive ordinals we may cease to have an axiomatizable theory, so that Gödel's theorem in its standard form does not apply. I don't know what the possibilities are of constructing axiomatizable theories that would extend axiomatic second-order logic still further or what might be natural as a semantic consequence relation that would validate sentences proved in such an extension of higher-order logic.16 However, it seems certain that something could be found to answer Etchemendy's apparent wish for something intermediate between the consequence relation yielded by the Henkin semantics for axiomatic secondorder logic and the consequence relation yielded by standard semantics. One suggestion is that one might restrict the standard semantics to constructible sets, although from Etchemendy's point of view that would have the disadvantage of validating R → CH.17 The problem is more likely to 15

Solomon Feferman and Kurt Schütte, in their classic analysis of predicative provability, considered systems of second-order arithmetic with orders indexed by ordinals generated "autonomously," that is, the orderings involved could be proved to be wellfounded in a theory assuming only the well-foundedness of a shorter ordering. Whether such an idea would be applicable in the present situation I do not know. In the Feferman-Schütte case, the ordinals involved are recursive ordinals. 16 In a discussion of these matters in a seminar at UCLA, D. A. Martin and Yiannis Moschovakis remarked that coming up with really new second-order principles was a quite non-trivial problem. They suggested Projective Determinacy as an example. The idea would be to formulate a sentence (or more likely a schema) PD such that R → PD is valid in the standard semantics if and only if Projective Determinacy holds as a set-theoretic proposition. It follows that its validity could be seen only using strong large cardinal assumptions. Although there are very good reasons for accepting Projective Determinacy, so that its status is not exactly that of CH, still the intuition of many will be that R → PD is not a logical truth. In any event, neither it nor another principle suggested to me by Peter Koellner would be proposed as a logical axiom. 17 Another possibility would be to consider as valid those second-order sentences whose validity for standard models can be proved in ZF or ZFC, thus making a settheoretic metalanguage do the work that we suggested would be done by a detour

167 be that there are too many possibilities, and set theory may throw up anomalous consequences for all of them.18 Although I believe that stronger arguments for this claim ought to be given, axiomatic second-order logic is entangled with set theory to a very significant degree, if not in quite such a problematic way as is the standard semantics. Thus far I have considered only second-order logic with full comprehension, which intuitively tries to capture validity in the standard semantics. In accord with the variety of interpretations that can be given to the second-order language, it is well also to look briefly at weaker comprehension assumptions. In the context of second-order arithmetic, a great variety have been studied by proof theorists.19 For illustration, however, it will suffice here to mention only one such logic, which has the advantage of predicativity. That is where one has only first-order comprehension; that is, the formula A in the comprehension schema is not allowed to contain bound second-order quantifiers. The intended models are very clear and straightforward: given a domain D of individuals, the n-place secondorder variables range over first-order definable subsets of Dn. Such a logic and its extensions have been used mainly to develop predicative systems of analysis and set theory. The sense of predicativity involved has been predicativity given the natural numbers. However, in more recent years they have found their way into logicist constructions, where the natural numbers are not assumed. The work involved is chiefly by John Burgess, Allen Hazen, and Richard Heck. Such work does something to answer a natural objection to Wright's neo-Fregean program: Full second-order logic, with an through higher-order logic. Such a notion would be generically invariant in the sense of Koellner (2010). It is faithful if ZF is Σ1-sound. 18 Attention should be called to the important paper Väänänen (2001), which discusses second-order logic both as an axiomatic theory and semantically, A point emphasized in the paper is that if second-order logic is used as a deductive instrument, it does not distinguish between standard and Henkin models. Väänänen considers a fixed deductive system and thus does not enter into reflections like those above on standard validities unprovable in axiomatic second-order logic, and on extensions of the theory to make them provable. This is probably due to his seeing that such extensions as would be at all natural would be third- and higher-order logics. 19 See for example Simpson (1999).

168 axiom forcing the domain to be infinite, is a very blunt instrument for developing basic arithmetic. What one can do with what I would call strictly predicative means still yields only a weak arithmetic, so that it may not be blunt enough. However, it does get the subject off the ground.20 IV The final example I wish to consider is plural logic. This derives from George Boolos's proposal made in 1984 of a way of reading the language of monadic second-order logic. I won't go into detail about that and its motivation, but the general idea was that the plural in natural language should not be regimented by the singular as is typically done in paraphrasing English sentences in logical notation, and attention to the plural brings to light a number of sentences that are not firstorderizable but can be rendered in monadic second-order logic. Plural logic has in the succeeding twenty years come to be a subject in its own right, and although the usual plural logics are virtually notational variants of monadic second-order logic, it is best to present them independently. What is called a PFO language (following Agustín Rayo (2002)) has in addition to the usual first-order apparatus plural variables xx, yy, zz, … and possibly plural constants. Logical predicates are the usual identity and a relation ∝, where t ∝ T is a formula if t is a singular (first-order) term and T a plural term, which can be read as 't is one of the Ts'. Quantifiers are admitted for plural variables in addition to singular. Primitive non-logical predicates can take only singular arguments. Identity of "pluralities" can be defined on a familiar model, xx == yy being defined as ∀z(z ∝ xx ↔ z ∝ yy). ∃xxA(xx) will be read as 'there are some xx's such that A(xx)'. An example might be "George Boolos ate some Cheerios". This might be rendered as '∃x(x is a Cheerio ∧ GB ate x)', but of course that is true if Boolos ate at least one Cheerio, while the plural logician will claim reasonably that what the sentence intends is more likely to be that there were some Cheerios, say in a bowl, and Boolos ate them. So the sentence will be rendered as '∃xx(xx are Cheerios ∧ GB ate xx)'.

20

See Burgess and Hazen (1998) and Burgess (2005), chapter 2.

169 To formulate a comprehension principle, we have to decide whether the truth of ∃xxA(xx) requires the "plurality" xx to have more than one element, or at least one, or is allowed to be empty. The third alternative makes PFO simply a notational variant of the usual monadic second-order logic. Although the first is natural given the talk of pluralities and answers to widely held intuitions about the English plural, most writers adopt the second.21 Then the comprehension schema takes the form ∃xA(x) → ∃yy∀x[x ∝ yy ↔ A(x)]. The need for this antecedent is no obstacle to an easy translation of PFO into the language of monadic second-order logic, and vice versa. For example, if A(F) is translated as A*(xx), then we render ∃F A(F) as ∃xxA*(xx) ∨ A**, where A** results from A*(xx) by replacing each formula t ∝ xx by t ≠ t. The restriction of nonlogical predicates to singular arguments may seem artificial, and indeed there is no obstacle to admitting predicates with plural arguments. The resulting languages are in Rayo's terminology PFO+. For the logic, this raises the question of extensionality. In keeping with intuitions about plurality, it is usual to assume that places occupied by plural arguments are extensional and therefore to assume a schema of extensionality: xx == yy ∧ A(xx) → A(yy). Predicates with plural arguments correspond to second-level predicates in second-order logic.22 21

However, Schein (2006), §0.1, offers examples to show that plural phrases in English can designate empty "pluralities," which would favor the first alternative. The most convincing is "The moons of Venus are zero in number." 22 I am assuming here, with Rayo's PFO+, that the same predicate cannot take both singular and plural arguments. That preserves the connection to second-order logic. But it deviates from the use of the plural in English and related languages. That point is emphasized in Schein (2006) and in the plural-logical work of Byeong-Uk Yi (2005) and (2006).

170 The plural reading of the variables xx, … seems to some writers to make the comprehension principle particularly obvious. For example John Burgess writes: For plural comprehension just says that, for any formula A(u), there are some objects, the xx, such that any object u is among them if and only if A(u). And of course there are such objects, namely the objects for which A holds.23

The example set by Boolos, of insouciance about comprehension assumptions when this logic or the parallel reading of monadic second-order logic is used, has been widely followed. I will argue that this attitude is another example of neglect of the entanglement of logic and mathematics. Suspicion is aroused by some of Boolos's examples from mathematics, in particular set theory. Consider the following: There are some sets of which every set that is not a member of itself is one, but there is no set of which every set that is not a member of itself is a member, as the reader, understanding English and knowing some set theory, is doubtless prepared to agree.24 … once we cut through the verbiage, we do find it trivial that there are some sets none of which is a member of itself and of which each set that is not a member of itself is one.25

The force of these examples comes from the apparent obviousness of the claim that there are some sets such that …, even though there is not a set of such sets, so that, unless we take the domain of individuals to be restricted in some way, we cannot interpret the plural quantifier in the usual settheoretic way of interpreting second-order quantifiers. Boolos, however, seems to take the meaning of these formulations to be more or less transparent, although he does include the qualifier "knowing some set theory" about his reader. In spite of my own concerns, I am ready to agree as well. But how could Boolos be sure that a reader who knows some set theory is not relying on the fact that (taking 'there are some sets' as a familiar se23

(2005), p. 213. I have changed the notation to fit my own. Cf. Burgess (2004), p. 197. 24 Boolos (1984), p. 66 of reprint. 25 Ibid., p. 69.

171 cond-order quantifier) the statements amount to trivial theorems of set theory with predicative classes (and not dependent on "higher" axioms such as Power Set or Replacement)? One can think of this theory as formulated in second-order logic with first-order comprehension, which does not force the interpretation of the second-order variables as ranging over classes rather than some alternative. Boolos was surely aware that, in spite of the translation he spelled out using the English plural, set theory with monadic second-order logic is not a conservative extension of ZF as is GB (i.e. ZF plus predicative classes) and is in fact proof-theoretically stronger. It's somewhat hard to believe that the step from one to the other is made on the basis of intuitions about the plural in English. That the meaning of the statements he puts forth is not transparent without assuming quite a bit of mathematics might be made clearer by looking at a more complex example: There are some numbers such that any number n is one of them if and only if there are not some numbers such that 1 is one of them, for every number x that is one of them, 2x is one of them, for every e (as code of a Turing machine) such that for every number x {e}(x) is defined and is one of them, 3 x 5e is one of them, and n is not one of them. This is a plural translation of the comprehension axiom of second-order number theory asserting the existence of Kleene's class O of recursive ordinal notations. Unlike the examples cited from Boolos, it is impredicative. Boolos introduced plural logic in tandem with a claim about ontological commitment, that monadic second-order quantifiers read plurally do not introduce any new ontological commitment, over and above what a first-order theory is committed to. That this claim is convenient for nominalists is obvious. Boolos was no nominalist, but he did use it to argue for one nominalist-like thesis, concerning proper classes in set theory, about which he had long been skeptical. The claim was that adding monadic second-order apparatus to set theory did not introduce commitment to classes or the like, including Fregean "concepts". In other cases, many of them

172 natural language sentences similar to those used in logic exercises, he had argued quite effectively that the plural locutions that he used to read the language of monadic second-order logic did not involve reference to sets or classes. Above I have presumed that plural variables in plural logic range over "pluralities", but although cognates of that term have been used in the history of logic and set theory (for example Cantor's Vielheit), it is not very informative. Moreover, it leads to regimentation of the plural by the singular, which was one of the objections to readings in terms of sets or classes. Boolos's claim, which following Øystein Linnebo we might call the Ontological Innocence Thesis, was criticized early on in spite of the already mentioned convenience for certain programs. Quine, who coined the term "ontological commitment", famously insisted that the ontological commitment of a theory can be assessed only when the theory is regimented in terms of first-order logic. That might seem excessively rigid, but if the (more or less) logical apparatus of the theory is extended, it seems reasonable to say that this also gives rise to an extended notion of ontological commitment. Many years ago I maintained just that for a modal language, and indeed what is called modal nominalism has come to be distinguished from the strict nominalism of Goodman.26 The suggestion might then be made that for the case of plural logic, we should introduce a notion of plural commitment. "Some critics live in Cambridge" would be committed to critics, but not plurally committed to them, while "Some critics admire only one another", because it is not firstorderizable, would be plurally committed to critics. In English, a mark of such commitment would be the use of plural pronouns to indicate crossreference; for example, we make the latter sentence more explicit by paraphrasing it as: Some critics admire no critic who is not one of them.27

26

See my (1971). Although there is much in the paper that I no longer agree with, that does not touch the present point. 27 Or possibly: Some critics admire no one who is not one of them. The logical issues are unaffected by which reading is chosen.

173 If we accept this suggestion, we will have to admit that plural commitment admits of degrees. Although plural logicians do not find any motivation for restrictions on comprehension to insure predicativity, the fact remains that a first-order version of the comprehension schema of plural logic is substantially weaker in the context of a mathematical theory such as arithmetic or set theory than is a stronger predicative schema, and the latter is still weaker than the full impredicative schema. Since they differ with respect to for what A one is prepared to assert that there are some objects (e.g. numbers, sets) x such that x is one of them iff A(x), it is hard to see why one should say that the stronger schemata do not postulate additional entities. If they do not, then ontological commitment may just not have the significance that both nominalists and many of their opponents attribute to it, or than Boolos seems to attribute to it in the case of proper classes. That might be a victory for the Innocence Thesis, but it would be a Pyrrhic victory.28 The main thesis of this section is that the reliance on linguistic intuitions concerning the plural to judge the truth and commitments of axioms used in powerful mathematical theories represents neglecting the entanglement of logic and mathematics. This claim might be reinforced by the observation that the semantics of plurals is a matter of disagreement and uncertainty among linguists. A disagreement relevant to the issues discussed above is whether plurals are "predicational", that is whether the logical form of plural terms should contain predicates.29 A positive answer to 28

John Burgess has urged for a number of years that ontological economy does not have the significance for science and mathematics that nominalists attribute to it. See for example Burgess (1983) and Burgess and Rosen (1997). Nonetheless, he seems to embrace the Innocence Thesis. I am not sure of his reasons, but one might guess from the treatment of the status of second-order logic at the end of Fixing Frege that he sees no other way of understanding the second-order language of set theory. My own view of the matter is that quantification over absolutely all sets cannot be taken quite at face value; see my (1983), essays 8 and 9, and (2006). 29 This is the view of Schein in his earlier work (1993), one of the most extensive treatments of plurals. A predicational view was maintained in an earlier joint paper with James Higginbotham, but Higginbotham later abandoned this view; see Higginbotham (1998). In that essay Higginbotham favors a primitive concept of plurality, although he prefers the Russellian term 'class as many'. It appears from (2006), §1.4.2,

174 this question would give a motivation for preferring a predicative plural logic, while a negative answer might undermine such motivation, thus favoring the view of most advocates of plural logic. Some of the difficulties about the semantics of plurals might be attributed to a "singularist" bias in the tools at semanticists' disposal.30 But even the supporters of the two views that are my target here, the Innocence Thesis and the claim that full comprehension in plural logic is obvious, would surely wish to see the semantics of plurals better understood. V In section II we proposed that the entanglement of standard secondorder logic with higher set theory is a reason why it should not be considered logic "in the basic sense." One might well ask what that means. That could be the subject of more than one other paper. A short answer would be: a normative canon of reasoning about objects and subject matters very generally, perhaps even for anything that can properly be called thought. On the latter reading, it would have the property of Generality that John MacFarlane extracts from his study of Frege and Kant.31 That would suggest that basic logic should have a minimum of potentially controversial presuppositions. Given the arguments about the law of excluded middle, it is not evident that even classical first-order logic satisfies that condition. A related question closer to the main themes of this paper is, "When is the entanglement of a proposed logic with mathematics problematic?" The answer is surely that being problematic is a matter of degree, so that we should not expect a sharp line between unproblematic and problematic entanglements, any more than there is a sharp line between unproblematic and problematic mathematics. In the above I have treated the entanglement of first-order logic as unproblematic. But one might object that full polyadic first-order logic is problematic because the construction of countermodels can be infinite and that Schein still holds a version of the predicational view. But the issue is not at center stage in his intricate discussion. 30 Schein (2006), §1.4.1, seeks to overcome this by using the plural in the metalanguage. 31 MacFarlane (2002).

175 noneffective, with the consequence that validity is undecidable. Someone uncomfortable with this degree of entanglement of logic and mathematics might wish to argue that only monadic logic, or possibly something still more limited, is truly logic. I don't think there is a knock-down argument against such a position. It might even have its relevance to some applications of logic to natural language. Against it one can cite the apparent fact that the mathematics involved in constructing countermodels to nonvalid formulae is virtually uncontroversial among mathematicians and mathematical logicians. However, the matter is not so simple as it seems. We have a general method for constructing a countermodel to a formula on the assumption that it is not derivable in a standard deductive system. But this assumption is a consistency assumption, and in some cases, such as that of sentences of higher set theory, a quite powerful theory can be required to prove consistency. The case is hardly different for second-order logic with first-order comprehension or with only a small number of layers of predicative classes. In the monadic case, however, the nonderivability assumption does not pose a problem, because there is a straightforward decision procedure that either yields the answer of validity or constructs a finite countermodel. From some points of view even that is not wholly unproblematic, since the size of the countermodel can grow exponentially.32 There is a new set of problems when impredicative comprehension is combined with an infinite domain of individuals, since there one loses completeness for the most natural semantics. As we have seen in section III, the project of extending second-order logic leads us gradually into higher set theory. If we pose questions in terms of standard semantics, fairly simple examples can lead into issues in set theory that are far from being resolved or even shown to be resolvable. Satisfiability can be equivalent not to the consistency of statements we do not know how to decide, but to their truth or falsity. The case of CH illustrates the distinction of the two, since we know that CH and ¬CH are consistent with ZFC and known large cardinal axioms, so long as the latter are consistent. One can hardly 32

A matter I have not been able to investigate so far is the advocacy by some German writers of a much narrower conception of logic. The most sophisticated is probably Michael Wolff, in his (1995) and (2004). I believe that one of his motives is a wish to avoid the entanglement that exists even for first-order logic.

176 doubt that the entanglement of second-order logic with mathematics is more problematic than that of first-order logic.

References Bernays, Paul, 1930. Die Philosophie der Mathematik und die Hilbertsche Beweistheorie. Blätter für deutsche Philosophie 4 (1930), 326-367. Reprinted with Postscript in Bernays (1976), pp. 17-61 and with English translation in Bernays (forthcoming). Bernays, Paul, 1969. Bemerkungen zur Philosophie der Mathematik, In Akten des XIV Internationalen Kongresses für Philosophie, Wien 1968, III, 192-198. Wien: Herder. Reprinted in Bernays (1976), pp. 170-175 and with English translation in Bernays (forthcoming). Bernays, Paul, 1976. Abhandlungen zur Philosophie der Mathematik. Darmstadt: Wissenschafliche Buchgesellschaft. Bernays, Paul, forthcoming. Essays on the Philosophy of Mathematics. Wilfried Sieg, W. W. Tait, Steve Awodey, and Dirk Schlimm, eds. Chicago and La Salle, Ill.: Open Court. Boolos, George, 1984. To be is to be the value of a variable (and to be some values of some variables). The Journal of Philosophy 81, 430-449. Reprinted in Logic, Logic, and Logic, pp. 54-72. Cambridge, Mass.: Harvard University Press, 1998. Burgess, John P., 1983. Why I am not a nominalist. Notre Dame Journal of Formal Logic 24, 93-105. Burgess, John P., 2004. E pluribus unum: Plural logic and set theory. Philosophia Mathematica (3) 12, 193-221. Burgess, John P., 2005. Fixing Frege. Princeton University Press. Burgess, John P., and Allen Hazen, 1998. Predicative logic and formal arithmetic. Notre Dame Journal of Formal Logic 39, 1-17. Burgess, John P., and Gideon Rosen, 1997. A Subject with no Object: Strategies for a Nominalistic Interpretation of Mathematics. Oxford: Clarendon Press.

177

Etchemendy, John, 1990. The Concept of Logical Consequence. Cambridge, Mass.: Harvard University Press. Reissued Stanford, California: CSLI, 1999. Etchemendy, John, 2008. Reflections on consequence. In Douglas Patterson (ed.), New Essays on Tarski and Philosophy, pp. 263-299. Oxford University Press. Higginbotham, James, 1998. On higher order logic and natural language. Proceedings of the British Academy 95, 1-27. Hilbert, David, and Paul Bernays, 1939. Grundlagen der Mathematik II. Berlin: Springer. 2d ed. 1970. Koellner, Peter, 2010. Strong logics of first and second order. The Bulletin of Symbolic Logic 16, 1-36. MacFarlane, John, 2002. Frege, Kant, and the logic in logicism. Philosophical Review 111, 25-65. Parsons, Charles, 1971. Ontology and mathematics. Philosophical Review 80, 151176. Reprinted with Postscript in Parsons (1983).33 Parsons, Charles, 1983. Mathematics in Philosophy: Selected Essays. Ithaca, N. Y.: Cornell University Press. Parsons, Charles, 2006. The problem of absolute universality. In Agustín Rayo and Gabriel Uzquiano (eds.), Absolute Generality, pp. 203-219. Oxford: Clarendon Press. Parsons, Charles, 2011. Quine's nominalism. American Philosophical Quarterly 48 (2011), 213-228. Quine, W. V., 1970. Philosophy of Logic. Englewood Cliffs, N. J.: Prentice-Hall. 2d ed. Cambridge, Mass.: Harvard University Press, 1986. Rayo, Agustín, 2002. Word and objects. Noûs 36, 436-464. Schein, Barry, 1993. Plurals and Events. Cambridge, Mass.: MIT Press. 33

Some disturbing typographical errors in the reprint of this essay are corrected in the paperback edition of the book (2005).

178 Schein, Barry, 2006. Plurals. In Ernest Lepore and Barry C. Smith (eds.), The Oxford Handbook of Philosophy of Language, pp. 716-767. Oxford University Press. Simpson, Stephen G. 1999.

Subsystems of Second-Order Arithmetic.

Springer-Verlag,

Väänänen, Jouko, 2001. Second-order logic and the foundations of mathematics. The Bulletin of Symbolic Logic 7, 504-520. Wolff, Michael, 1995. Die Vollständigkeit der kantischen Urteilstafel. Klostermann.

Frankfurt:

Wolff, Michael, 2004. Klostermann.

Frankfurt:

Abhandlung über die Prinzipien der Logik.

Woodin, W. Hugh, 2001. The continuum hypothesis. Notices of the American Mathematical Society 48, 567-576, 681-690. Yi, Byeong-Uk, 2005. The logic and meaning of plurals I. Journal of Philosophical Logic 34, 459-506. Yi, Byeong-Uk, 2006. The logic and meaning of plurals II. Journal of Philosophical Logic 35, 239-288.

Validity of Inferences‡ Dag Prawitz*

Abstract Deductive inference gives us conclusive grounds for beliefs and assertions, and may even compel us, as is generally acknowledged since the time of Plato and Aristotle. But there is no generally accepted account of what is to be required of an inference in order that it is to have this justificatory and compelling power. That the inference is valid is clearly not a sufficient condition, if validity of an inference is defined as truth preservaton for all variations of the content of the non-logical constants involved. It has been suggested that the agent has to be required to know the inference to be valid, but it may be argued along the lines of Lewis Carroll's regress that neither is this a sufficient condition. In any case it is too stringent, since justification would never get off the ground if we first had to show that the inferences used in a justification are valid. The paper suggests that we have to rethink the notion of inference to account for how we acquire knowledge by making inferences. An inference is first of all an act in which an agent operates on given grounds for the premisses ‡

This is a revised version of the paper I gave at “the 2nd Lauener Symposium on the Occasion of the Presentation of the Lauener Prize 2006 to Dagfinn Føllesdal”. It is not directly related to Føllesdal’s works, but shares with many of them an interest in modality. The main ideas of the paper were also presented in a somewhat different form at talks I gave in the same year at conferences at Pisa and Stockholm and at the Kant lectures at Stanford and, in a form more similar to the present one, in 2007 at talks at Stockholm, Bologna, and Pisa. I thank participants at these occasions for stimulating discussions and professors Cesare Cozzo, Per-Erik Malmnäs, Per Martin-Löf, and Dugald Murdoch for commenting on early versions of this paper. (Added 2012): The ideas of this paper about how to account for the fact that beliefs and assertions are justified by valid inferences have been followed up by a series of other papers, which were presented at later occasions, but have already been published. They are: "Inference and Knowledge", in: The Logica Yearbook 2008, M. Pelis (ed), pp 175-192, College Publications, London 2010; "Proofs and Perfect Syllogisms", in: Logic and Knowledge, C. Cellucci et al (eds), pp 385-402, Cambridge Scholars Publishing, Newcastle upon Tyne 2011; "The Epistemic Significance of Valid Inferences", Synthese 187, pp 887-898. (URL = http://dx.doi.org/10.1007/s11229011-9907-7). * Stockholms Universitet

180 in order to get a ground for the conclusion. It can be defined as valid if the operation does yield a ground for the conclusion when applied to grounds for the premisses. Getting in possession of a ground for the conclusion by performing the act, the agent becomes justified in holding the conclusion true, and if aware of this possession, she even gets compelled to do so.

When the systematic study of inferences began with Aristotle, there was in Greek culture already a flourishing argumentative practice with the purpose of supporting or grounding one’s assertions or discrediting assertions made by others. Such a practice naturally gives rise to the question what characterizes a good argument, one that really justifies its conclusion. In the discussion of this question, the remarkable idea arose that there are deductive arguments or proofs that conclusively establish the truth of their conclusions. Logic started as a reflection over the inferences that such maximally good arguments are built up of. I want to claim that there is still a need of such reflections. I shall here focus on two features of valid inferences: their power to justify assertions and beliefs and their capability of bringing into being a kind of modality. It is noteworthy that both features were invoked from the very beginning in the study of inferences. In an often quoted passage in Prior Analytics, Aristotle says: “A syllogism is a form of speech in which, certain things being laid down, something follows of necessity from them, i.e. because of them, i.e. without any further term being needed to justify the conclusion.”1

Assertions may be made on grounds of various kinds and various strengths. I take it to be a fact, acknowledged by most people at least since the time of Aristotle, that the making of inferences is one way to acquire grounds or justifications and that there is a kind of inference that delivers conclusive grounds, grounds of maximal strength. Not only may an inference justify an assertion or belief, it may also compel us to hold its conclusion true. The idea that an inference in some

1

Ross (1949) p. 287.

181 way compels us is present already in Parmenides and Plato2. It is reflected today in the common practice of inserting the word “must” when announcing an inference. For instance, after having derived a contradiction from the assumption that the square root of 2 is rational, a textbook may continue: “Hence, √2 must be irrational”.3 As one also says, in a more explicit recognition of the compelling force of a valid inference, “by pain of irrationality”, one has to accept the conclusion of a valid inference, given that one has accepted the premisses. To say that there are deductive inferences that give rise to conclusive and even compelling grounds or proofs is of course not to say that there are infallible roads to knowledge. One can never rule out that one is mistaken about what one thinks is a ground or a proof of a sentence. But if the sentence turns out be false, then we say that we did not really have a proof, what we took to be a proof was in fact not a proof. Hence, the possibility of mistakes does not have any bearing on the existence of conclusive proofs. It does not detract from the fact that a valid deductive inference does deliver a conclusive ground for its conclusion given conclusive grounds for the premisses. However, it is one thing that there is this idea of justification by inference and that we take it to be a conceptual truth that a valid deductive inference delivers a conclusive ground for the conclusion given conclusive grounds for the premisses, and another thing to explain how, if at all, there can be such justifications. Presumably such an explanation must rest on what it is for an inference to be valid. It is a task for philosophy to account for the validity of an inference in such a way that it follows or is directly included in the account that a valid inference lends itself for justifications – in particular, that a valid deductive inference delivers a conclusive ground for its conclusion, given conclusive grounds for its premisses. In this paper I shall restrict myself to deductive inferences and conclusive grounds, and henceforth, when I speak of inferences or grounds, I always have in mind deductive inferences or conclusive grounds, respectively. To have a ground for a sentence A, as I use the term, will always mean to be justified in holding A true and to know that A. 2 3

For an instructive discussion of some of Plato's views on this, see Cozzo (2005). This is an authentic example from a textbook of mathematical analysis.

182 1. Logical consequence as defined by Tarski In logic, as well as in philosophy in general, the dominant view concerning the notions of logical consequence and valid inference has been for long that the essential questions concerning the former notion were settled by Alfred Tarski in his paper “On the Concept of Logical Consequence”4, and that the latter notion is accounted for by saying that an inference is valid if and only if its conclusion is a logical consequence of its premisses in the sense defined by Tarski. Tarski himself did not think that he had settled completely how the concept of logical consequence is to be defined adequately. He especially mentioned the open problem of how to draw the line between logical and non-logical terms, a problem which is of course crucial in this context, since the whole idea of Tarski’s definition is to vary the content of the nonlogical terms, defining a sentence A to be a logical consequence of a set Γ of sentences if and only if every variation of the content of the non-logical terms occurring in the sentences involved that makes the sentences of Γ true makes A true as well. Furthermore, it should be said that Tarski was not concerned with the validity of inferences but restricted himself to the concept of logical consequence. Nevertheless, Tarski thought that his definition quite well caught what it means to say that a sentence follows logically from other sentences. He especially stressed that a modal element, namely the one discussed above in connection with valid inferences, was taken care of, saying: “In particular, it can be proved, on the basis of [my] definition, that every consequence of true sentences must be true”5 (my italics).

One main claim of my paper is that the concept of valid inference cannot be obtained in this way from the notion of logical consequence as defined by Tarski. This is not an original claim. Critical voices against Tarski’s definition of logical consequence, in particular when thought of as also defining the validity of inferences, have been raised before, for in4

Published 1936, first in Polish and then the same year in German under the title “Über den Begriff der logischen Folgerung”. The English translation appears 20 years later in Tarski (1956). 5 Tarski (1956), p. 417.

183 stance by John Etchemendy, Göran Sundholm and myself6. One aim of this paper is to isolate what I think is the main shortcoming of the dominant view. Focusing on inferences as a means to get grounds for assertions and beliefs, I shall argue that this raison d´être of inferences cannot be explained or accounted for when their validity is equated with the relation of logical consequence holding between premisses and conclusion as defined by Tarski. The issue will be kept apart from the dispute between classical and constructive reasoning. 2. Valid and logically valid inferences I am not denying that the idea of varying the content of the non-logical terms is a natural and fruitful approach even when one is concerned with inferences. The idea of such variations goes back to Aristotle whose standard way of showing that an inference is not valid was exactly to vary the content of the non-logical terms in the involved sentences in order to get a counterexample where the premisses are true and the conclusion is false. It was taken up by Bernard Bolzano and is the main element in his definition of logical consequence7. Bolzano's definition differs in some respects from Tarski's8, but as far as we are concerned in this discussion, the two definitions come essentially to the same in making the non-existence of a counterexample to a condition for logical consequence – not only a necessary condition, but a sufficient one as well. So far I have not spoken of an inference being logically valid, only of it being valid. When I said that Tarski’s analysis of the notion of logical consequence was, and still is considered to answer the question what it is 6

For instance, Etchemendy (1990), Sundholm (1998a) (a paper much influenced by Per Martin-Löf, see e.g. Martin-Löf 1985), Prawitz (1985) and (2005). I also want to mention a talk by Per Martin-Löf presented at the conference in Pisa mentioned in footnote ‡ on the first page of the paper. 7 Bolzano (1837). His term is "Ableitbarkeit". As already his choice of words suggests, Bolzano was interested in epistemological aspects of this notion. He explicitly connects it with the correctness of an inference, saying for instance that the relationship of logical consequence is "a remarkable thing since, knowing that it applies, it makes us able to infer immediately from the known truth of the [antecedents] the truth of the [succedent]". 8 One of the differences is referred to below.

184 for an inference to be valid, I should have said logically valid. But, for obvious reasons, when adopting the view that Bolzano-Tarski’s analysis of the notion of logical consequence settles what it is for an inference to be correct, one usually does not distinguish between the two notions. Yet, from an intuitive point of view, such a distinction is natural to make. There are inferences that one would be inclined pre-theoretically to consider as valid but not as logically valid; a simple natural language example is the inference from the two premisses "Adam is longer than Beatrice" and "Beatrice is longer than Carlo" to the conclusion "Adam is longer than Carlo", a mathematical example is inference by mathematical induction over the natural numbers (as usually formulated in first order logic, where the concept of natural number is not defined impredicatively but is taken as a primitive notion). What does it then mean to say that an inference is logically valid? An obvious suggestion is that an inference is logically valid if its validity does not depend on the non-logical terms but only on the logical terms that occur in the sentences involved. This may be spelled out by using the very idea of the Bolzano-Tarski definition of logical consequence in this context, saying that the logically valid inferences are those that remain valid regardless of how we vary the content of the non-logical terms. We may think of a variation of the content of the non-logical terms of a sentence along the lines of Bolzano as being obtained by simply replacing non-logical terms in the sentence by other terms of the same syntactical category. Letting φ stand for such substitutions, i.e. for assignments to the non-logical terms in the sentences A1, A2, …, An, and B of other non-logical terms of the same syntactical category, and denoting the reφ φ sult of making such a substitution φ in Ai and B by Ai and B , respectively, we may make the explicit definition: An inference from A1, A2, …, and An to B is logically valid, if and onφ φ φ φ ly if, for every φ, the inference from A1 , A2 , …, and An to B is valid.

185 Or, better, we may follow the way of Tarski and let φ stand for valuations, i.e. assignments of arbitrary values (of appropriate categories) to the nonlogical terms, and define: An inference from A1, A2, …, and An to B is logically valid, if and only if, for every φ, the inference from A1, A2, …, and An to B is valid under φ. What is obtained by following the Bolzano-Tarski approach in this way is thus a reduction of logical validity to validity. The crucial task is then to define what it is for an inference to be valid (or valid under an assignment). We could in the same way distinguish between consequence and logical consequence, and reduce the latter notion to the first one by saying that B is a logical consequence of Γ, when B remains a consequence of Γ for every variation of the content of the non-logical terms occurring in A and in the sentences of Γ.9 But if we just stay within the framework of Bolzano and Tarski, this distinction become pointless, because the notion of consequence will then collapse into that of truth of the corresponding material implication. The same holds of course, if we make the distinction between valid and logically valid inference proposed above, but do not move outside the framework of Bolzano and Tarski when it comes to analyzing the notion of valid inference. It would then only remain to say that an inference is valid if either one of the premisses is false or the conclusion is true, but clearly no one is interested in equating the validity of an inference with such a relation between the truth-values of the involved sentences. Already the inability of the Bolzano-Tarski approach to distinguish in a satisfying way between consequence and logical consequence should make one suspicious of the possibility of analyzing the concept of (logically) valid inference by using their notion of logical consequence. But the weakness of such an analysis is seen most clearly when we ask why an inference whose conclusion is a logical consequence of the premisses in the Bolzano-Tarski sense should have the power to justify a belief in the truth of the conclusion. One can hardly expect that there is a 9

This is a proposal made in Prawitz (2005).

186 satisfactory answer to that question by referring to their notion of logical consequence if one bears in mind the following facts. To say that the conclusion of an inference is a logical consequence of its premisses in the sense of Bolzano-Tarski amounts just to saying two things: 1) that the inference preserves truth (which only means here that if the premisses are true then so is the conclusion) and 2) that all inferences of the same logical form do the same. Now, although the first property is certainly relevant for the question whether the inference has the power to justify a belief in the conclusion (being a necessary condition for that), it is obviously not sufficient for the inference to have this power. The second property seems not even relevant to this question; it concerns other inferences of the same logical form as the first one, and is relevant to the question whether they have the power to justify their conclusions. Property 2 is admittedly a necessary condition for the inference to be logically valid, but is not sufficient for the simple reason that it only amounts to the requirement that all inferences of the same logical form satisfy property 1, and, as already said and as anyone agrees, property 1 is not sufficient for an inference to have the power to justify beliefs or to be valid. Nevertheless, the view that an inference is (logically) valid if and only if its conclusion is a logical consequence of the premisses in the sense of Bolzano and Tarski is, as noted, the prevailing one. In the sequel, I shall for the sake of brevity sometimes refer to it as the Bolzano-Tarski view or notion of valid inference (even if Tarski was primarily concerned with the notion of logical consequence and did not explicitly speak about valid inferences). In consideration of the dominance of this view, I shall devote part of the following sections to showing in more detail why it does not seem possible to account for the use of inferences to justify beliefs from that point of view. Admittedly, it is not immediately obvious what such an account is to consist of. The requirement that some such account should flow from an analysis of the concept of valid inference is certainly a fair one, but we must discuss how it is to be formulated more precisely. This is an important issue for any concept of valid inference and will also be a topic of the next sections.

187 3. How valid inferences do not justify their conclusions What conditions on inferences have to be satisfied in order that a person is to be justified in holding the conclusion of an inference true? This question is more difficult to answer than is sometimes realized. Starting from the Bolzano-Tarski view of valid inferences, shall we say simply: (1) If an inference J from a sentence A to a sentence B is valid, then a person who is justified in holding A true is also justified in holding B true. This is sometimes affirmed. It is even said that if B is a logical consequence of A, then without further investigation we may conclude that B is true given that we know that A is true. However, it is enough to consider an inference with a great epistemic distance between the premiss and the conclusion to see that implication (1) does not hold in general. Let B be a difficult mathematical theorem, for instance Fermat’s last theorem, and let A be the conjunction of axioms or the starting points for a proof of the theorem. We may grant that Andrew Wiles was justified in holding true all the propositions from which he started when proving Fermat’s last theorem and that the one-step inference from these starting points to Fermat’s last theorem is valid. But these two assumptions are clearly not enough to make Wiles or anybody else justified in holding Fermat’s last theorem true; they were, we may assume, satisfied long before Wiles gave his proof. It is instructive to reflect upon the fact that Wiles first had to withdraw a purported proof because of a discovered gap. Let us suppose that when he later filled the gap, there was no expansion of the starting points and that he was justified in holding them true when he gave his first incomplete proof. Then, on these assumptions, the one step inference that we are considering, that is the one inferring Fermat’s last theorem from the conjunction of the starting points of this withdrawn proof, was indeed valid, although, as long as the gap remained, one did not see the validity. Hence, according to implication (1), Wiles would have been justified in asserting Fermat’s last theorem as soon as he gave his first incomplete proof, contrary to the fact that in reality he soon afterwards had to withdraw it because of the discovered gap.

188 This clearly shows that it is not the validity of the inferences of a proof per se that makes one entitled to assert or believe in the conclusion of the proof. What makes one so entitled is a sufficiently detailed proof, in other words, a proof without gaps. But what is a gap-free proof? Should we say that each inference of the proof is to be seen to be valid, and what would that mean? Perhaps one should conclude from what is said above that the antecedent of implication (1) has to be strengthened to say not only that the inference is valid but also that the person in question knows it to be valid. Let us consider this proposal: (2) If a person knows the inference J from a sentence A to a sentence B to be valid, and is justified in holding A true, then she is also justified in holding B true. From an intuitive understanding of what it is for an inference to be valid, we certainly expect this implication to hold. However, we should also ask whether the antecedent of this implication describes a way to acquire new knowledge; otherwise, the fact that the implication holds is of little interest. In other words, we should ask whether the following holds: (3) We can become justified in holding a sentence B true (and hence get to know that B is true) by first getting to know two things: 1) that an inference J from a sentence A to the sentence B is valid, and 2) that A is true. If we think that this is an adequate description of how we acquire new knowledge by using inferences, we may test a proposed definition of the concept of valid inference by investigating whether it is able to support (3), i.e., whether the statement comes out true when we replace “valid” with what it is to be valid according to that definition. Such a test on the Bolzano-Tarski notion of valid inference is in effect suggested by John Etchemendy10. According to him the test comes out negatively, because in gen10

Etchemendy (1990), p. 93: ”A logically valid argument must, at the very least, be capable of justifying its conclusion. It must be possible to come to know that the con-

189 eral we are not able to know that J is valid in the proposed sense without first knowing that the conclusion B is true.11 This is an interesting argument, but I think that it is not at all obvious how it can be supported.12 However, this is not something that I shall discuss here, because I think that the statements (2) and (3) are on the wrong track when we are to describe how we normally use inferences to justify beliefs. Implication (2) was formulated as a possible response to the problem that arose when noting that the mere validity of an inference whose premisses are known to be true does not justify a person in asserting the conclusion, it then being suggested that it is only when a person sees the validity of the inference that she is so justified. Identifying 'seeing the validity of an inference' with 'knowing the inference to be valid', such knowledge was suggested as a necessary condition for an inference to justify a belief, which we may formulate as follows: (4) It is only when a person knows an inference to be valid and its premisses to be true that the inference justifies her in holding the conclusion true. clusion is true on basis of the knowledge that the argument is valid and that its premisses are true.” 11 Etchemendy’s argues that the Bolzano-Tarski view of valid arguments ”leaves such arguments impotent as a means of justifying their conclusions” saying: ”Tarski’s account equates validity with the joint truth preservation of a collection of arguments. … In general, it will be impossible to know whether an argument is a member of such a collection, hence whether it is “valid”, without antecedently knowing the specific truth-values of its constituent sentences”. 12 Of course, Etchemendy does not want to deny that in many cases we can come to know that the conclusion of an inference is a logical consequence of the premisses in the sense of Bolzano and Tarski without knowing antecedently the truth-values of the involved sentences. What he is saying is that this is not possible in general, in other words that there are cases where the validity of an inference J in the proposed sense cannot be established without first establishing the truth of the conclusion B or the falsity of A. To show that this is so, we have thus to find sentences A of B of that kind and then develop a theory about the order in which things can be demonstrated, showing that the fact that B is a logical consequence of A in Bolzano-Tarski’s sense cannot be demonstrated before having demonstrated that B is true or A is false. We would then have shown that there are valid inferences in the proposed sense that cannot be used to gain new knowledge.

190

But further reflection shows that it is doubtful that we could ever use inferences to acquire knowledge, if we had first to establish their validity. Furthermore, as far as the Bolzano-Tarski view of valid inference is concerned, we need not bother about (3),13 because, as will appear form the next section, not even implication (2) seems possible to support when one assumes that view and (4). 4. Some regress arguments There are several regress arguments that seem to show that we would never be able to justify beliefs by making inferences if we had first to establish the validity of the inferences. The best known regress of this kind is the one arising in the tale told by Lewis Carroll (1985) about Achilles and the Tortoise14, but a regress of the same structure was already described by Bernard Bolzano15. Noting that when we infer a sentence B from a sentence A, the validity of that inference, call it J, is a prerequisite if A is to be a ground for B, Bolzano asks whether besides A, we should not count the validity of J as an additional ground for B in order to get the complete ground for B. To do so, Bolzano argues, is to say that B is true because A is true and J is valid, which is really to replace the original inference of B from A by a new inference to the truth of B from two premisses: (a) A is true and (b) if A is true, then B is true. By the same reasoning, Bolzano continues, we should now say that to get the complete ground for B, we should also count the validity of the new inference to the truth of B from (a) and (b) as part of the grounds for B, thereby giving us yet another inference with three premisses: besides (a) and (b), we have now also the premiss (c)

13

if (a) A is true and (b) if A is true, then B is true, then B is true.

and the connected problems mentioned in footnote 12. Carroll (1895). 15 Bolzano (1837). My attention to this was drawn by Sundholm (2000). 14

191

The regress that this reasoning leads to is now obvious. Carroll’s regress arises instead from the question that was put in the introduction of this paper: why are we compelled to hold true the conclusion of a valid inference whose premisses we accept? In Carroll’s tale, construed as a race between Achilles and the Tortoise, the Tortoise has agreed to the truth of a sentence A from which a sentence B immediately follows (actually being one of the first steps in Euclid), and Achilles is given the task of compelling him to accept the conclusion B as well. Achilles starts by asking the Tortoise if he is not willing to accept that if A is true then B must be true. The Tortoise does not object to that. He thus accepts not only that A is true but also that if A is true then B must be true, but still does not see why he must then accept B. Achilles argues that B follows logically from what the Tortoise has already accepted, and that therefore the implication (c) above holds (with “must” inserted). The Tortoise accepts this implication too, still not seeing why he must accept B, and all that Achilles has to offer him as argument for doing so is again that if all the premisses, now (a) - (c), are true, then B must be true. And so the regress continues almost exactly as in Bolzano, the only differences being that Carroll strengthens “B is true” to “B must be true” and that the two regresses start from different questions. By slightly changing Carroll’s story, letting Achilles’ task be instead to show why the Tortoise is justified in holding B true, we can make it directly relevant to the issue with which we are now concerned: the regress will now question that we can establish the truth of the implication (2) above. Assume that a person, call her P, knows a sentence A to be true and let J be an inference from A to a sentence B. Assume further not only that J is valid but that P knows this, as required in (4), in order that the inference J is to justify her in believing that B. Why should P now be justified in holding B true? Suppose that we argue as follows: since P knows that J is valid, P knows that if A is true then B is true (an implication that Achilles gets P to accept), and therefore P may just apply modus ponens and conclude that B is true. But to be an argument showing that P is justified in holding B true, it must also be assumed because of (4) that P knows modus ponens to be a valid inference. Assuming that P has that knowledge, we infer in the same way as before that P knows (a), (b), and (c) as stated

192 above in Bolzano’s and Carroll’s regress, and given all this, we may try to argue that P is justified in holding B true. But, clearly, we have got involved in very same regress as Bolzano and Carroll. To stop this regress we must be able to infer directly that P is justified in holding B true given that P knows the inference J to be valid and A to be true. In other words, it must be right to say that P is entitled to infer B from A, and is thus justified in holding B true, just because P knows J to be valid and is justified in holding A true, without having to assume the additional premiss that P knows the validity of the inference from the two premisses (a) and (b) to the conclusion that B is true. But with what right can we stop the regress in that way? Presumably that must depend on what it means that the inference J is valid. If the validity of the inference J only means that the implication “if A, then B” is true for all variations of the non-logical terms in A and B, then it is difficult to see that knowing J to be valid supplies one with some information relevant for being justified in holding B true, except just the information that if A is true then B is true, which as we have seen does not stop the regress. As a lesson from the Bolzano-Carroll regress16 we have to conclude, it seems, that if an inference must be known to be valid in order to justify one’s belief in the conclusion, then the implication (2) cannot be derived from the validity of the inference taken in the sense of Bolzano-Tarski. The failure to support (2) does not depend on the assumed fact that the Bolzano-Tarski notion of logical consequence lacks a genuine modal ingredient,17 because the same regress seems to arise if we take the validity 16

The conclusion that Bolzano draws from the regress is that his initial question must be answered negatively, that is, the validity of the inference is not to be counted as a part of the ground for the conclusion. But this is just a matter of terminology, which leads to a distinction between grounds and justifications, contrary to the terminology that I declared in the introduction to this paper. The validity of the inference remains a condition (a necessary but not a sufficient one if (4) is assumed, as we have seen) to be justified in holding the conclusion true. As Bolzano puts it, the validity of the inference is a prerequisite for the premiss to be a ground for the conclusion. Carroll does not draw any conclusion from the regress but leaves it as a riddle why the Tortoise is compelled to accept the conclusion of an inference whose premisses and whose validity he is willing to accept. 17 The view of Etchemendy (ibid) seems to be that the failure does depend on such an absence.

193 of the inference to mean that the conclusion is a necessary consequence of the premisses; at least, is not easy to see what kind of necessity would help the situation. As we saw, in Carroll’s regress, it is assumed that the Tortoise accepts, not simply that ‘if A is true, then B is true’, but that ‘if A is true, then B must be true’, without it being possible to see that this stronger assumption gives rise to some more information relevant to the question of being justified in holding B true. To avoid misunderstanding, I should repeat that implication (2) certainly holds from a pre-theoretical stance. Any reasonable explication of what it is to be justified in holding a sentence true and what it is for an inference to be valid should support implication (2), I think. If we take the validity of an inference in the Bolzano-Tarski sense, then the implication (2) should still hold on a reasonable understanding of what it is to be justified in holding the conclusion B true. What has been shown so far is only that a regress arises when we combine the Bolzano-Tarski notion of validity with the idea expressed in (4), that knowledge of the validity of an inference is not only a sufficient but also a necessary condition of being able to use an inference to justify a belief. This is of no immediate concern for the Bolzano-Tarski notion of validity since one does not need to subscribe to (4), and should not do that, I think. However, in the next section it will be seen that the same regress lurks even when we impose a less stringent condition than knowing the inference to be valid, as long as we clinch to the Bolzano-Tarski notion of validity and do not make the condition so weak that the problem encountered in connection with implication (1) reappears. There are other objections, not connected with any particular view on the validity of inferences, against the requirement expressed in (4) that one has to know the validity of an inference if one is to be justified by the inference in holding the conclusion true. Such a requirement seems directly to give rise to a circle or a new regress, perhaps a more straightforward one than the regress noted by Bolzano and Carroll. It appears as soon as we assume that the knowledge of the validity of the inference has to be explicit and ask how we know the validity. If the validity is not immediately evident, it must come about by a demonstration, whose inferences must again be known to be valid according to the requirement expressed in (4). These inferences must either be of the same kind as the inference whose validity we try to demonstrate or be of another kind, and so we get either into a cir-

194 cle or a regress. It hardly seems reasonable to think that this circle or regress can be avoided by saying that, for sufficiently many inferences, their validity is immediately evident. On the contrary, one must expect that generally the validity has to be shown by arguments, perhaps short ones but requiring some inference steps, based on some definition or explication of what it is for an inference to be valid.18 As a matter of fact, we normally do not demonstrate the validity of an inference before we use it, and, as has been argued here, a requirement to the effect that we should do so in order that the inference is really to justify us in holding the conclusion true would block the possibility of acquiring knowledge by inference.

5. An abstract scheme for how valid inferences justify their conclusions In the light of the discussion above, the idea that the validity of an inference has to be known for it to justify its conclusion may be looked upon as an overreaction to the problem encountered in connection with implication (1). In order that a valid inference is to justify us in holding its conclusion true, it is certainly not enough that there just happens to be such an infer18

The unavoidable circularity in any attempt to justify deductive principles has been discussed by many people. Michael Dummett uses the term pragmatic circularity to emphasize that it is not a question of establishing the correctness of a deductive principle by assuming what is to be proved, which would be a gross circularity, but only of using the same deductive principle in the demonstration of its correctness; see for instance Dummett (1991), chapter 9, which also contains a response to my discussion of this kind of circularity in Prawitz (1985). Carlo Cellucci (2006) has recently discussed this kind of issue and argues that deductive inference cannot be justified, invoking among other things Carroll's regress. As seen from the above, I place that regress in a different context, but agree with Cellucci and many others that an attempt to give a suasive argument for the validity of basic deductive principles is doomed to contain a pragmatic circularity in the end. My point here is however the quite different one that we should not demand a justification of an inference before it is legitimately used to justify its conclusion (contrary to what is claimed in (4)). (It follows, as will be argued in the last section 7, that the same form of inference may be used, on the meta-level, to justify a belief in the validity of the inference.)

195 ence of which we are not aware, but it is enough, one may suggest, that we actively make the inference, in other words, that we infer the conclusion from premisses known to be true. Whether this is a reasonable way out depends on what it is to make an inference or infer a conclusion. A simple inference from a sentence A to a sentence B is commonly announced by saying something like “B because of A” or “A, hence B”. If all that is meant by making an inference is that the conclusion is asserted or believed on the ground of the premisses or because of the premisses, as expressed in such announcements, then it is not enough “to make a valid inference” in order to be justified in holding the conclusion true. In the discussion of implication (1) at the beginning of section 3, we have already refuted that idea. A gap in a proof may be located precisely to an inference in that sense, where a sentence B is asserted on the ground of an already established premiss A; although, unknown to us, the inference from A to B may be valid, the proof has to be withdrawn because of a discovered gap, which means that the assertion of B has not been given sufficient grounds or, in other words, is not considered to have been justified. Therefore, in order to be justified in holding a sentence true, one must fulfil something more than the condition of having asserted the sentence as the conclusion of a valid inference from premisses known to be true. What could that stronger requirement be? Could there be something like recognizing the validity of an inference, understood as less demanding than knowing but as something of sufficient substance to imply that one is justified in holding the conclusion true? Put more formally, we need to find a relation R between a person P and an inference J in terms of which we can state a condition that satisfies the following demands. On one hand, it has to be substantial enough so that, unlike the antecedent of the implication (1), it implies that the person is justified in holding the conclusion of the inference to be true; in other words, the following implication should hold: (5) If an inference J from a sentence A to a sentence B is valid, then a person who is justified in holding A true and stands in the relation R to J, is justified in holding B true

196 But, on the other hand, it is not to be so strong that, like the antecedent of implication (2), it cannot be satisfied when taken as a necessary condition for an inference to justify a belief. It must be possible to state a condition that is both necessary and sufficient for a person to be justified in holding the conclusion of an inference true. We should therefore be able to describe such a relation R and derive the implication (5) from what it means for an inference to be valid. As long as the validity of an inference is equated with the conclusion being a logical consequence of the premisses in the sense of Bolzano and Tarski, the prospects of describing such a relation seem slim. In particular, it seems possible to run again the Bolzano-Carroll regress for a relation R of the required kind. We must ask why a person should be justified in holding a sentence B true because of being justified in holding a sentence A true and standing in the relation R to a valid inference J from A to B. Let us go back to the less formal expression “P recognizes the validity of J” instead of “P stands in the relation R to J”, and see what we can infer from the assumption that P recognizes the validity of J. When “valid” means that the implication “if A, then B” is true for all variations of the content of the nonlogical terms of A and B, then that is what the person recognizes, and what seems relevant here is just that she recognizes the truth of this implication (without any variation of the content). From this we want to conclude that she is justified in holding B true. Trying to infer that by using (5), we may assume that she recognizes the validity of modus ponens, which implies that she recognizes the implication (c) in section 5. But there we are again with a type of argument that leads to a regress of the Bolzano-Carroll kind. At this point it may be relevant to remark that we must beware of requiring of a person that she demonstrate that something is a justification of a belief in order to be justified in holding the belief. It would be incoherent to require that a person justify that something is a justification before it is counted as a justification. In the discussion above I have made no such requirement; on the contrary I have criticized the idea that one has to know that an inference is valid before it can be used to justify a belief. But in accounting for our deductive practice philosophically, we must be able to explain how valid inferences can yield justifications of beliefs. In this account we must use what it is for an inference to be valid and should be able to derive from that, on the meta-level so to say, that the conclusion of a

197 valid inference is justified when appropriate conditions are satisfied. It has been argued here that it does not seem possible to give such account starting from the Bolzano-Tarski notion of valid inference. 6. Inferences as operations on grounds To get a fresh approach to the concept of valid inference we should reconsider the concept of inference. As just remarked, a typical way of announcing an inference is to make an assertion and state at the same time a ground for the assertion, saying for instance "B, because A" or "A, hence B". But there are examples of a more complicated kind. For instance, the premiss A need not always be asserted categorically, but may have been asserted only conditionally, that is, under some assumptions, in which case the conclusion may be asserted under the same or fewer assumptions. An example: Having arrived at a contradiction under an assumption A, we conclude that the assumption is false, saying “hence, by reductio not-A”. In that case the conclusion of the inference is asserted categorically or, at least, not any longer under the assumption A, which has been discharged. In the last example, the assertion of the conclusion was accompanied by an indication of some kind of operation that is taken to bring about a justification. This is consonant with an intuitive understanding of an inference as consisting of something more than just stating a conclusion and some premisses. Although the conclusion and the premisses may be all that we make explicit, there is also some kind of operation involved thanks to which we see that the conclusion is true given that the premisses are. Sometimes we vaguely refer to such an operation as in the example above, but it is mostly left implicit. My suggestion is that in analysing the validity of inferences, we should make these operations explicit, and regard an inference as an act by which we acquire a justification or ground for the conclusion by somehow operating on the already available grounds for the premisses. Here I can only quite briefly and roughly outline such an alternative approach.19 Its main idea is thus to take an inference as given not only by 19

My own first work on such an approach is Prawitz (1970). A similar but partly different approach is followed by Martin-Löf in his type-theory; for an application in a context of the present kind, see for instance Sundholm (1998b).

198 its premisses and conclusion (and discharged assumptions if there are such) but also by an operation defined for grounds for the premisses. As before I use here the term ground for a sentence to denote what a person needs to be in possession of in order to be justified in holding the sentence true; hence, the premisses from which a conclusion is inferred do not constitute grounds for the conclusion in the sense I use the term20 – rather the premisses have their grounds, and it is by operating on them that we get a ground for the conclusion. Seen as an act, an individual inference is individuated by some premisses and their grounds, a conclusion, an operation performed on the grounds for the premisses, a person who performs the operation, and a time or situation in which it is performed. The act consists in applying the operation to the given grounds for the premisses, claiming that the result obtained is a ground for the conclusion in view of the given grounds for the premisses. Of course, in logic we usually abstract from the person who performs the operation and the time at which it is performed. We may also abstract from the particular premisses, their grounds, and the conclusion, leaving only the operation and the relation that is to hold between the premisses and the conclusion. We may then speak of an inference form. Modus ponens associated with an appropriate operation is an example. Finally, we may also abstract from the operation left in an inference form, and may then speak of an inference figure or schema. How the validity of an inference is to be seen now suggests itself: An individual inference is valid if and only if the given grounds for the premisses are grounds for them and the result of applying the given operation to these grounds is in fact a ground for the conclusion. 20

Their truth (or a justified belief in their truth) alone does not justify a belief in the conclusion. But those who nevertheless prefer to use the term ground for the truth of the premisses may substitute “justification” when I speak of ground. As for Bolzano’s use of the term ground compare footnote 16. Since the premisses and the conclusion of an inference may be held true only under some assumptions, we need to be able to speak of what we may call unsaturated grounds that become grounds when supplied with grounds for these assumptions.

199

An inference form is valid if and only if all its instances are valid, which is also to say that when the operation in question is applied to grounds for sentences that may occur as premisses of the inference, it yields as value a ground for the corresponding conclusion. An inference scheme is valid if and only if it can be associated with an operation so as to get a valid inference form. Although there is no room here for a precise development of these ideas, the general idea can be illustrated by some examples. Let us consider the inference form of mathematical induction, in which it is concluded that a sentence A(n) holds for an arbitrary natural number n, having established the induction base that A(0) holds and the induction step that A holds for the successor n' of any natural number n given that A holds for n. The ground for the induction step may be thought of as a chain of operations that results in a ground for A(n') when applied to a ground for A(n). The operation that is involved in this inference form may roughly be described as the operation which, for any given n, takes the given ground for A(0) and then successively applies the chain of operations given as ground for the induction step n times. Obviously, given that the arguments to which this operation is applied are grounds for the premisses, i.e. the induction base and the induction step, the result of applying this operation is a ground for A(n), for any n, and hence the inference form is valid in the sense defined. This way of looking at the validity of mathematical induction accords quite well with the way in which this inference form is often explained to a beginning student, saying for instance that A(n) clearly holds for any n, since having established the induction base, we know that it holds for 0, and then we can apply the proof of the induction step as many times as needed. That the operation associated with the inference form of mathematical induction yields a ground for the conclusion is immediately clear, given that the grounds to which the operation is applied are of the stated kinds. But in some cases the validity of an inference must depend on what we take as a ground for the conclusion. Consider the simple example of conjunction introduction – the premisses are here two arbitrary sentences A

200 and B, and the corresponding conclusion has the form A & B. The operation, which we may call &I, brings together given grounds for A and B, say g and h. To carry out this inference is to apply &I to g and h and to claim that the result &I(g,h) is a ground for A & B. How do we see that conjunction introduction so described is valid? Somewhere the claim that something is a ground for a sentence must rest on what the sentence means. That we have a ground for a conjunction A & B when we bring together a ground for A and a ground for B is an example of this: we have to take it for granted because of what conjunction means. To give an account of valid inferences for a specific language, one must therefore also specify for sentences of various forms that certain things count as grounds in virtue of what the sentence in question means. When inferences are understood in the way suggested, it is immediate that a person who makes a valid inference obtains a justification for holding the conclusion true, given that she has grounds for holding the premisses true. To make an inference is to apply a certain operation to the given grounds for the premisses, and that the inference is valid is now defined just to mean that the result of applying this operation to grounds for the premisses is a ground for the conclusion, and hence it justifies the person in question in holding the conclusion true. We have thus found a relation R between a person and an inference of the kind sought for in section 5: it consists simply in the person making the inference. But to make an inference has now been given a meaning more substantial than to claim only that the conclusion is true because of the truth of the premisses. It now also means that a specific operation is applied to given grounds for the premisses. The carrying out of this operation may not result in a ground for the conclusion, the inference then being invalid, of course. But if the inference is valid, then by the definition of what that means, the result is a ground for the conclusion. Therefore, as already said, when a person makes a valid inference, she gets in possession of a ground for the conclusion, and we must thus grant that she is justified in holding the conclusion true. What it is to apply an operation to some arguments is something fairly well known from other fields, for instance arithmetic, but should anyway be discussed more thoroughly than I can do here; it is best done in parallel with a more systematic development of the ideas concerning

201 grounds for sentences of various forms. But let us consider one further example, beyond the ones already given. The inference form conjunction elimination exists in two forms: the premiss is always a conjunction, say A & B, and the conclusion is then either A or B. Corresponding to these two forms of conjunction elimination, we have two operations, call them &E1 and &E2. They are applicable to all grounds for conjunctions, and the value of applying the operations to such a ground is given by the equations &E1(&I(p,q)) = p and &E2(&I(p,q)) = q, respectively. That a person applies one of these operations implies that she is able to handle the equation in question. By the meaning of conjunction, g is a ground for A & B, if and only if g = &I(p,q) for some p and q such that p is a ground for A and q is a ground for B. Applying for instance &E2 to a ground &I(p,q) for A & B thus yields q, which by assumption is a ground for B, as required for this inference form to be valid. As this should illustrate, to make a valid inference is to apply a specific operation to given grounds and to evaluate the result, and is not to show that the inference is valid. But by simple reflection on what one has done, one realizes that one is in possession of a ground for the conclusion, and by some further reflection that the inference is valid. Knowledge of the validity of an inference has not been made a prerequisite for using the inference to justify a belief, but, as seen, such knowledge is easily obtained by reflecting upon the inference act. When making such a reflection explicit, it will involve a number of inferences, and typically among them will be an inference of the very kind that we are reflecting on. But because of the fact that a valid inference justifies a belief in the conclusion without it being known that the inference is valid, there is no vicious circle involved here. 7. The compelling force of an inference Let me finish by discussing briefly, presupposing the view of inferences sketched here, what kind of necessity is involved in connection with a valid inference. My suggestion is that this necessity has to do with the compelling force of a valid inference, which, as remarked in the introduction, is

202 a feature of a valid inference. Can we account for this force of an inference from the present perspective? To connect again with Carroll’s story about Achilles and Tortoise, but now in its original formulation, where Achilles is given the task of making the Tortoise compelled (or with Carroll’s words, “to force [the Tortoise], logically”) to accept as true a sentence B that obviously follows from a sentence A that the Tortoise has accepted, we recall that Achilles’ strategy was quite unsuccessful: all that Achilles succeeds in doing is to get the Tortoise to accept a series of additional premisses in the form of implications, starting with “if A is true, then B must be true”. The new sense that we have given to what it is for an inference to be valid does not change the futility of this strategy: it is still pointless to invoke the validity of the inference from A to B, because by just accepting the validity, the Tortoise would again be sitting with a number of additional premisses, which in itself leads nowhere. The right strategy must be instead to ask the Tortoise to act, for instance: “Please, infer B from A; you know how to do it!” And then: "Look at what you have got! You see that it is a ground for B, don’t you!” I think that the strategy suggested is a good illustration of how one may get compelled to form a belief. The first step is to act and make an inference. If the inference is valid and one has grounds for the premisses, then the inference act does in fact result in a ground for the conclusion. One may say that this fact does not yet have any compelling force. But we just saw in the case of conjunction elimination how, knowing the meaning of the sentences involved and reflecting upon the inference that one has made, one easily realizes that one has got in the possession of a ground for the conclusion. At least at this point, when we have become aware of having a ground for the conclusion, it would be irrational not to hold that sentence true; by pain of being irrational, we have to hold the sentence true, and are in that sense compelled. As I have argued elsewhere,21 this is a kind of necessity that can suitably be called necessity of thought – it refers to how one should or must think. It is clearly not an ontic necessity – there is no reference to all possible worlds – and the necessity has not to do with variations of the content 21

Prawitz (2005).

203 of non-logical terms. The necessity is not particularly connected with logically valid inferences; simply valid inferences are equally compelling. This kind of necessity makes it quite appropriate to place the word “must” in front of the conclusion.22 However, when placed in that position, saying “hence B must be true” should be understood to say, not that it must be the case that B is true, but rather that we must hold B true.

References Bolzano, B. 1837, Wissenschaftslehre I-IV, Seidel, Sulzbach. Carroll, L. 1895, “What the Tortoise said to Achilles”, Mind IV, pp 278-280. Cellucci, C. 2006, "The Question Hume Didn't Ask", in Demonstrative and NonDemonstrative Reasoning in Mathematics and Natural Science, (eds) C. Cellucci and P. Pecere, Edizione dell'Università degli Studi di Cassino, Cassino, pp 207-235. Cozzo, C. 2005, "Can a Proof Compel Us?", in Mathematical Reasoning and Heuristics, (eds) C. Cellucci and D. Gillies, King's College Publications, London, pp. 191-211. Dummett, M. 1991, The Logical Basis of Metaphysics, Duckworth, London. Etchemendy, J. 1990, The Concept of Logical Consequence, Harvard University Press, Cambridge, MA. Martin-Löf, P. 1985, “On the Meaning of the Logical Constants and the Justification of the Logical Laws”, in Atti degli Logica Matematica 2, Scuola de Specializzazione in Logica Matematica, Dipartimento de Matematica, Università di Sie22

I think that even non-deductive inference may have a compelling force. If overwhelming evidence points to the truth of a sentence, it would be irrational not to hold the sentence true, no matter whether the evidence is conclusive or not. As linguists have observed, in several indo-european languages “must” or its synonym is often inserted when an assertion is not based on observations but is reached by inferences. Admittedly, the compelling force may be quite weak in such cases, and inserting “must” then comes to indicate, not so much that the grounds are compelling, but rather that they are indirect; see for instance Prawitz (1994) or (1977).

204 na, Siena, pp 203-281. (Republished in Nordic Journal of Philosophical Logic, 1, pp 11-60.) Prawitz, D. 1970, “Constructive semantics”, in Proceedings of the 1st Scandinavian Logic Symposium Åbo 1968, Filosofiska Föreningen och Filosofiska Institutionen vid Uppsala Universitet, Uppsala, pp 96-114. — 1977, “Logisk intuitionism, sanning og mening”, Norsk Filosofisk Tidskrift -77 nr 3, pp 139-172. — 1985, “Remarks on some Approaches to the Concept of Logical Consequence”, Synthese 62, pp 153-171. — 1994, “Meaning and Experience”, Synthese 98, pp 131-141. — 2005, “Logical Consequence from a Constructivist Point of View”, in The Oxford Handbook of Philosophy of Mathematics and Logic, (ed) S. Shapiro, Oxford University Press, Oxford, pp 671-695. Ross, W. D. 1949, Aristotle’s Prior and Posterior Analytics, Oxford University Press, Oxford. Sundholm, G 1998a, “Inference Versus Consequence”, The LOGICA Yearbook 1997, Czech Acad. Sc., Prague. — 1998b, “Proofs as Acts and Proofs as Objects: Some questions for Dag Prawitz”, Theoria LXIV, pp 187-216. — 2000, “When, and why, did Frege read Bolzano?”, LOGICA Yearbook 1999, Filosofia Publishers, Czech Academy of Science, Prague, pp 164-174. Tarski, A. 1936, “Über den Begriff der logischen Folgerung” in Actes du Congrès International de Philosophie Scientifiques 7, Hermann et Cie, Paris, pp 1-11. — 1956, Logic, Semantics and Metamathematics, Oxford University Press, Oxford.

Reason and Rationality‡ Jon Elster*

Abstract Some years ago, Dagfinn Føllesdal and I had an exchange in Oslo over the nature of rationality. We disagreed, but at the time the source of our disagreement was not obvious to me. Since then I’ve thought some more about the matter, and I think I know why we disagreed. Føllesdal said, if I remember him correctly, that he opposed some of my conclusions because they went against his intuition that “rationality” is what we call in Norwegian a “honnør-ord”, which means something like “a positively laden term”. Now, whereas I do believe that reason is such a word, I do not think rationality is or at least not in the same sense. Like courage or perseverance, rationality is a neutral virtue in the sense that it can be harnessed to any end, good or bad. In this paper, I shall try to clarify this distinction. Whether the argument will satisfy Føllesdal remains to be seen.

In the analytical approach to human behavior, the same Latin word, “ratio”, is at the origin of two very different traditions in Western thought, which are both quite different and linked to each other in several ways. On the one hand, there is a tradition that opposes reason to passion and, more recently, to interest. Seneca’s treatise On Anger, for instance, is organized around the opposition between reason and passion, to which the French moralists of the 17th century added the category of interest. An influential statement of the relation among these three motivations is that of La Bruyère: “Nothing is easier for passion than to overcome reason; its great triumph is to conquer interest.”1 As I shall argue, reason in this sense is closely linked to the ideas of justice, of the public interest and of the common good. ‡

This article draws on my inaugural lesson in 2006 at the Collège de France, “Raison et raisons”. * Columbia University 1 La Bruyère (2007), p. 98.

206 On the other hand there is the more recent notion, which emerged some time in the last third of the nineteenth century, of rational choice as opposed to various forms of irrationality. A rational agent is one who acts on sufficient reasons (in the plural); more specifically, who acts on the beliefs and desires that rationalize the action. This idea is spelled out more fully later. The idea of rationality is often, but misleadingly, linked to the exclusive pursuit of self-interest.2 Scholars are of course free to define their concepts any way they want. Yet it would be strange if one were not allowed to say that a person who tries to alleviate poverty in the world should do so in a rational manner, by making sure that his donations do not line the pockets of dictators or bloat the bureaucracy of the charitable organization. To act in conformity with reason, in the singular, and to act for good reasons, in the plural, differ in that reason is objective whereas the reasons are subjective. From the external point of view, in Bernard Williams’s sense, one can assess an action as more or less in agreement with reason.3 From the internal point of view, one can assess it as being rational or irrational. Of these two ideas, only rationality can enter into the explanation of behavior. The demands of reason can generate and explain specific actions only to the extent that the agent internalizes them. Although different, the two norms have a common obstacle, namely the passions. Although there have been several attempts in recent years to argue that emotion can make our choices or more rational or can help us attain the goals of reason, I do not believe they have succeeded. The observation made by Antonio Damasio that people with emotional deficiencies are also deficient in their decision-making may be due to non-causal correlation rather than to causation. 4 Although one might think that a passionate commitment to justice would promote the attainment of that end5, the cognitive distortions induced by passion might undermine it. The work and life of Karl Marx may serve as an example. His passionate commitment to the liberation of the working class subverted the clear-headed thinking needed to attain that end. 2

See for instance Aumann (2006). Williams (1981). 4 Damasio (1994); for comments see Elster (1999), pp. 291-98. 5 For an argument to that effect see Rudenfeld (2001). 3

207 The norm of rationality and the norm of reason also have something in common, namely the requirement of acting on beliefs that are well grounded in the evidence. Finally, the two norms have in common that they are the object of a certain deference on the part of the agent. The origin and the nature of the deference differ, but in both cases it is a matter of deferring to a source of normativity. As we shall see, the effect of the mechanisms of deference is sometimes to subvert the object of the deference. Whereas the theory of rational choice has been the object of highly elaborate developments, this is not the case for the idea of reason. The conception I shall present is a synthesis, which I hope is not too idiosyncratic, of various classical texts. Let me start with another observation by La Bruyère: “To think only about oneself and about the present, is a source of mistake in politics.”6 A century later, James Madison referred to “the mild voice of reason, pleading the cause of an enlarged [i.e. not selfish] and permanent [i.e. not shortsighted] interest”.7 Instead of the partial perspectives of selfishness and myopia, one has to substitute a doubly impartial attitude. The idea that reason requires an impartial treatment of individuals has been known at least since Leibniz, who said that one should “put oneself in the place of all”8, and probably goes back much further. A fact that will turn out to be important for my purposes is that impartiality is a very general idea that can be spelled out in many different ways. The choice behind the veil of ignorance, which is usually thought of as a way of representing the idea of impartiality, can be understood in at least four distinct ways, corresponding to Harsanyi, Rawls, Dworkin and various meritocratic conceptions, which are philosophically less central but politically quite important. Conceptions of reciprocity and of universal human rights are also instances of this general notion of impartiality. Reason also requires an impartial treatment of temporal instants. No date can claim privileged treatment simply because of its location on the temporal continuum, e.g. because of its proximity to the present moment. 6

La Bruyère (2007), p. 358. Federalist # 42. 8 Leibniz (2002), p. 124. 7

208 To prefer an early small reward to a larger delayed reward is not in itself contrary to reason. I might reasonably prefer a hundred dollars now over five hundred in a year if my life expectancy is less than one year or if I need the hundred dollars for survival. But a preference that is due only to the temporal proximity is contrary to reason. From the objective point of view, or an observer’s point of view, individuals who ignore the remote consequences of their present behavior are acting stupidly, because their lives are likely to be short and unhappy. I shall argue, however, that we are not entitled to infer that they are also acting irrationally. A third component of reason, as I understand the idea, is that of acting on beliefs that are well-grounded in the agent’s evidence. Exactly what this criterion means is hard to spell out, and I shall not try to do so. At the very least it excludes systematic bias in belief formation, whether hot or cold, motivated or unmotivated. Rational choice, as we shall see, has a similar requirement. We should not assume, however, that a rational agent would form the same beliefs as those required by reason. Although the normative links between evidence and belief are the same in both cases, the normative requirements for how much evidence to gather are not the same, as we shall see shortly. Let me move on to the much better-understood idea of rational choice, summarized in Figure 1 below. In the diagram, the arrows represent both optimality relations and causal relations. Thus for instance, a belief is rational if it is caused by the evidence in light of which it is optimal rather than, for instance, being the result of two opposite biases that exactly cancel each other.

209

The blocked arrow is intended to exclude any direct impact of desires on beliefs, as occurs in wishful thinking. As we shall see shortly, however, there may be an indirect impact that is mediated by the process of evidence-gathering. The optimal amount of evidence is determined partly by the agent’s prior beliefs about the expected costs and benefits of gathering it. Sometimes, the beliefs about the benefits are modified by what is found in the search itself, as shown by the loop in the diagram. At the same time, how much information it is rational to gather also depends on the desires of the agent. A selfish person is not likely to invest much in information about how the action is going to affect other people. Similarly, a myopic person is not likely to invest much in information about the long-term consequences of his choice. If reason requires us to think about others and about the remote future, it may therefore induce more investment in informationgathering and hence different beliefs than mere rationality requires, except, as I said, if the rational person has internalized the demands of reason. I want to emphasize the radically subjective character of the notion of rationality, as I understand it. The rational person is one who does as well as possible by his or her own lights, whatever these may be. He or she

210 gathers the optimal amount of information, forms accurate beliefs on that basis, and then chooses the action that will best realize his or her desire, given those beliefs. The observer may then try to explain the behavior on the hypothesis that each of these three optimizing operations has been carried out correctly. In this perspective, there is no room for assessing the lights themselves. The desires are the unmoved movers of choice, which are not themselves chosen. At least, this is true for a rational agent. If an agent desires to change his desires, it must reflect some form of irrationality. That statement refers to fundamental desires, not to induced ones. The desire to do X as a means to achieve Y may of course change if the agent finds that Z is a better means to Y. Yet Y remains the unmoved mover throughout. I shall illustrate these claims by a hypothetical example. Suppose I suffer from a severe inability to defer gratification, that is, from being unable to take account of future consequences of present behavior. And suppose scientists came up with a discounting pill, which would increase the weight of future rewards in present decisions. You may, if you wish, substitute psychotherapy for the pill. If I take the pill, my life will go better. My parents will be happy I took the pill. In retrospect, I will be grateful that I did. But if I have a choice to take the pill or not, I will refuse if I am rational. Any behavior that the pill would induce is already within my reach. I could stop smoking, start exercising or start saving right now, but I don’t. Since I do not want to do it, I would not want to take a pill that made me do it. Another way of putting the matter is to say that a person who wants to be motivated by long-term consequences ipso fact has that motivation already.9 That is why, with an exception to be noted shortly, the idea of meta-preferences seems pointless. In this example I have tacitly assumed that the agent discounts the future in a consistent manner. By this I mean the following. If at one point in time the present value of one future reward is larger than the present value of another future reward, then its present value is larger when seen from any point of time. It can never happen, that is, that as we move on in time, the preferred reward becomes the less preferred for no other reason than the passage of time. This seems to be a requirement of rationality. 9

See also Skog (2001).

211

One form of time discounting with this consistency property is exponential discounting, shown in the upper panel of Figure 2. The curves representing the exponentially discounted values of two future rewards can never cross each other. Until recently, economists widely assumed that exponential discounting is a good description of how people actually make intertemporal choices. An increasing amount of empirical evidence suggests, however, that people discount the future hyperbolically, as in the lower panel of Figure 2.10 The present value of one unit of good (utility or welfare) t periods into the future is 1/(1+kt), where k is a positive constant. With hyperbolic discounting, the present-value curves can cross. Suppose that on March 1 I wake up with a toothache and make an appointment with my dentist for March 15. On March 14 I call him to say that because of a (fictitious) funeral in the family I cannot keep it. Except for the sheer passage of time, no change has occurred in the interval. In particular, the pain from toothache is the same. Once this has happened to me a few times, I may understand what is going on and take steps to prevent it from happening again. I might, for instance, act on the external reward system by authorizing my dentist to bill me three times the normal amount if I fail to show up for an appointment. But I might also try to change my time preferences. Suppose I have the 10

See notably Loewenstein, Read and Baumeister, eds. (2003).

212 choice between a reward of 10 at time 5 and a reward of 30 at time 10. The table below gives the present values of these rewards at all times up to time 5. Approximate present values (k = 1) Time

0

1

2

3

4

5

Reward 10 at time 5

1.7

2

2.5

3.3

5

10

Reward 30 at time 10

2.7

3

3.3

3.8

4.3

5

We observe that a crossover occurs some time between time 3 and time 4 – that is when I call my dentist to cancel. Suppose now that I take a pill, or go into therapy, for the purpose of changing the parameter k to a value of 0.3. Approximate present values (k = 0.3) Time

0

1

Reward 10 at time 5

4

4.5

Reward 30 at time 10

7.5

8.1

2

3

4

5

5

6.3

7.7

10

8.9

9.7

10.7

12

In this case, the delayed larger reward remains preferred throughout the whole period up to the time when the small reward becomes available, and the agent will therefore be able to stick to his resolution. Even if he has to pay for the pill or the therapy up front, he is willing to do so as long as the cost is less than 2. In this case, the agent may rationally want to change his preferences, but only because they are inconsistent. Although they remain inconsistent after the change, the inconsistency does not do any damage since the agent is now able to carry out his plan of choosing the larger delayed reward. By

213 contrast, an agent who discounts the future consistently has no reason to engage in character planning, at least not for the sake of enabling him to make certain choices. He might, perhaps, choose to take a discounting pill if he values patience for its own sake, as an admirable character trait.11 Again, however, this would only be because he values it, not because it is objectively valuable. Choices, in other words, need to be seen through the eyes of the agent. A myopic person who loses his glasses may be prevented by his myopia from finding them. He is “trapped”. The person who discounts the future very much, but in a consistent manner, is also trapped. His current preferences offer him no reason to change his preferences. Similarly, a rational agent may find himself in a “belief trap” that leaves him stuck with a false belief, namely if the believed costs of testing the belief are too high. In Poland it has been widely believed that anyone who drinks when using disulfiram (Antabuse) implanted under the skin will die.12 As a matter of fact, implanted disulfiram is pharmacologically inert.13 The false belief might nevertheless (rationally) deter people from testing it. As I said at the outset, reason – in the sense of an enlightened concern for the long-term good of the community – is a positively laden term. I believe this is a fact of empirical anthropology, not of philosophical anthropology or, less grandly, of conceptual analysis. To explain what I mean by this claim, let me first note that in all societies we observe a normative hierarchy of motivations, in the sense that people may be praised or blamed for acting on a certain motivation independently of praise or blame for the action itself. In ancient Greece, for instance, patriotism on behalf of the polis was at the top of hierarchy; next came vengeance for a personal affront; further down we find material self-interest; and at the bottom of the hierarchy are envy and hubris.14 Tocqueville claimed that in the America he observed around 1830, self-interest was more highly regarded than altruism.15 In some societies, the desire for personal vengeance is more 11

Skog (2001). Osiatynski (1997). 13 Johnsen and Mørland (1992). 14 For examples see Hansen (1991), p. 195; Aristotle, Politics, 1311b, 1315a. 15 Tocqueville (2004), p. 611; see also Miller (1999). 12

214 highly valued than impartial motives, whereas in modern Western societies vengeance probably ranks below self-interest. The existence of this hierarchy induces a desire to represent a given action, to oneself or to others, as performed for a motivation that is highly ranked in the hierarchy. As Seneca said, “Reason wishes the decision that it gives to be just; anger wishes to have the decision which it has given seem the just decision”.16 The latter wish is a meta-motivation, not to be confused with meta-preferences.17 An example of the latter would be a person with two different preference orderings, one for eating over dieting and one for dieting over eating, and a meta-preference favoring the latter. I fail to see how this amounts to more than a simple preference for dieting. Following La Bruyère’s insight that “Men are very vain, and of all things hate to be thought so”18, a meta-motivation could amount to a preference for preferring dieting over eating on grounds of health over having the same preference for dieting on grounds of vanity. Let me bracket the question of the audience to which the agent is representing his or her motivation. It could be another person, or it could be the agent him- or herself. In one case it would be deception, in the other self-deception. For my purposes here what matters is that in either case the agent would display a deference towards the higher-ranked motives. In societies that value reason highly, people will try to present behavior that is inspired by interest or passion as really motivated by impartial concerns. Conversely, Tocqueville claimed, the Americans he observed tried to present their spontaneous altruistic actions as really motivated by self-interest. The deference to reason, when it is observed, may cause motivational conflicts. We may be torn between our self-interest and our self-image as being someone who is not moved only by self-interest. The conflict may be resolved in one of several ways. One solution is that one of the two motives, the first-order one or the second-order one, is sacrificed to the other. Another is to strike a compromise. A third is to try to have it both ways. If I want to have a good self-image but also to keep my money for myself, I can achieve both ends by adopting an appropriate impartial principle. If 16

On Anger I. 18. Sen (1977). 18 La Bruyère (1990), p. 318. 17

215 other people donate much, I can donate little on the utilitarian ground that my contribution will not have much of an impact. If others give little, I can invoke a principle of fairness that will justify small donations on my part as well. Why should I be a moral saint? I have argued that the deference to reason is not invariant across all societies. By contrast, I believe that the deference to rationality is transculturally and transhistorically invariant. As Føllesdal emphasizes in the writings I cited in my Personal Memoir, rationality is a norm. Specifically, it is a norm in the sense that if we want to realize a desire, we ipso facto want to realize it as simply and efficiently as possible. As a Norwegian proverb says, “Don’t cross to the farther bank of the river when you go to fetch water”. (The proverbial English example of bringing coal to Newcastle is more ironic, since the American who was induced to do so by rival merchants actually made a huge profit due to a local miners’ strike.) I believe that any counterexample to this proposition would simply involve a richer specification of the desire of the agent. I said earlier that the deference to reason and to rationality might subvert their objects. Consider for instance wage bargaining. Although this is an arena where it is perfectly acceptable to appeal to mere interest, it may be strategically useful to appeal to principles such as equity, equality or desert. Since a concession on a matter of principle is weightier than a concession on a matter of interest, it requires a larger concession by the other side. The problem is that if both sides engage in such strategic uses of argument, the bargaining may break down, to the detriment of both parties. The appeal to reason can undermine reason.19 Similarly, the appeal to rationality can undermine rationality. Let me define hyperrationality as the search for the solution that would have been optimal if we ignored the cost of the search itself. For a trivial example, consider comparison shopping. People on holiday in Southern France cross the border into Spain to buy cigarettes, as if the gas were free and the time spent in the traffic jams had no opportunity costs.20 For a more substantial example, consider the use of the principle of “the best interest of the child”

19 20

Elster (1989 a), p. 170-72. L’Indépendant (Perpignan), August 13, 2005, p. 2.

216 in contested child custody cases.21 There is by now a great deal of evidence that the time-consuming and highly conflictual process of determining which parent is best suited to take care of the child is very much against the child’s interest. Instead of this hyperrational process, it would make more sense to award custody by the flip of a coin or, if this should prove unacceptable, revert to the presumption of maternal custody. To the extent that there is some truth to the claim that rationality is a specifically modern or Western idea, it applies only to this idea of hyperrationality.22 In modern Western societies, reason as I have defined it is indeed a positively laden term. It is at the top of our normative hierarchy, as it was also in ancient Greece. Rationality, by contrast, is not something we value in itself, but rather an aspect of what it means to act to realize one’s goals, whatever they might be. Even in the case of a person who lives only for the present and for himself, and who is rapidly destroying his health, finances and personal life, we may ask whether he goes about it in a rational manner. This statement may be counterintuitive, but I have tried to argue that it is not wrong.

References Aumann, R. (2006), “War and peace” (Nobel Prize lecture), in K. Grandin (ed.), Les Prix Nobel 2005, Stockholm: The Nobel Foundation. Babcock, L., Wang, X. and Loewenstein, G. (1992), “Choosing the wrong pond”, Quarterly Journal of Economics 111, 1-20. Damasio, A. (1994), Descartes’s Error, New York: Putnam Elster, J. (1989 a), The Cement of Society, Cambridge University Press Elster, J. (1989 b), Solomonic Judgments, Cambridge University Press Elster, J. (1999), Alchemies of the Mind, Cambridge University Press

21 22

Elster (1989 b), Ch. III; also Babcock, Wang and Loewenstein (1992). For a thoughtful discussion, see Wiener (1998).

217 Hansen, M. H. (1991), The Athenian Democracy in the Age of Demosthenes, Oxford: Blackwell. Johnsen, J. and Mørland, J. (1992), “Depot preparations of disulfiram: experimental and clinical results”, Acta Psychiatrica Scandinavica 86, 27-30. La Bruyère (2007), Characters, Whitefish MT: Kessinger Publishing Loewenstein, G., Read, D. and Baumeister, R. (2003), Time and Decision, New York: Russell Sage. Leibniz, G. W. (2002), Le droit de la raison, Paris: Vrin. Miller, D. T. (1999), “The norm of self-interest”, American Psychologist 54, 1053-60. Osiatynski, W. (1997), Alcoholism: Sin or Disease?, Warsaw: Stefan Batory Foundation Rudenfeld, J. (2001), Freedom and Time, New Haven CT: Yale University Press. Sen, A. (1977), “Rational fools”, Philosophy and Public Affairs 6, 317-44. Skog, O.-J. (2001), “Theorizing about patience formation: The necessity of conceptual distinctions”, Economics and Philosophy 17, 207-19 Tocqueville, A. de (2004), Democracy in America, New York: Library of America. Wiener, J. (1998), “Managing the iatrogenic risks of risk management”, Risk: Health, Safety and Environment 9, 39-82. Williams, B. (1981), “Internal and external reasons”, in his Moral Luck, Cambridge University Press, pp. 101-13.

Essays Part IV Meaning

On “Meaning and Experience” Olav Gjelsvik ∗

Introduction Dagfinn Føllesdal’s wonderful paper “Meaning and Experience” pursues a general position on meaning, and it does so from a Husserlian background while responding both to Quine’s and to Davidson’s philosophical work on the topic. The present paper aims to assess the philosophical significance of this contribution. It is natural for me to engage with Føllesdal’s position by looking at the ways it interacts and also contrasts with Quine or Davidson. I shall mainly look at the last interaction, but also a little bit at the first. The structure of this paper is as follows: I shall first introduce the problems that I shall pursue, and place Føllesdal’s points and contributions in a philosophical context. I will then supply some of my own thoughts on the same cluster of problems, and on this background evaluate his contribution.1 Føllesdal’s main critical point against Davidson’s view on radical interpretation Let me start directly with the main critical point Føllesdal makes about Davidson in 1975. According to Føllesdal 1999 he first made the point to Davidson when they were walking together in the hills in the neighbourhood of Biel in 1973. At the time Davidson’s main instruction for radical interpretation was “Maximize Agreement”. This rule needs tempering ∗

CSMN, Universitetet i Oslo I develop my thoughts about Davidson’s main project on meaning and radical interpretation in my “Knowledge and error: a new approach to radical interpretation”, in Donald Davidson on Truth, Meaning and the Mental, ed. G. Preyer, Oxford, OUP, 2012, pp.167-191. That paper contains some of the same discussions of Davidson as this chapter, but takes things much further and places them in a general systematic context.

1

222 with, Føllesdal argues. Imagine that a big tree is between the speaker and the only rabbit present, and that the interpreter sees this. In that case one should not expect agreement between speaker and interpreter about whether a rabbit is present. One should therefore not maximise agreement at all costs, one should limit agreement somewhat; at least to the cases where disagreement is not to be expected. What is present or absent to the senses of the speaker and the interpreter respectively plays an important role in the identification of the cases where disagreement is to be expected. Note that if one were engaged in radical interpretation, entertained the hypothesis that “gavagai” meant rabbit, and really maximized agreement in this situation, one would most likely throw away a correct translation/interpretation hypothesis. Davidson immediately responded to Føllesdal that radical interpretation should be based on weighted agreement. Føllesdal found this promising, but also pointed out that such a theory “does not relieve the translator and student of meaning from considering what is happening at our sensory surfaces. On the contrary, a main source of evidence for a rational person is his sensory evidence. For this reason, a special premium must be put on agreement between similarly placed observers. And this notion of similarly placed observers brings us directly back to the problem of intersubjective comparisons of stimulations we discussed with Quine.”2 Looking at the discussions of Quine in the same paper, i.e. the discussions Føllesdal refers to in the quote above, we see that Føllesdal agrees with Quine about the importance of sensory evidence, but disagrees with Quine on what sensory evidence is. Føllesdal’s main charge against Quine on the latter point is that Quine’s identification of stimulations with triggerings of nerve endings sets us off on the wrong track. It does not preserve Quine’s own insight that meaning in language is a social phenomenon. Føllesdal suggests that we instead of going inside the skin should “identify stimulations with the pattern of chromatic irradiation, sound vibration, etc., just outside, or at the sensory surfaces.”3 Føllesdal sums his position up by saying that a theory of meaning should include at least two things; both the structure of a truth-theory and a 2 3

Føllesdal (1975), p. 40. Føllesdal (1975), p. 35.

223 condition concerning weighted agreement, in which what is going on at our sensory surfaces plays an important role. In addition he points to many other things that play a role. Now, I think we should all agree that what is going on at the sensory surfaces in this sense plays an important role that a theory of meaning should recognize one way or other. But I also see a need to say quite a bit more about what exactly this role is or should be in an account of meaning and content. It looks to me as Føllesdal here in 1975 might be siding with Quine against Davidson on the issue of proximal versus distal stimuli or input, as this question has been pursued in the later debates between Davidson and Quine. My own view is that we should in a way not side with any of them, and that this is important. I hope Føllesdal is with me in this, and also with me when I pursue what role, exactly, we should give to sensory evidence. Davidson’s response to Føllesdal Let me now turn to Davidson’s later thoughts about error and input. Here is Davidson about error in his Dewey-lectures, “The Structure and Content of Truth”. He is here commenting on the difficulties for his own distal (as opposed to Quine’s proximal) input theory when explaining error. The distal inputs are out there in the environment of the speaker: The difficulty with the distal theory is that it makes error hard to explain, the crucial gap between what is believed true and what is true. ... The solution depends on two closely related interpretative devices. An interpreter bent on working out a speaker's meanings notes more than what causes assents and dissents, he notes how well placed and equipped the speaker is to observe aspects of his environment, and accordingly gives more weight to some verbal responses than to others. This provides him with the rudiments of an explanation of deviant cases where the speaker calls a sheep a goat because he is mistaken about the animal rather than the word. 4

Note how Davidson’s formulation in 1990 is quite close to Føllesdal’s position in 1975. More than assent and dissent is needed, Davidson says; one also has to look at how well placed the speaker is for observation. This is 4

See Davidson (1990), p.321.

224 the basis for the weighting. What Davidson says tallies closely with Føllesdal’s wording that “a special premium must be put on agreement between similarly placed observers”. If there is a difference between Føllesdal and Davidson in 1990, it relates to the role of causality, and whether we can get at what objects we speak about and refer to by identifying causal interactions between the speaker and the environment. This further relates to the ambitions in Davidson’s project of 1990. This project seems quite ambitious in that he says he aims to bridge the gap between the propositional and the non-propositional by starting with preferences for the truth of uninterpreted sentences in a physical environment where the radical interpreter observes causal interactions. It seems to me that Føllesdal is very clear that ambitions towards bridging this gap are misplaced. Føllesdal is furthermore very clear that just speaking of weighting will not do, one needs to say things directly about the role of evidence when explaining the principles of the weighting. I shall now turn to another side of this, and that is what remains of the issue about distal or proximal input. In this context I find Davidson’s notion of being well placed and equipped for observation interesting. It seems to me, however, that the notion we need in radical interpretation is not “being well placed and equipped for observation”. What we need is the notion of being well placed and equipped for knowing by observation, or so I shall argue below. I shall indeed argue that this change does the trick in sorting out what is going on in the discussions about distal or proximal input. I shall first explore the issue of explicating and explaining error by looking at the case Davidson suggests in the long quotation from him. The case is the case of sheep and goat, where the speaker is mistaken about the animal and not about the word. Mistakes can be of very many types. The case of the rabbit behind the tree exemplifies one type of case, where one is mistaken about whether an animal is there at all. If the present case is a case where the sheep has been dressed up as a goat, and it is almost impossible to spot the real nature of the animal just by looking, then it is a case where one is expected to be mistaken about the type of animal. Assumption now: The case of sheep and goat is of this sort. Note then that the speaker in this case, as opposed to the rabbit case, is in a plain sense very well placed and equipped for observing this animal since the animal is plainly in sight and the speaker’s sight is good. The striking feature is that

225 it is not easy at all to know by observation what animal it is. In fact it is a case where the speaker, in case the speaker possesses the concepts of sheep and goat, is strongly expected to believe that the animal is a goat. We can perhaps think of this case as a case where we cannot explain that the speaker S does not make the relevant error about the world.5 By Føllesdal’s terminology from 1975, we can say that it is a case where the surface stimulations are identical. Same surface stimulations, same beliefs and same words. Davidson’s phrase is, however, “how well placed and equipped the speaker is to observe aspects of his environment”.6 My question here is whether this phrase captures what he needs to capture. Davidson’s explanation of error should according to this phrase, I presume, work by explaining error by pointing out that the speaker is not well placed or equipped to observe in the erratic case. Note that in the case of the rabbit behind the tree, introduced by Føllesdal in 1973, we can explain the error quite well by employing Davidson’s phrase. Perhaps the phrase has been coined in response to that example. If, however, we were to explain the false belief in the present case of the sheep and goat by employing that same phrase of Davidson’s, then it will not work so easily. This is because the speaker in a plain sense is well placed to observe. There is no problem in observing the animal. The intuitive difficulty in this case is that the speaker who is well placed to observe is not well placed or equipped to 5

When things are like this, they will have specific implications for interpretation of his words if the speaker insists on calling this sheep a sheep even when he ought to call it a goat according to our tentative interpretation. A different – weaker and intermediate – case is the case where it is understandable and explainable that the speaker mistakes the sheep for a goat, but many speakers, also this one, does not always make a mistake here. Then the case will have less obvious implications for the interpretation of words if the speaker according to our tentative interpretation says that the animal in front is a sheep. There will be a variety of such cases, and the interpreter will work towards some trade-off in interpretation, look at many different cases etc. The point I will stress is only that if we epistemically expect error, and there is no reasonable explanation of why the error is not made, as in the case above with the dressed up sheep, not distinguishable from a goat, and there is no slip of the tongue or any other slip that gets corrected by further questioning, then we do have a case which has very obvious implications for attributions of meaning. 6 My italics.

226 know by observation what sort of animal it is. We therefore have a prima facie case for replacing Davidson’s phrase with the phrase “being well placed or equipped to know by observation”. If that is what we need to explicate error in cases of this sort, then there is a need to employ the concept of knowledge in explicating error. These matters deserve, however, a much closer look.7 Preparing to explicate error: Simple and not so simple attributions Of primary interest when doing radical interpretation are statements about what the speaker S sees, hears, smells, etc. Such statements provide an essential part of the story of what causes assent and dissent. I shall call such attributions simple attributions. We can think of the case where a simple attribution is true as a case where the subject is provided with evidence in the matter at hand. One first point is now this: An explanation of false beliefs works by attributing to S not only evidence in this sense of evidence, but also by attributing epistemic sensitivity to such evidence, or to the reasons for belief the evidence provides. We typically explain error by pointing to how things looked or appeared to S, rather than plainly pointing to what S saw. When we point to the latter we do that in a way that precisely indicates or exhibits how things must have looked to the speaker. A man believes falsely that he saw a bear, because he saw a man dressed up like a bear. We might just say that he saw a man that looked exactly like a bear and leave it at that. Then we assume that S was sensitive towards this evidence (things 7

The notion of triangulation plays an important part in late Davidson. To me it is somewhat unclear what use Davidson makes of it. It can be taken as a further specification of his traditional radical interpretation set-up, as further details about how to move from attribution of preferences between uninterpreted sentences in situations to an attribution of a full semantic theory and corresponding beliefs. In that case it is just more about what I do in my text, in particular it provides details about how to get to the simple perceptual attributions I describe. It can also be taken as a more ambitious philosophical theory that is meant to go some way towards accounting for what intentional contents are, or towards how intentional content is possible at all, i.e. it might move towards a full-blooded theory of meaning in M. Dummett’s sense of the latter. This latter project is not my project. It is not even a project I am in a position to recommend.

227 looking like bears) and formed the corresponding (false) belief that there was indeed a bear. In the normal case a sheep looks like a sheep. Still, how things appear to S may not be captured by a simple description of what S sees. When these things come apart, then the state described by the former type of description is typically, but not exclusively, caused by what S sees, hears, smells, etc. We have in fact general knowledge of such causal patterns, and we have no empirical grip of ascribing how things look and appear to S in isolation from ascriptions of what S sees, hears, etc. This general type of knowledge of such connections must in interpretation be applied by us, and obviously this general knowledge can only be applied in the light of what S sees, hears etc. It is therefore a fact that we need attributions with the notions “S sees...”, (example: S sees a sheep), “S hears....”, and others on our way towards interpretative hypotheses. The simple attributions a radical interpreter makes, the attributions of what the speaker sees and hears, have two fundamental properties that make them very appealing to a radical interpreter. Firstly, they are attributions which can be made by the interpreter without any implications about what concepts the speaker possesses, or any assumptions about the finegrained propositional attitudes of the speaker. Secondly, if you see a sheep eating grass, we can validly infer that there is a sheep eating grass that you see. Simple attributions are factive and imply truth in this sense: The simple attribution cannot be true unless the thing that you see or hear is there, with the properties ascribed to it in the noun phrase of the simple attribution. Look then at this complex attribution, and think of that as typical: “It appears to the speaker that there is a sheep there.” The semantics of this attribution is entirely different from the simple attributions in the paragraph above. In this case, if the attributions are correct, the speaker must necessarily master all the concepts in use after the “that”. Secondly, such attributions do not imply that the content, that a sheep is eating grass, is a true one. In some cases we attribute this content knowing perfectly well that the purported sheep is a goat dressed up like a sheep. These attributions are clearly fully loaded with intentionality, even intentionality in the Husserlian sense. Let us proceed towards explication of error.

228 Explications of error Take again the case where there is a sheep dressed up exactly as a goat. What is needed then in order to explain the error is to make use of the fact, if it is a fact, that it “appears to S that he is seeing a goat”. We attribute epistemic sensitivities toward seeing a dressed up sheep looking like a goat (“It looked to S exactly as a goat”), and on that basis we attribute the false belief to S that “There is a goat”. The role of “S sees a sheep” is fundamental in this explanatory enterprise, (and it accounts for the falsehood), but equally fundamental is the attribution of the evidential sensitivities to the speaker towards the fact that the animal in question looks like a goat. We have to bring in all of these things to attribute the false belief in the radical interpretation framework, for then to make use of that attribution of false belief in making adjustments in interpreting words on the basis of the preference for the truth of sentences. It is agreed that we in an interpretative hypothesis typically tentatively attribute very many concepts and a host of true beliefs to the speaker. I claim we must in a similar fashion make all the simple attributions, and on that basis attribute epistemic sensitivities related to the concepts we also attribute, albeit tentatively. When we attribute the epistemic sensitivities in these cases (“It looked to S exactly as a goat”), we do, I claim, tentatively attribute some knowledge to S, namely knowledge about goats; what they are, and how they look and appear. Without such knowledge, the epistemic sensitivities that underlie the attributions of false beliefs seem to be missing. To sum up: The explication of error works by two types of attributions, the simple attributions of “seeing a sheep dressed up like a goat”, in conjunction with the complex attribution of evidential sensitivities which involve “knowing how a goat looks”. Davidson thought, erroneously, that you could explain error by the phrase “being well placed or equipped to observe”. But in our case of error it is not as if error is to be ruled out as long as this predicate applies. Only if we read the phrase in question as “being well placed or equipped to know by observation”, do we really get on our way towards capturing what we need for the explication of this type of error. If this is right, it follows directly that error in these central cases amounts to failed knowledge. The notion of failed knowledge must then be

229 understood against the background where we normally do know by observation. Pulling together in a larger picture It seems really obvious by now that we need more than simple attributions to explicate error; we need to ascribe a conceptual repertoire and evidential sensitivities to the speaker. In order to account for error, we therefore need to move past the stage I like to identify as the initial stage of radical interpretation. This initial stage limits itself to verbal behaviour in situations in which the speaker is very well placed and equipped to know by observation the relevant aspects of his environment. The contents in question are such that we hold that were they to be entertained, the speaker would know them to be true. This first stage of radical interpretation concentrates on the cases where there is no reason to expect a mismatch in content between what is the case, simple attributions, and tentative complex attributions about how things appear. These are the cases where we see sheep and goats plainly in view, sheep and goats that look like sheep and goats. From this starting point we can move towards tentatively interpreting an interesting, albeit limited, part of the speaker's language. It also gives us resources for being able to start to account (still tentatively, though) for this type of error. Attributions of error can be conceived of as occurring on a second stage where we go beyond such simple situations, and adjust our interpretative hypotheses in the new terrain. The attribution of error takes place against the background of the attributions of knowledge, the knowledge that is crucial to explicate error. This procedure gives first priority to the verbal behaviour in the cases where we with reason can be said to know by observation. By employing the concept of knowledge by observation we identify these cases where there is indeed a match between the distal and proximal input, the initial phase of interpretation. We move so to speak beyond the defining battle ground, Quine’s and Davidson’s battle ground over whether we should go for distal or proximal stimuli, by first employing a concept which holds these two other concepts, the proximal and the distal, together in the right way. This is what the concept of knowledge here does. My argument is therefore meant to show the need for a concept of this exact nature, and for starting radical interpretation with cases identified by employing this con-

230 cept. When we move beyond these simple cases where knowledge by observation is easy and natural, then we also have the resources to give the right weight to how things look to the speaker, and we are able to identify explicable errors. We can sum this up by saying that we ought to make tentative attributions of types of knowledge to get well started in radical interpretation, and we also need to be able to attribute types of knowledge in order to be able to make attributions of false belief rather than adjusting our interpretative hypothesis in the more complex cases. There is a clear sense in which many errors can be seen as failed knowledge. Is the introduction of the concept of knowledge necessary? Most people find it incredible that we should employ the concept of knowledge at the very basic level in radical interpretation. This is very hard to digest if one thinks of knowledge as justified true belief. And I openly state that my aim here is a larger one of giving knowledge its rightful role in an account of content. Some people commenting my view have said that Davidson should instead speak of “being well placed and equipped to acquire true belief by observation.” This latter move by Davidson would be very welcome to me. However, Gettier cases show us that we are unable to provide reductive definitions of knowledge, and they also illustrate that true beliefs acquired by observation is a problematic notion. If the case at hand, facing the radical interpreter, is a Gettier case, should it be taken as belonging to this first stage of radical interpretation? I think not, but will not argue the case in detail. Note that the amendment to Davidson’s view seems to have to let the Gettier case in among the simple cases, and that is deeply problematic. Let me instead turn to the attribution of knowledge of how things look in the explanation of error, and our example. Perhaps we should not attribute knowledge here, but just beliefs about how goats appear. Not just any beliefs, of course, these beliefs would also have to be mainly true. If not, the explanation of error would not work. That explanation assumes that the sensitivity to the evidence is really there. For that to be so, we can hardly allow any central false beliefs about how goats look and appear, for instance the belief that goats are only turquoise in colour. True beliefs about how the animals appear need not be constitutive of having the con-

231 cept of a goat; that is true. The concept of goat must, however, be correctly attributed for the explanation of error to work, and so must the epistemic sensitivity towards the animal which in this case looks like a goat. For the latter, the sensitivity, we need to attribute many true beliefs about how goats look. Is Davidson’s position that we should at this point only attribute many true beliefs, and no knowledge, about how goats look to the speaker? I must admit that I cannot see clearly how Davidson’s view in that case would be motivated. We have, I stress, no general reason against attributing knowledge. Refraining from attributing knowledge of how goats look in the case of someone with the concept of goat is normally something we do just in case there is a specific reason for not attributing knowledge about how goats look. That is not so in our case. In fact, we normally think of the fact that a person does know how goats look as what accounts for the fact that the person has a large number of true beliefs about how they look. Knowing how goats look is therefore naturally taken as what underlies having the right sort of evidential sensitivity when you are well placed and equipped to see or hear a goat; the sort of evidential sensitivity which in its turn underlies the fact that you will be expected to believe falsely that you see a goat when presented with a sheep dressed up to look almost exactly like a goat. Knowing how a goat looks means being familiar with the way goats look, and presupposes having the concept of a goat. The person who is attributed the knowledge about how they look, is typically attributed the concept of a goat on the basis of discriminative behaviour and epistemic sensitivity towards them, and that is in radical interpretation attributed partly on the basis of attitudes towards utterances of sentences in simple situations. The simple situations are naturally identified as the situations where we can easily know by observation, and we can be attributed knowledge of the truth of the content if attributed entertainment of the content in question. Føllesdal’s contribution seen from my point of view Føllesdal stressed in 1975 the importance for interpretation of what is going on at the sensory surfaces. In stressing this, his position was a great improvement on both Quine and Davidson at the time. We can now see that the sensory surfaces are very important indeed. But we should also in-

232 vestigate the way they are important. Føllesdal also stressed that perception has propositional content, and as he says, that it can “thereby serve as evidence for judgment”.8 How do these two things hang together? I agree fully with Føllesdal that having propositional content is necessary for serving as evidence. But this feature in itself is not sufficient for identifying the perceptual states that are to count as evidence in the beginning phases of radical interpretation when we get the first tentative interpretative hypothesis reasonably well established. This seems to be the morale of the example with a sheep dressed up as a goat. We are engaged in the project of giving perceptual evidence a significant role in an account of meaning and content, and in this sense we have learned from Quine. My suggestion is that we do that step-wise. There are elements of that two-step thinking also in Quine and Davidson, in particular in Quine, with his stress on observation and occasion sentences and the special roles for such data. In fact, my approach would perhaps start from many of the same situations as the one’s Quine would start from, even if the conception of the initial phase is very different. The discussion above seems to support this thought: Evidence in the required sense is simply what we would know by looking if we had the concepts involved in the attribution. In this case this can be generalized to the unproblematic perceptual knowledge we have in the cases when we can take as evidence what we see, hear, smell and feel. I agree that some times the evidence is how things look and appear only. I agree that some times it is easier to know how things look than to know what they are. But these cases are complex cases, and we should get to them later in the game. The states we should start with are the states we think of as providing perceptual knowledge. These states are factive states, as are also, in my view, the state of experiencing something happen and the state of observing something happen. Knowledge is the most general factive state, and all these other states are instances of knowledge. Among these various factive states the ones captured by the simple attributions stand out by the fact that they do not imply much about the speaker’s sensitivity towards the evidence provided. That is why someone engaged in radical interpretation should be so keen on them. A lot more is implied in cases like attributions of experience or ob8

Føllesdal (1999), p. 725.

233 servation, and that ties in with the fact that the latter attributions imply many things about one’s conceptual repertoire. What is going on at the sensory surfaces of a person is not normally evidence that this person has. It is not even conceptually structured. But knowledge about what is going on at the sensory surfaces is of vital importance for the radical interpreter when making simple attributions in order to get rudimentary interpretation going, and for giving explicable error the right role when interpretation develops further. It is a central part of the causal knowledge we need in order to make simple attributions, and thus in order to start to explicate error. But its role and status is a causal one, and in order to see its relevance we cannot just look at such causal facts alone, we need to see their relevance as a vital part of the picture when identifying what we are well placed to know by observation. What matters is the attribution of what we see and hear, and of how things look and sound. The work of causality and the causal facts we assume are always structured by the conceptual structure we tentatively attribute to the speaker. It is also this same conceptual structure that provides structure to what we hypothesize the speaker can know by observation. Our causal knowledge is then vital for determining in particular cases whether the speaker’s states are in fact of this or that observational sort. Our causal knowledge is essential for attributing, tentatively, knowledge to the speaker of the world around her/him. On the present approach, attributions of meaning should be seen against the backdrop of such simple knowledge attributions. That brings out the role of evidence in the account of meaning. In saying this, I have moved much beyond Dagfinn’s actual formulations, and I have in my particular way, by employing knowledge, stressed the distal more than he did in 1975, and perhaps more than he still does. I do not expect him to agree with me in all I have said. But I think I have pointed to a way that takes us beyond the opposition between proximal and distal. Looking back at the debate between Quine, Davidson and Føllesdal from my perspective, we can see that Føllesdal is the one who first formulated the direction where the debate ought to go about meaning and content. He built on both Quine and Davidson, and he built more on Quine

234 than on Davidson. What made him able to point to the direction forward, Føllesdal might say, was the way he had studied Husserl.9

References Davidson, Donald (1990) “The Structure and Content of Truth”, Journal of Philosophy, Vol. 87, 1990, pp. 279-328. Davidson, Donald (1999) “Reply to Føllesdal”, in The Philosophy of Donald Davidson, The Library of Living Philosophers, edited by L.E. Hahn, Chicago and La Salle, Open Court, 1999, pp. 729-732. Føllesdal, Dagfinn (1975) “Meaning and Experience”, in Samuel Guttenplan (ed.) Mind and Language, Oxford, Clarendon Press, 1975, pp. 25-44. Føllesdal, Dagfinn (1999) “Triangulation”, in The Philosophy of Donald Davidson, The Library of Living Philosophers, edited by L.E. Hahn, Chicago and La Salle, Open Court, 1999, pp. 719-728. Gjelsvik, Olav (2012) “Knowledge and error: a new approach to radical interpretation”, in Donald Davidson on Truth, Meaning and the Mental, ed. G. Preyer, Oxford, OUP, 2012, pp.167-191. Preyer, Gerhard, ed. (2012) Donald Davidson on Truth, Meaning and the Mental, Oxford, OUP, 2012. Quine, W.V. (1960) Word and Object, Cambridge, Mass., MIT-Press, 1960.

9

I thank Pascal Engel, Dagfinn Føllesdal, Peter Pagin, Bjørn Ramberg and Tim Williamson for helpful comments on this material.

Essays Part V Reference

Føllesdal and Quine’s Slingshot John Perry*

Abstract The slingshot is a family of arguments to the effect that the semantic contributions of statements are limited to their truth-values. If we accept the conclusion, and wish to maintain the intelligibility of non-truth functional, statement-embedding contexts like ‘makes’ and ‘necessarily’, we must suppose that the ingredient structure of the complex statements they yield to be illusory. Gödel (1972) was puzzled by the argument. Church (1956) used it to confirm Frege’s (1892) decision, that the ‘Bedeutung’ or ‘denotation’ of a sentence is its truth-value. Davidson (1967a) used it to argue against positing facts or states of affairs in semantics, as Reichenbach (1947) had done. It was this last use that mainly interested Jon Barwise and I, in our paper ‘‘Semantic Innocence and Uncompromising Situations” (1981), in which we named the argument ‘‘The Slingshot.” However, years before Davidson deployed the slingshot, Quine (1953a) had employed the argument in formulating his objections to modal logic. And years before Barwise and I criticized the slingshot, Føllesdal (1961, 1966, 2004) had shown how to evade Quine’s argument. It is this episode I wish to explore, in order to understand Quine’s use of the argument, to better understand what is wrong with it, and to give overdue recognition to Føllesdal’s accomplishment.

Introduction An expression is an ingredient in a larger expression in which it is embedded, if its syntactic and semantic structure remains intact within the larger, embedding, expression; basically, if it is used, not merely mentioned. By the semantic contribution of a statement (closed sentence) I mean those factors that can affect the truthvalues of larger statements of which it is an ingredient. One such factor is clearly the truth-value of the statement itself. ‘~(Obama is *

University of California, Riverside

238

a Democrat)’ is false, because ‘Obama is a Democrat’ is true. Any context, that embeds statements to form new ones, where the truthvalues of the new statements are sensitive only to the truth-values of the embedded ones, is truth-functional. Many linguistic contexts that embed statements are not truthfunctional. For example, ‘Necessarily, Obama is a Democrat’ is false, although ‘Necessarily, Obama is a human’ is true; ‘necessarily’ appears to be sensitive not only to the truth-values of the embedded statements, which are the same, but also to the modality of the connection between their subjects and predicates, which differs. Moreover, it appears that the embedded statements are syntactically and semantically ingredients of the containing statements; the words are used, not mentioned. Consider another example, in which the apparent contribution of the embedded statement is not modality. Assume ‘Mary weeps’ and ‘Elwood grins’ are both true. Nevertheless, ‘Harry makes Mary weep’ might be true, and ‘Harry makes Elwood grin’ false. ‘Makes’ is sensitive not only to the truth-value of the statements it embeds, but also to their subject matter, whom and what they are about. It is plausible, then, that statements semantically contribute more than truth-values to larger statements of which they are ingredients, and some linguistic contexts are sensitive to these other factors. The slingshot is an argument, or a family of arguments, to the effect that, contrary to what just seemed plausible, the semantic contributions of statements are limited to their truth-values. Thus, if we wish to maintain the intelligibility of non-truth functional contexts like ‘makes’ and ‘necessarily’, we have to suppose that the ingredient structure they appear to have is illusory. The slingshot can be thought of as an argument that the following three principles are jointly inconsistent, and, since (A) and (B) are undeniable, (C) must be abandoned. (A)

(Redistribution) Statements S and S’ make the same

239

(B)

(C)

semantic contribution, if logic alone can show that they share truth-values. (Substitution) Statements φ(a) and ψ(b) make the same semantic contribution, if a = b, and φ and ψ predicate the same conditions. (Diversity) Statements S and S’ may make different semantic contributions, although they have the same truth-value.

The slingshot proceeds by providing a sequence of statements, each member of which makes the same semantic contribution as its successor, according to (A) or (B). The first and last members of the sequence share truth-values, but differ in some other factor one might think formed a part of their semantic contribution. The sequence purports to show that this factor is cannot a part of a statement’s semantic contribution after all. For example, consider: (D) (E) (F) (G)

Obama is human. [The object x such that Obama is human & x = Obama] = Obama (by (A)). [The object x such that Obama is a Democrat & x = Obama] = Obama (by (B)) Obama is a Democrat (by (A)).

(D) and (G) agree in truth-value, but not in modality. But their semantic contribution is the same, given principles (A) and (B). Hence modality is not, as we might have thought, part of a statement’s semantic contribution. Gödel (1972) was puzzled by the argument. Church (1956) used it to confirm Frege’s (1892) decision, that the ‘Bedeutung’ or ‘denotation’ of a sentence is its truth-value. Davidson (1967a) used it to argue against positing facts or states of affairs in semantics, as Reichenbach (1947) had done. It was this last use that mainly interested Jon Barwise and I, in our paper ‘‘Semantic Innocence and Un-

240

compromising Situations” (1981), in which we named the argument ‘‘The Slingshot,” for we wanted to follow Reichenbach in this regard.1 However, years before Davidson deployed the slingshot, Quine (1953a) had used principles like (A) and (B) to argue against (C), in the context of his objections to modal logic. And years before Barwise and I criticized the slingshot, Føllesdal (1961, 1966, 2004) had shown how to evade his argument. It is this episode I wish to explore, by way of an overdue appreciation of Føllesdal ‘s accomplishment. Quine and the Aristotelian Essentialist Logician I’ll focus on Quine’s essay, ‘‘Three Grades of Modal Involvement.” The slingshot comes up in his critique of a character I will call the ‘Aristotelian Essentialist Logician’, or ‘AEL’. The AEL holds four theses. First, the AEL believes that there is a distinction between statements that are necessarily true and those that are only contingently true, e.g. (1) (2)

Obama is human Obama is a Democrat

where (1) is necessary, and (2) merely contingent. Second, the AEL thinks this difference is not a matter of connection of meaning between the singular term used to identify the subject and the predicate; necessity is not analyticity. It is rather a relation between the subject itself, and the predicate. Obama himself is necessarily human, but only contingently a Democrat. The fact that ‘The leader of the Democratic party is a Democrat’ is, arguably, analytic, does not show that Obama, who is the leader, is necessarily a Democrat. The fact that ‘The object behind the podium is human’ is synthetic, does not show that Obama, who is the object behind the podium, is only contingently human. 1

See also Perry,1996.

241 Third, the AEL thinks there is no reason why operators such as ‘☐’ and ‘‘’ should not be sensitive to this factor of modality; that is, why modality should not be part of the semantic contribution of statements like (1) and (2). The AEL realizes and accepts that such operators will be ‘extensionally opaque’ in Quine’s terminology; that is, substitution of co-extensional predicates in the embedded statement, may affect the truth-value of the whole statement. To adapt Russell’s (1903, §125) classic example, consider: (3) ☐ (Obama is a featherless biped) (4) ☐ (Obama is a human) Even if being a featherless biped is coextensive with being human, as Russell invited us to suppose, (3) may be false even though (4) is true, on the assumption that it is possible for Obama to grow feathers. The operators are sensitive to the connection between Obama and the predicates, so the difference in predicates is relevant, even if their extensions are the same. Fourth, the AEL thinks that because modality is a relation between individuals and predicates, rather than singular terms and predicates, it will make sense to quantify across ‘☐’ and ‘‘’ as in the existential generalization of (4): (5) ∃x ☐ (x is human) This makes sense, because it makes sense for a the open sentence ‘☐(x is human)’ to be satisfied by an individual. Obama, for example, satisfies it, because he is necessarily human. A philosopher, whose modal tendencies lead him no further than using language like, (6)

‘Obama is human’ is analytic

242

can remain at what Quine calls the ‘first level of modal involvement’. It’s further than Quine is comfortable with, because of his reservations about analyticity, but it’s no more unintelligible than analyticity. If the philosopher’s modal tendencies lead him to express (6) with a sentence like (4), he will have sunk to the second level of modal involvement. Quine sees it as inherently confusing and most likely motivated by confusion, for one is embedding a sentence in (4) that is more properly mentioned as in (6); the embedded sentence appears to be an ingredient, but really isn’t. The AEL does not see necessity as analyticity, does not think (4) is confusing or confused, and aspires to sink to the third level of modal involvement, where quantifiers reach across ☐ or ‘ to bind variables, as with (5). The AEL thinks necessity is a matter of individuals having properties, and hence satisfying predicates and open sentences in a certain way. He thus sees nothing unintelligible about (5). Someone might sink to the third level by a different route. One might, for example, stick to the interpretation of necessity as analyticity, but reinterpret quantification substitutionally in order to make sense of (5). But the AEL does not wish to reinterpret quantification. He thinks (5) makes sense, because of his ‘‘Aristotelian Essentialism,” which, at this point in the career of the term, meant only recognizing different ways, or modes, in which individuals may come to satisfy open sentences. Quine disapproves of even this modest metaphysics. But the argument Quine makes isn’t straightforwardly against the metaphysics of the AEL. It is against the aspiration to implement this conception of modality by extending predicate logic, creating the illusion that the well-understood apparatus of identity, variables, quantifiers, names and other singular terms can be supplemented with modal operators with no loss of intelligibility. Quine argues to the contrary. The AEL will either have to give up the orthodox interpretation of identity, variables and quantifiers, or do without singular terms other than variables, or see modal distinctions, the

243 whole point of the exercise, collapse. It is in making this argument that Quine uses the slingshot. Some Terminology Before we consider Quine’s slingshot and Føllesdal’s response, let’s fix some terminology; my use differs slightly from theirs. I use ‘designate’ as the all-purpose term for the relation of singular terms to the objects they pick out, rather than ‘refers’. Later I will use ‘refers’ with a more specific meaning that is in line with distinctions that Føllesdal makes, and associates with the phrase ‘genuine singular term’. I’ll say that predicates, general terms, open sentences, and closed sentences (statements) have extensions; the extension of a predicate or general term is the set of objects that fall under it, of an open sentence, the set of objects of which it is true, and of a statement, its truth-value. Names, variables, descriptions, and class-abstracts are singular terms, and designate objects. Variables are assigned to objects, and free variables designate the objects they are assigned to. The designations and extensions of expressions containing variables will thus also be relative to a variable assignment, and the terms ‘extension’ and ‘designation’ will be used with the understanding that this is so. An expression α is an extensional ingredient in a larger expression β, if all that α provides to determining the designation or extension of β is its extension; in this case, substitution of another co-extensional expression for α will not change the designation or extension of β. Truth-functionality is a species of extensionality. An operator O is extensionally transparent, or e-transparent, if it does not affect the extensionality of the ingredients of expressions to which it is appended; that is, if α is extensional in β, it remains extensional in Ο(β). Otherwise, it is e-opaque. Anticipating Føllesdal, we do not assume that having a designation must be a species of having an extension, as was assumed by Frege (1892) (taking his Bedeutungen to correspond to extensions),

244

Carnap (1942), Quine (1959), and many others. So we need parallel definitions for designation. A singular term α is a designational ingredient in β, if all that α contributes to determining the extension or designation of β is its designation. An operator O is designationally transparent or d-transparent if it does not affect whether the ingredients of expressions to which it is appended are designational; that is, if if α is designational in β, it remains designational in O(β). Otherwise, it is d-opaque.2 The AEL wants his modal operators to be e-opaque; otherwise, Obama could not be necessarily human, without being necessarily a featherless biped. If modal operators were e-transparent, the modalities would ‘‘collapse”. But for sentences like (5) to make orthodox sense, the variable x must remain designational. (5) should be true if for some variable assignment to x, (7) is true. (7)

…(x is human)

And that should should just depend on the object assigned to the variable, and whether that object is a human contingently or necessarily. This combination of e-opacity and d-transparency can be sustained, Quine thinks, so long as the AEL limits himself to variables as the only expressions in the system that have designation; that is, eschews names, descriptions and abstracts. But within such a limited system the AEL could not provide an account of why one might suppose …(Barack Obama is male) 2

An operator might be e-transparent or d-transparent with respect certain ingredients in the expressions it embeds, while being e-opaque or d-opaque with respect to others. This complication isn't directly relevant to the issues I am discussing, so I will ignore it.

245

is true, while …(The President is male) is false. But, Quine argues, if the AEL adds singular terms to his system, and understands them in the orthodox way the project is doomed, for the combination of e-opacity and d-transparency cannot be sustained. At any rate, if not impossible, it is ‘‘... less easy than one at first supposes,” for Extensionality does not merely recommend itself on the score of simplicity and convenience; it rests on somewhat more compelling grounds, as the following argument will reveal. (Quine, 1953a: 163)

The ‘‘following argument” is Quine’s slingshot. Quine’s Slingshot Quine gives us, in effect, an all-purpose slingshot.3 Pick any two statements, p and q, that share a truth-value, but differ in any other factor that one might suppose formed a part of a statement’s semantic contribution. Then consider: (H) (I) (J) (K)

p {x|x = Ø & p} = {Ø} {x|x = Ø & q} = {Ø} q

According to principle (A), (H) and (I) make the same semantic contribution, as do (J) and (K). According to principle (B), keeping in mind that p and q have the same truth-value, (I) and (J) make the 3

By using class abstraction and the empty set, Quine achieves a certain elegance. But use of set theory is not an essential ingredient of the slingshot; see (D) – (G).

246

same semantic contribution. Thus (H) and (K) make the same semantic contribution. So the factors that differ, in virtue of the differences between ‘is human’ and ‘is a Democrat’, cannot be part of the semantic contribution. Now our AEL holds that ‘…(Obama is human)’ is true, while ‘…(Obama is a Democrat)’ is false. The predicate ‘is human’ applies to Obama necessarily, while ‘is a Democrat’ does not. The difference in predicates means a difference in modality, even though the truth-value stays the same. How can the AEL evade the slingshot? Consider (L) Obama is human (M) {x|x = Ø & Obama is human} = {Ø} Once we absorb what (M) means, it seems that (L) and (M) do have the same modal status. Since Obama is necessarily human (L) is necessary. Now the only way that the set {x|x = Ø & Obama is human} could fail to be {Ø}, is for Obama to fail to be human. So if Obama can’t fail to be human, the identity asserted by (M) cannot fail to be true, and (M), like (L) is necessary. Similarly, (N) (O)

{x|x = Ø & Obama is a Democrat} = {Ø} Obama is a Democrat

are both contingent. Since it is contingent that Obama is a Democrat, it is contingent that the identity asserted by (N) holds. Thus there is no reason for the AEL, at least so long as modality is the issue, to deny (A). The problem is with principle (B), which tells us that (M) and (N) make the same semantic contribution. This the AEL must deny, since he considers (M) necessary and (N) contingent. As we saw, the AEL gladly gives up extensionality. But principle (B) does not depend on extensionality. It depends rather on

247 substitutivity, the substitutivity of ‘{x|x = Ø & Obama is human}’ for ‘{x|x = Ø & Obama is a Democrat}’, given that they both designate the set with the empty set as its only member. Føllesdal ‘s Diagnosis Prior to criticizing it, Føllesdal defends Quine’s argument in two ways. First, he claims the conclusion of Quine’s argument follows from its premises. If we accept Quine’s premises, we cannot have a modal logic that is rich enough to formulate sentences like (5), includes singular terms such as names, descriptions and set abstracts, and adheres to an orthodox interpretation of quantification, variables, and identity. Second, he claims that the existence of the then existing systems of quantified modal logic did not refute Quine’s argument, because they either adopted an unorthodox treatment of quantification and variables, or an unorthodox treatment of identity, or did not incorporate singular terms, or incorporated them but treated them in an unorthodox way. But Føllesdal nevertheless does undermine the slingshot, and Quine’s argument against the AEL, because he finds an assumption of Quine’s that is not required by orthodoxy, and provides a principled reason for rejecting it. According to Føllesdal, Quine in effect assumes that an aspect of the (then) standard philosophical conception of how singular terms worked was required for an orthodox interpretation of logic. The assumption is one we mentioned above, that having a designation is a species of having an extension, or, as Quine puts it in Methods of Logic, ‘‘the primacy of predicates” (1959: 218). Frege had much earlier combined designations of singular terms, the extensions of predicates, and the truth-values of sentences under one general heading, ‘Bedeutungen’ (1892). This unifying idea was carried over by later writers, in the concept of extension. It might seem like a bit of harmless theoretical streamlining. But Føllesdal showed that it was more substantive. Statements, descriptions, and other complex singular terms are built out of predicates, and their designata vary with the exten-

248

sions of the predicates in them. So the idea that their designations are basically also extensions is very natural. But variables and names are not built out of predicates, and so cannot vary with the extensions contained in them. So to treat their designations as extensions is a further step, and a substantial one. If designation is a species of extension, then d-transparency must reduce to e-transparency. Then the step that is at the heart of Quine’s challenge to the AEL is irresistible: the AEL cannot reject e-transparency and retain d-transparency. The first thing Føllesdal does in his dissertation is to point out that there is however a principled reason for distinguishing and d-transparency from etransparency. He shows that there is a good argument to the effect that an operator cannot be both d-transparent and e-opaque. The reason is that operators do not embed variables and singular terms directly, but only as they occur as ingredients in embedded general terms and sentences.4 Føllesdal shows that if the operator is dopaque, then an expression that directly or indirectly embeds it will be e-opaque. But an analogous argument will not show that an operator cannot be both d-transparent and e-opaque, because operators can embed sentences and general terms which are not themselves ingredients in singular terms. So there is at least the possibility of operators that are, as the AEL needs them to be, d-transparent but eopaque. Føllesdal then goes on to dilineate the elements essential to an orthodox interpretation of singular terms, to which the AEL must adhere. Orthdoxy requires that quantifiers and variables be treated as they are in the predicate calculus. And it requires that the identity of objects that are the values of variables be ordinary objects; the object that serves at the witness to ‘(x is married & x is president) 4

Føllesdal does not claim that this is necessarily so, but only that there seem to be no exceptions to it in the constructions the AEL wishes to model.

249

should be an ordinary object, like Obama, the same sort of object that can witness the truth of x is married & x is president and not some sort of special ‘‘intensional” object, or an object that can ‘‘split” when we consider alternative possibilities. But, according to Føllesdal, the treatment of designation as a species of extension is not essential to meeting these criteria. Føllesdal does not claim that the reduction of designation to extension is inappropriate for complex terms like descriptions and class abstracts, but only that it is inappropriate for what he calls ‘genuine singular terms’. Singular terms whose semantics works like general terms, such as class abstracts and definite descriptions, are non-genuine. If proper names are hidden definite descriptions, they too are non-genuine. The key property of genuine singular terms is that they stick with their designata. Føllesdal explains the difference by considering the different behavior of genuine singular terms and nongenuine singular terms as we consider different possible worlds. ‘The President of the United States’ designates Obama in the actual world (at the time I write), but it would have designated McCain, had the 2008 election turned out differently. So it is not a genuine singular term. A genuine singular term for Obama would designate Obama no matter what possible world we consider. Genuine singular terms thus have the property Kripke was to call ‘rigidly designating’ (1972). I think it is important to keep in mind, however, that the crucial thing is that the designation of a genuine singular term persists as the term is embedded in larger expressions that include operators like ‘…’ and ‘‘’. It behaves just as a variable does, once the assignment is fixed. Thus the distinction does not depend on using possible worlds in one’s analysis of necessity and possibility.

250

The terminology I prefer reserves ‘referring’ for genuine singular terms and ‘denotes’ for singular terms with an extensionbased semantics, including descriptions and class abstracts. On this picture, designation comes in two forms, referring and denoting. This is why I have deviated from the Føllesdal /Quine use of ‘refers’. Føllesdal points out in his preface to the reprint of his dissertation that his work did not fully anticipate modern referentialism. He did not claim that ordinary proper names are genuine proper names, as Kripke (1972) and Donnellan (1970) were, in effect, to do, but left this question open. Nor did he provide something like a causal theory of reference, as Geach (1969), Donnellan, and Kripke were to do. A causal theory provides an account of how genuine singular terms can have the designations they do, which serves as an alternative to the picture that they work like general terms. Whatever the case with ordinary proper names, Føllesdal saw that there is no apparent logical or philosophical reason why there could not be genuine names; that is, singular terms that, like variables, are simply directly assigned to objects, but unlike variables have a designation that is fixed once and for all. There is no reason our AEL cannot incorporate such genuine singular terms into his system. Let us say an operator is referentially transparent, or rtransparent, if it does not affect whether genuine singular terms are designational. An operator may be r-transparent, without being dtransparent. In Obama = the President both ‘Obama’ and ‘the President’ are designational; that is, they only contribute their designation, Obama, to the determining the truth value of the statement. But in

251 ‘ ~ (Obama = the President) their status differs. ‘Obama’ is still designational (assuming it is indeed a genuine singular term), but ‘the President’ is not. Its semantic contribution is not just Obama, but also the property of being President. We have r-transparency, since the contribution of genuine singular terms is not affected. But we do not have dtransparency, since the contribution of the non-genuine terms is affected. Back to the Slingshot Føllesdal points out that Quine’s argument assumes that the AEL’s system will meet the following conditions (using my terminology). (a) (b) (c)

(d) (e)

Quantification and identity are interpreted in the orthodox way; ‘‘’ and ‘…’ are d-transparent; Interchange of embedded sentences that are logically equivalent does not effect the truth value of modal statements; Complex singular terms for classes are allowed; Transparency entails substitutivity.

All of the conditions are made quite explicit by Quine, except the last. And, Føllesdal insists, if we give Quine these premises, we must accept his conclusion, that the modalities collapse. Equipped with Føllesdal’s distinctions, however, the AEL need not accept the premises. Premise (b) is too strong. To exploit the apparatus of quantification and identity, the AEL need only claim that ‘‘’ and ‘…’ are r-transparent, so that variables and other genuine singular terms remain designational through the addition of these operators. Premise (e) now needs to be qualified, in recognition of the change in premise (b). R-transparency guarantees substitutivity of genuine singular terms, but not of all singular terms.

252

With these changes, the key step in Quine’s slingshot, from (M) {x|x = Ø & Obama is human} = {Ø} to (N)

{x|x = Ø & Obama is a Democrat} = {Ø}

can be rejected. The AEL’s holds that ‘‘’ and ‘…’ are rtransparent, by our modification of (b). And he holds that rtransparency guarantees substitutivity of genuine singular terms, by our modification of (e). But the move from (M) to (N) requires substitution of non-genuine singular terms. So the AEL need not accept that (M) and (G) have the same modal status. Now consider principle (B), (B)

(Substitution) Statements ϕ(a) and ψ(b) make the same semantic contribution, if a = b, and ϕ and ψ predicate the same conditions

It must be amended to be plausible; ‘a’ and ‘b’ must be genuine singular terms. So amended, it can’t support the slingshot.

Redistribution and Subject Matter Føllesdal pointed out that there are many constructions for which Quine’s argument threatened catastrophe: [Quine’s argument] was independent of the modalities and applied to any attempt to single out by help of an operator a proper subclass of the true sentences...Quine’s argument could be repeated for belief, causality, counterfactuals, probability, and the operators in ethics such as ‘‘it is obligatory that”, ‘‘it is permitted that”. They would all collapse. The argument was simply too disastrous to be correct. (2004: x--xi.)

253

At the outset, I gave one non-modal example of a non-truthfunctional contexts, ‘makes’, as in ‘Harry makes Mary weep’. With this example it seems that principle (A), Redistribution, is the culprit, and not principle (B). Like ‘necessarily’, ‘makes’ distributes over conjunctions. Unlike ‘necessarily’, but like many other non-truth-functional contexts, ‘makes’ distributes over disjunctions. If Harry makes Mary weep or Elwood grin, then he either makes Mary weep or he makes Elwood grin. Now consider this ‘‘mini-slingshot,” that does not depend on principle (B). (O) (P)

Mary runs Mary runs & [Elwood grins ∨ ~ (Elwood grins)]

By principle (A), (O) and (P) make the same semantic contribution. Hence it seems that we can infer (9) from (8) (8) (9)

Harry makes Mary weep Harry makes {Mary weep & [Elwood grin ∨ ~(Elwood grin)]}

But then, because ‘Harry makes’ distributes over conjunctions, we can infer (10) Harry makes [Elwood grin ∨ ~(Elwood grin)] and then, because it distributes over disjunctions, we get (11) Harry makes Elwood grin, ∨ Harry makes ~(Elwood grin). But of course, simply from the fact that Harry makes Mary weep, it does not follow that he makes Elwood do or not do anything.

254

Since ‘possibly’ does distribute over disjunction, we might expect that the mini-slingshot would pose a problem for it. But it doesn’t, because for every statement S, we have either ‘S or ‘ ~ S. So with ‘necessarily’ the mini-slingshot argument doesn’t get started, and with ‘possibility’ the final step doesn’t lead to a problem. The semantic contribution that is obscured by the move from (8) to (9) is what I call “subject matter”. (8) tells us of a relation between Harry and Mary, and Harry and the issue of Mary’s weeping or not; principle (A) allows us to import completely new subject matter, Elwood, and the issue of whether he grins or not. It seems that getting rid of (B) will not solve all of the problems of intensional logic. We need also to get rid of (A). Logical equivalence does not preserve semantic contribution, because it allows us to import extraneous subject matter. This inability to track subject matter seems to be built into the standard interpretation of the predicate calculus, in terms of total models. Every model that makes ‘Mary weeps’ true, will also make the conjunction ‘Elwood grins & ~ (Elwood grins)’ true. This feature of the predicate calculus, and the associated concept of logical equivalence, were used by Nelson Goodman (1961) to argue that we have no good concept of what a statement is about. Standard semantics for intensional logic, in terms of total possible worlds, which provide answers for every issue that can be framed in the base language, also loses track of subject matter. The intension of ‘Mary runs’ and ‘Mary runs and ‘Mary runs & [Elwood jumps ∨ ~(Elwood jumps)]’ will be the same set of possible worlds (See Perry, 1989). The sentence ‘Mary weep’ in ‘Harry makes Mary weep’ is an example of what is sometimes called a “naked-infinitive”. The idea is that ‘Mary weep’ and ‘Elwood grin’ are sentences that lack tense; the time of the jumping and running is the same as the time of seeing. This is a plausible but not completely uncontroversial idea.

255 Other examples seem not to require any controversial syntactic views. For example, ‘the x-ray shows that Gretchen has a broken leg’ does not imply that ‘the x-ray shows that Elwood has a fractured skull or the x-ray shows that Elwood doesn’t have a fractured skull’. Barwise and I (1983) tried to handle the slingshot within situation theory. By exploiting the various forms of partiality built into the theory, we denied both (A) and (B). Recognition of genuine reference is already a step in the direction of partiality, since the reference of a genuine singular term does not depend on the extensions of the predicates in the language. Still, I cannot see that recognition of genuine reference by itself can handle the problems that are due to (A). However, it is difficult to be certain, and Føllesdal’s dissertation may contain the ideas for way of blocking the redistribution step, or the ideas for explaining why doing so is not necessary. But, for the time being, I’ll take some solace in that it seems there was part of the story about the slingshot still left to tell when Barwise and I wrote our article.

References Barwise, Jon and John Perry, 1981. Semantic Innocence and Uncompromising Situations. Midwest Studies in Philosophy, Vol VI: 387--403. Barwise, Jon and John Perry, 1983. Situations and Attitudes. Cambridge: MIT/Bradford. Reprint with additions: Stanford: CSLI Publications, 1999. Carnap, Rudolf, 1942. Introduction to Semantics. Cambridge: Harvard University Press. Church, Alonzo, 1956. Introduction to Mathematical Logic. Princeton University Press.

Princeton:

Care, Norman S. and Robert M. Grimm, 1969 (eds.). Perception and Personal Identity. Cleveland: The Press of Case Western University.

256

Cohen, Robert S. and Marx W. Wartofsky (eds.), 1965. Boston Studies in the Philosophy of Science, Volume II. New York: Humanities Press. Clark, A., J. Ezquerro, and J. Larrazabal (eds.), 1996. Philosophy and Cognitive Science: Categories, Consciousness, and Reasoning. Dordrecht: Kluwer Academic Publishers. Davidson, Donald. 1967. Truth and Meaning. Synthese, 17. Davidson, Donald and Gilbert Harman (eds.), 1973. Semantics of Natural Language. Dordrecht: Reidel. Davidson, Donald and Jaakko Hintikka (eds.), 1969. Words and Objections. Dordrecht: Reidel. Donnellan, Keith, 1970. Proper Names and Identifying Descriptions. Synthese 21: pp. 335-358. Føllesdal, Dagfinn, 1965. Quantification into Causal Contexts. In Cohen and Wartofsky, 1965: 263--274; reprinted in Linsky, 1971: 52--62; page references to the latter. Føllesdal, Dagfinn, 1969. Quine and Modality. In Davidson & Hintikka, 1969: 175--85. Føllesdal, Dagfinn, 1961, 1966, 2004. Referential Opacity and Modal Logic. New York: Routledge. Originally published 1961, Ph.D. thesis, Harvard University Department of Philosophy. Published (in mimeographed form), Oslo, Oslo University Press, 1966. Page references are to the edition published in New York, Routledge, in 2004, which includes additional introductory material by Føllesdal. Frege, Gottlob, 1892. On Sense and Reference. Originally appeared in 1892 as Über Sinn und Bedeutung, Zeitschrift für Philosophie und philosophische Kritik L (1892). In Translations from the Philosophical Writings of Gottlob Frege, ed. Max Black and Peter Geach, trans. Max Black, 2nd edition Oxford: Basil Blackwell.

257 Geach, Peter. 1969. The Perils of Pauline. Review of Metaphysics 23: pp. 287--300. Gödel, Kurt. 1972. Russell’s Mathematical Logic. In David Pears, ed., Bertrand Russell: A Collection of Critical Essays. Garden City, N.Y.: Anchor Books. Goodman, Nelson. 1961. About. Mind 70: 1-24. Hintikka, Jaakko, 1969a. On The Logic of Perception. First published in Care and Grimm, 1969 : 140-173. Reprinted in Hintikka, 1969: 151183. Hintikka, Jaakko, 1969. Models for Modalities, Dordrecht: Reidel. Kripke, Saul, 1980. Naming and Necessity. Cambrdige, MA.: Harvard University Press. First published in Davidson and Harman, 1973: 253355, 763-769. Linsky, Leonard, (ed.), 1971. University Press.

Reference and Modality. Oxford: Oxford

Perry, John, 1989. Possible Worlds and Subject Matter: Discussion of Barbara H. Partee’s Possible Worlds in Model-Theoretic Semantics: A Linguistic Perspective. In S. Allen (ed.). Possible Worlds in Humanities, Arts and Sciences: Proceedings of Nobel Symposium, August, 1986. Berlin and New York: Walter de Gruyter. Reprinted in Perry, 2000. Perry, John, 1996. Evading the Slingshot. In A. Clark, J. Ezquerro, and J. Larrazabal (eds.), 1996. Reprinted in Perry, 2000. Perry, John, 2000. The Problem of the Essential Indexical and Other Essays, Enlarged edition. Stanford: CSLI Publications. Quine, Willard van Orman, 1953. Three Grades of Modal Involvement. Proceedings of the Eleventh International Congress of Philosophy: vol. 13, pp. 65--81. Reprinted in Quine, 1976: 158-176. References are to the reprint.

258

Quine, Willard van Orman, 1976. The Ways of Paradox, and other essays. Second edition. Cambridge: Harvard University Press. Quine, Willard van Orman, 1953. From a Logical Point of View. Cambridge, MA.: Harvard University Press. Quine, Willard van Orman, 1959. Methods of Logic, revised edition. New York: Holt, Rinehart and Winston. Reichenbach, Hans, 1947. Elements of Symbolic Logic. New York: The Free Press. Russell, Bertrand, 1903. The Principles of Mathematics. Cambridge: Cambridge University Press.

Føllesdal and Frege on Reference Øystein Linnebo∗

0. Introduction In a recent essay on reference, Dagfinn Føllesdal writes the following. One must distinguish between on the one hand a two-sorted semantics, where terms that refer are treated quite differently from general terms and other nonreferring expressions, and on the other hand a theory of reference, that is an account of the relation between referring expressions and their objects. What I argued for in my dissertation was the former. I recognized that talk about modality, knowledge, belief, causation, change, probability, ethics, etc., makes sense only if the referring expressions in our language have a semantics that is very different from that of general terms. At that time I thought that the referring expressions actually succeed in relating to their objects regardless of how the world and our theories about the world change, and I was at a loss as to how to explain how this could happen. (Føllesdal 1997, p. 359; see also Føllesdal 2004, p. xxviii.)

In this important passage Føllesdal distinguishes between semantics proper and the theory of reference. He explains how in his dissertation he developed a novel “two-sorted semantics” where singular terms are treated very differently from general terms: where a general term can apply to different objects in different possible worlds, a singular term applies to the same object in all possible worlds (at least in which the object exists). But Føllesdal also admits that early in his career he didn’t know how to develop a theory of reference, that is, an account of the relation that obtains between a referring expression and its referent. In this essay I first identify some desiderata for a theory of reference that appear in Føllesdal’s later work (Section 1). We will see that these desiderata are highly Fregean in character. Next I outline a Frege-inspired theory of (a certain core form of) reference that I have been developing in my own ∗

Birkbeck, University of London

260 work (Section 2). Finally I argue that this theory (as far as it goes) satisfies Føllesdal’s desiderata (Section 3). In particular, I argue that, when the distinction between semantics proper and the theory of reference is carefully heeded, my Frege-inspired theory of reference can be seen to be fully compatible with Føllesdal’s “two-sorted semantics.” The Fregeanism that I advocate is thus highly selective: its only direct concern is the theory of reference, not semantics proper. 1. Føllesdal on Singular Terms In his 1961 dissertation, written at Harvard under the supervision of W.V. Quine, Føllesdal investigates Quine’s famous objection to quantified modal logic. A careful examination of Quine’s objection leads Føllesdal to formulate the notion of what we now know as a rigid designator (which Føllesdal calls a genuine singular term). A rigid designator is, as we all know, a term that denotes the same object in all possible worlds (or at least in all worlds in which this object exists). Føllesdal shows us how Quine’s objection can be avoided by carefully distinguishing between rigid and non-rigid designators and by holding that singular terms, unlike definite descriptions, are rigid designators. This is an important contribution to semantics in which Føllesdal articulates and defends one of the most important doctrines now usually associated with Saul Kripke. To understand the similarities and differences between Føllesdal and Kripke, it will be useful to distinguish between what I will call semantics and meta-semantics. This distinction generalizes and subsumes the distinction Føllesdal draws in the passage just quoted between a semantics and a theory of reference.1 Semantics typically takes the form of a theory of semantic values, where the semantic value [E] of an expression E is the contribution that E makes to the truth-values of sentences in which it occurs.2 Following Frege, it is argued that semantic values are subject to a principle of compositionality, according to which the semantic value of a complex expression 1

My distinction between semantics and meta-semantics is thus the same as Stalnaker’s distinction between “descriptive” and “foundational” semantics. See e.g. Stalnaker 1997. The same distinction is found in the work of various other philosophers as well, for instance Michael Dummett, Richard Heck, and Jason Stanley. 2 I will use boldface for all meta-linguistic variables.

261 is determined as a function of the semantic values of its individual subexpressions. For instance, the semantic value of an atomic sentence P(a1, … , an) is functionally determined as [P]([a1], … , [an]). Following Frege again, the semantic value of a sentence is often taken to be just its truthvalue, and the semantic value of a proper name, its referent. If these two assumptions are accepted, then the principle of compositionality will allow us to determine the kinds of semantic values had by other types of expressions; for instance, the semantic value of a one-place predicate must be a function from objects to truth-values. I said above that Føllesdal’s discovery of the notion of a rigid designator was an important contribution to semantics. When I said this, I had in mind semantics in the precise sense that I have just explained. The semantic value of a rigid designator is an object. And this semantic value remains associated with the term throughout all possible worlds. This contrasts with the view known as descriptivism, which holds that the semantic value of a proper name is identical to that of some definite description associated with the name. It also contrasts with Carnap’s view that the semantic value of a name is an intension, that is, a function from possible worlds to objects. So concerning semantics, Føllesdal and Kripke appear to be largely in agreement. In particular, both are fully committed to the rigidity of names and the principle of compositionality. Meta-semantics, on the other hand, is concerned with what is involved in an expression’s having the various semantic properties that it happens to have, such as its semantic structure and its semantic value. The expressions in question are purely syntactical items—ink marks on paper and vibrations in the air—and thus have no intrinsic semantic significance. What is it, then, that endows these intrinsically “dead” syntactic items with semantic structure and semantic values? These questions belong to metasemantics rather than semantics proper. The relation between semantics and meta-semantics can be compared with that between economics and what we may call meta-economics. Economics is concerned with the laws governing money; for instance, that an excessive supply of money leads to inflation. Meta-economics, on the other hand, is concerned with what is involved in various objects’ having monetary value; for instance, what

262 makes it the case that a piece of printed paper can be worth €100. Since neither semantic nor monetary properties are intrinsic to the items in question, there must be some account of what the possession of such properties consists in. This account is likely to draw on both psychological facts about the agents who operate with the items in question and sociological facts about these agents’ interaction. Our primary interest is in meta-semantic questions concerning singular terms. When a singular term refers to an object, what is the nature of this relation? What does this relationship consist in? The most influential answers to these meta-semantic questions are due to Kripke and Gareth Evans. According to Kripke, a name refers to its bearer in virtue of the historical chains by which the name has been passed down from people directly acquainted with the referent to the contemporary users of the name.3 People’s beliefs about a name and its reference thus play an absolutely minimal role in Kripke’s account of how the name comes to have its reference. Evans, on the other hand, chooses to build on the idea of information being transmitted in causal chains. Føllesdal repeatedly and forcefully distances himself from such historical and causal theories of reference. He even writes that “my view [on the tie between singular terms and their objects] is much closer to Frege’s than to Kripke’s” (Føllesdal 1986, p. 109; see also Føllesdal 1997, p. 362). At first this claim may seem puzzling. How can Føllesdal, who like Kripke has vehemently criticized the descriptivist view of reference typically associated with Frege,4 now suddenly declare that he wants a view that is closer to Frege’s than to Kripke’s? But in fact there is absolutely no contradiction here. The distinction between semantics and meta-semantics helps us to see why. What Føllesdal has been defending, as against Frege, is the semantic thesis that names are rigid designators. But the passage just quoted in no way goes back on this. Rather, what Føllesdal says in this passage is that he wants a meta-semantic account of the link between a name and its 3

As Føllesdal 2004 pp. xxix-xxx points out, this view is anticipated in Geach 1969. This interpretation of Frege has been challenged; see for instance Evans 1982 and McDowell 1977. For the purposes of this article I need not take a stand on this exegetical issue. 4

263 referent which is closer to Frege’s than to Kripke’s—while holding on to the semantic thesis that names are rigid designators. The following desiderata for a meta-semantic theory of reference can be extracted from Føllesdal’s writings. 1. Generality. The theory of reference must be general enough to cover reference to all kinds of objects, including abstract objects, with which there can be no direct acquaintance or causal interaction.5 2. Fregean sense. The theory must allow for some Fregean notion of sense; that is, for some mode of presentation of the referent.6 3. Reference failure. The theory must have something plausible to say about cases of reference failure, where the subject does his part of the job but the world fails to provide a unique referent. (Frege attempted to account for this by allowing a full-blown sense even in the absence of a referent.)7 4. Individuation. The theory must bring out the role that individuation plays in the determination of reference.8 5. Cognitive Constraint. In “non-parasitic” cases of reference, where a person refers to an object all by himself and not merely exploits the referential capacities of other members of his community, this person must know which object he is referring to; that is, he must possess some discriminating conception of the referent.9 5

See e.g. Føllesdal 2004 , p. xxxi. See e.g. Føllesdal 1986 , p. 111 and Føllesdal 1997, p. 364. 7 See e.g. Føllesdal 1986, pp. 110-111; Føllesdal 1997, p. 359; and Føllesdal 2004, p. xxxi. 8 See e.g. Føllesdal 2004, p. xxxii. 9 See e.g. Føllesdal 1986, p. 109 and Føllesdal 2004, p. xxxi. The desideratum of Cognitive Constraint is related to what Evans 1982 calls Russell’s Principle. But Cognitive Constraint is weaker than Russell’s Principle because of its explicit restriction to nonparasitic cases of reference. 6

264

Summing up this section, I first subsumed Føllesdal’s distinction between a semantics and a theory of reference under the more general distinction between semantics and meta-semantics. Then I explained how Føllesdal and Kripke agree on the semantics of singular terms: both defend the thesis that singular terms are rigid designators. Finally I explained how Føllesdal sharply disagrees with Kripke on the meta-semantics of singular terms: where Kripke defends his well-known historical account, Føllesdal wants a more Fregean theory of reference that satisfies the above five desiderata. 2. Towards a Fregean Theory of Reference Frege’s first serious investigation of the problem of reference appears to have been largely motivated by the desideratum of Generality. For at Grundlagen (Frege 1884) §62 Frege raises the following question. “How, then, are the numbers to be given to us, if we cannot have any ideas or intuitions of them?” The problem—which is still very much with us—is of course that numbers cannot be perceived or in any way experimentally detected. How then can we refer to such objects? I claim that Frege’s question at §62 belongs to what I have called metasemantics. To see this, recall that by this stage of Grundlagen Frege has already argued that the numerals refer to objects (as opposed to second-order concepts) and rejected the psychologistic view that these objects are purely mental items. So Frege is at this stage entitled to assume his own platonistic view that the numbers are independently existing objects. Given this assumption, the semantic question what objects different numerals refer to has a completely straightforward answer; for instance, the numerals ‘7’ and ‘VII’ refer to the number 7. Rather, Frege’s present concern is with the meta-semantic question what facts about reference to natural numbers consist in. How can our numerals “latch on to” the natural numbers, given that there is no perception or causal interaction that can serve as a link? Frege’s next sentence proposes a way of addressing this hard metasemantic question. “Since it is only in the context of a sentence that words have any meaning, our problem becomes this: To define the sense of a sentence in which a number word occurs.” The doctrine that words have

265 meaning only in the context of a sentence has become known as the Context Principle.10 What Frege proposes is that the Context Principle has an essential role to play in the explanation of reference, both in general and to numbers and other abstract objects in particular. The idea is to translate the problem of explaining what it is for a singular term to refer into the problem of explaining what it is for certain complete sentences involving this term to be meaningful. I will now outline a Frege-inspired theory of reference which is based on this proposal. Although I believe Frege anticipated many aspects of this theory, he would probably have disagreed with other parts of it. But my present goal is systematic, not exegetical. I begin by narrowing down the problem in two different ways. My first restriction is to focus on thought rather than on language.11 For the purposes of this paper I will thus not attempt to say anything about how linguistic expressions come to refer but rather focus on the corresponding problem concerning mental representations. My explanandum will thus be what is involved in someone’s capacity for singular reference to various sorts of objects. The Frege-inspired proposal that I will investigate is that an adequate explanation of this capacity for singular reference will take the form of an explanation of what is involved in the person’s capacity for understanding complete thoughts concerning objects of the sort in question.12 10

See also ibid. pp. x, 71, and 116. I have changed the translation of ‘Satz’ from ‘proposition’ to ‘sentence’. This is reasonable, given that Frege talks about words occurring in a “Satz.” 11 In doing so I am to a large extent leaving the historical Frege behind, given his emphasis on starting with language rather than thought. But this departure is less radical than it may seem. In particular, my approach to thought (roughly that of Evans 1982) is very different from “the psychologistic” one that Frege so forcefully criticized: I seek to show how thought can have an objective propositional content, which is (at least in principle) intersubjectively accessible. 12 Strictly speaking, I here collapse two steps. The first step is Frege’s suggestion that questions concerning singular reference be addressed in terms of analogous questions concerning complete thoughts. In particular, in virtue of what does a physical state of an agent have a particular thought as its content? The second step is to approach this question about thoughts in terms of the notion of understanding. Doing so is quite nat-

266

This first restriction allows us to concentrate on an individual person rather than on a whole language community. This is a huge simplification. For instance, we can now hold that reference involves some Fregean mode of presentation but allow this mode of presentation to vary with each individual act of reference.13 In contrast, if a notion of sense is to be attached to an expression of a public language, then this sense will have to be shared by every competent speaker of this language. My second restriction is to focus on canonical cases of singular reference.14 These are certain maximally direct ways of referring to objects, where the referent is “directly present” to the thinker. For instance, referring to a person whom I see immediately in front of me is canonical, whereas referring to Napoleon, with whom I am in no way acquainted, is not. (More examples of canonical reference will be presented shortly.) Having made these two restrictions, Frege’s proposal becomes the following: We can explain what is involved in someone’s capacity for canonical singular reference to objects of a certain kind by explaining what is involved in his or her capacity for understanding complete thoughts concerning such objects. A slight simplification of this is possible. We observe that it makes sense to begin by explaining someone’s understanding of identity statements before attempting to explain his understanding of thoughts more generally. This strategy is adopted by Frege himself in Grundlagen.15 The rationale is ural; for in order to stand in some propositional attitude to a thought, one presumably needs to understand that thought. 13 The resulting notion of mode of presentation may thus have more in common with Husserl’s notion of noema than with Frege’s notion of Sinn. 14 In the terminology of Evans 1982, my goal is to explain what our understanding of the relevant kind of “fundamental Ideas” consists in. Following Michael Dummett and Gareth Evans I believe non-canonical reference must be explained in terms of someone’s ability to recognize the referent when presented with it in a canonical way. See Dummett 1981, pp. 231-239 and Evans 1982, pp. 109-112. 15 Frege unfortunately abandons this strategy in Grundgesetze. For an analysis, see my Linnebo 2004.

267 that, before one can understand what it means for an object to possess properties and stand in relations, one needs to know how to distinguish the object from other objects and how to re-identify it when presented with it in alternative ways.16 When this observation is added, we arrive at what will be our official statement of Frege’s proposal: We can translate the problem of explaining our capacity for canonical singular reference into the related but different problem of explaining our capacity for understanding identity statements concerning the object in question. How should this proposal be carried out? Again Frege makes an ingenious suggestion. The core idea is that canonical reference has a rich and systematic structure. Firstly, objects are always presented to us only via some of their parts or aspects. And secondly, we have a grasp of how two such parts or aspects must be related for them to pick out the same object. Here are some examples.17 1. Physical bodies.18 A physical body is most directly presented in perception, where we causally interact with one or more of its spatiotemporal parts. Two such parts determine the same physical body just in case they are connected through a continuous stretch of solid stuff, all of which belongs to a common unit of motion.19 2. Directions. A direction is most directly presented by means of a line (or some other directed object) that has the direction in question. Two lines determine the same direction just in case they are parallel. 16

Cf. Evans 1982, who explains fundamental Ideas in terms of “fundamental grounds of difference.” 17 In a more complete treatment, each example would of course have to be developed in greater detail and defended against objections. My present goal is merely to sketch some promising examples in order to illustrate how the Fregean framework functions. 18 By “physical body” I mean, roughly, a cohesive physical object with natural boundaries. Apples and oranges are thus paradigmatic physical bodies. By contrast, proper parts of an apple and arbitrary mereological sums of apple stuff may be physical objects but fail to qualify as physical bodies. 19 I elaborate on this view and defend it against some natural objections in my Linnebo 2005.

268

3. Shapes. This case is analogous to that of directions: Shapes are most directly presented by things or figures that have the shape in question. Two such things or figures determine the same shape just in case they are congruent. 4. Syntactic types. Syntactic types are most directly presented by means of their tokens. Two tokens determine the same type just in case they count as equivalent (“equi-typical”) according to the relevant standards. 5. Natural numbers. A natural number is most directly presented by means of some member of a sequence of numerals. Two numerals determine the same number just in case they occupy analogous positions in their respective sequences.20 These examples suggest that canonical cases of singular reference are always based on two elements. First, there is an intermediary entity in terms of which the referent is most immediately presented. Let’s call this the presentation. Second, there is a relation which specifies the condition under which two presentations determine the same referent. Let’s refer to this as the unity relation. Finally, let’s call an ordered pair 〈u, ≈〉 consisting of a presentation u and a unity relation ≈ applicable to this presentation a referential attempt. Frege’s proposal is then that canonical reference is based on referential attempts. Does this proposal yield an adequate explanation of what someone’s understanding of singular reference consists in? A formal adequacy condition is obviously that the account be non-circular. It is easily seen that the form of our proposal allows it to be non-circular. 20

Related ideas are found in Parsons 1971. This view of the natural numbers as finite ordinals contrasts with the logicist view that the natural numbers are finite cardinals, individuated by Hume’s Principle (which says that two numbers are identical just in case the concepts whose numbers they are are equinumerous). However, both views are compatible with the Fregean account of reference. It is thus largely an empirical question which view best describes human thought about the natural numbers.

269 Consider for instance the case of directions. What I have proposed is that someone’s understanding of an identity statement concerning directions can be explained in terms of his being suitably related to lines (in terms of which directions are presented) and having a suitable grasp of parallelism (which is the unity relation). In this case there is no threat of circularity; for we can explain what it is for someone to be suitably related to lines and to have a suitable grasp of parallelism without presupposing any prior ability to explain reference to directions. Next, we observe that there is nothing in this example that is peculiar to the case of directions. My proposal is to explain someone’s understanding of an identity statement in terms of this person’s being suitably related to the relevant presentations and having a suitable grasp of the relevant unity relation. This explanation will of course have to include an account of what it is for a person to be suitably related to these presentations and to have a suitable grasp of this unity relation. But there is no general reason why this account should presuppose what we are trying to explain, namely reference to the sort of objects that are determined by these presentations and this unity relation.21 The material adequacy condition is that the account should capture what someone’s capacity for singular reference consists in. My argument that this adequacy condition is satisfied is based on two claims: first, that my account explains what the subject’s understanding of identity statements involving the referent consists in; and second, that this understanding explains the subject’s capacity for singular reference to the object in question. Let’s begin with the first claim. Consider a representation a purporting to make singular reference to some object. According to my account, this representation is associated with some referential attempt 〈u, ≈〉, which specifies how the referent is presented and when two such presentations deter21

This is of course not to say that there cannot be particular cases where such an illicit presupposition exists. In fact, elsewhere I suggest that some of the problems encountered by Frege’s proposal are caused by the use of presentations and unity relations an adequate grasp of which would presuppose an ability to refer to the entities in question, thus making the account viciously circular. See Linnebo 2006, pp. 166-168 and Linnebo forthcoming, Section 2.2.

270 mine the same referent. By operating with this referential attempt, the subject will be able to understand any thought of the form ┌a = b┐, where b is any other representation purporting to make singular reference to an object of the kind in question. For according to my account, b too must be associated with some referential attempt, say 〈v, ≈〉. Moreover, we are assuming that the subject operates correctly with these representations, namely in accordance with the following principle for the identity of their semantic values: (SV)

[a] = [b] ↔ u ≈ v

This means that the subject has an ability to track the referent of a and to distinguish it from other objects of the same sort. My second claim is that this competence is naturally described as knowing (in a non-parasitic way) what object the representation a refers to. Consider for instance the case of physical bodies. Assume someone is digging in the garden, hits upon something hard with her shovel, and as a result forms the thought: This body is large. Later she hits upon something hard again, one meter away from the first encounter, and as a result forms the thought: This body is identical to that body. Finally, our subject appreciates that this identity statement is true just in case the two chunks of solid stuff that she has hit upon are spatiotemporally connected in the suitable way. It is extremely plausible to describe this capacity as a capacity to refer to physical bodies. For instance, if a robot was equipped with perception-like mechanisms and programmed so as to operate with the appropriate unity relation, it would make sense to ascribe to the robot a basic capacity for referring to physical bodies. On the view that I am advocating, the unity relation ≈ implicitly defines a (partial) function f≈ that maps a presentation u to the referent, if any, that u picks out. This is encapsulated in what I will call principles of individuation: (PI)

f≈(u) = f≈(v) ↔ u ≈ v

271

Of course, when formulating principles of individuation, we philosophers make use of our own ability to refer to objects of the kind in question. But this is perfectly permissible. We are allowed to presuppose that we can refer to objects of the kind in question. What we are not allowed to presuppose is an explanation of what this ability consists in. But no such presupposition is made. 3. Assessing the Fregean Theory of Reference Recall from Section 1 the kind of account that Føllesdal wants of the phenomenon of reference. Firstly, he wants a semantics that incorporates both the principle of compositionality and his own discovery that names are rigid designators. Secondly, he wants a meta-semantic account (or what he calls a “theory of reference”) that satisfies the five desiderata that I identified. I will now assess the extent to which the Frege-inspired theory of reference outlined in the previous section delivers what Føllesdal wants. Obviously, my theory can at best be the beginning of the sort of theory Føllesdal wants. For I have explicitly restricted my attention in two ways: firstly to reference at the level of thought rather than at the level of language; and secondly, to canonical cases of such reference. By contrast, Føllesdal is concerned with the reference of linguistic expressions, including cases that can in no way be regarded as canonical. My question can thus only be whether my theory, as far as it goes, has those features that Føllesdal wants.22 If it does, then this theory may serve as a stepping stone for later attempts to develop the more ambitious sort of theory that Føllesdal ultimately wants.23

22

Henceforth, this qualification will tacitly be assumed to be in place unless otherwise stated. 23 Should it, on the other hand, turn out that no such ambitious theory is possible, then this would not automatically threaten my less ambitious theory. For core cases of some phenomenon may well enjoy a particularly nice explanation although this explanation cannot be extended to more peripheral cases.

272 3.1. The meta-semantic desiderata It is fairly straightforward to see that my Frege-inspired theory of reference satisfies the five meta-semantic desiderata listed at the end of Section 1. The first desideratum—Generality—is clearly satisfied. Indeed, the theory was explicitly designed so as to be able to accommodate reference to all kinds of objects, including abstract ones. Next, we observe that the theory incorporates a notion of Fregean sense or mode of presentation of the referent. For according to this theory, canonical reference to an object is mediated by a presentation and a unity relation. Note that this notion of sense is not spelled out in a descriptivist manner; that is, it does not identify the semantic contribution of a proper name with that of some description.24 Rather, on my account the mode of presentation is a feature of the meta-semantic mechanism by which a semantically simple item comes to refer to an object. A mode of presentation is thus a part of an explanation of how an item comes to possess a particular semantic content, but not part of this content itself. Further, I claim that the theory provides a plausible analysis of what is going on in cases of reference failure. For according to my theory, there will be presentations that fail to determine referents. There are for instance spatiotemporal parts that fail to determine unique bodies. For instance, if I point to the floor and say ‘this body’, I will probably fail to determine a unique body.25

24

For other attempts to articulate a non-descriptivist notion of sense, see Evans 1982 and McDowell 1977. 25 The possibility of reference failure has consequences for the unity relation. Since a unity relation gives the condition under which two presentations determine the same referent, this relation will always have to be symmetric and transitive. However, a unity relation will only be reflexive on those presentations that succeed in determining referents. For a referential attempt 〈u, ≈〉 succeeds in determining a referent just in case the presentation u bears the unity relation ≈ to itself. From this it also follows that the function f≈ that figures in the principles of individuation is a partial function whose domain is identical to the field of the unity relation ≈. (Recall that the field of a relation R is the set of objects which R relates. Thus, when R is dyadic, its field is the set {x |

273

The final two desiderata were Individuation—which requires that the notion of individuation play a role in the determination of reference—and Cognitive Constraint—which says that for a thinker to refer nonparasitically to an object, he must know which object he is referring to. These desiderata are clearly satisfied as well. 3.2. The principle of compositionality I now turn to a worry about the compatibility of my Fregean theory of reference with the principle of compositionality. This worry is based on the following observation. According to the principle of compositionality, the meaning of a complex expression is to be explained in terms of the meanings of its constituent parts. So here we are explaining a property of a complex expression in terms of the properties of its simple constituents. But according to my Fregean theory of reference, the referentiality of a singular representation is partially explained in terms of the meaningfulness of identities involving this representation. So here we are explaining a property of a semantically simple item in terms of the properties of more complex expressions of which this simple is a constituent. This means that “the order of explanation” associated with my Fregean theory of reference is the opposite of that dictated by the principle of compositionality. It thus appears that my theory conflicts with the important principle of compositionality.26 I would like to suggest that our distinction between semantics and metasemantics provides the key to resolving the apparent conflict. The principle of compositionality is concerned with the assignment of semantic values to complex expressions and thus belongs to semantics. My Fregean theory of reference, on the other hand, is concerned with what is involved in an ex∃y (Rxy ∨ Ryx)}. Note that it is a theorem of first-order logic that a symmetric and transitive relation is reflexive on all objects in its field.) One way not to resolve this apparent conflict would be by claiming that, whereas my theory is only concerned with thought, the principle of compositionality governs only linguistic meaning. This response is unacceptable because an analogous principle of compositionality is supposed to apply to the contents of thoughts.

26

274 pression’s having the various semantic properties it happens to have and thus belongs to meta-semantics. Since the principle of compositionality and our Fregean account of reference have completely different concerns, there is no conflict. What the principle of compositionality says is that the semantic value of a complex expression is determined by the semantic values of its simple constituents. But it says nothing about how other kinds of explanation—such as meta-semantic explanations of what it is for an expression to have a semantic value in the first place—should proceed. However, the most popular response among philosophers who seek to use some form of Frege’s Context Principle to explain reference has been to concede that the apparent conflict is genuine and therefore to argue that the principle of compositionality has to be rejected or at least weakened.27 But rejecting or weakening the principle of compositionality is obviously a steep price to pay. Why, then, have so many philosophers found this response inevitable? I believe the answer has to do with a dangerous ambiguity in the wording of Frege’s proposal. “Since it is only in the context of a sentence that words have any meaning,” Frege writes, “our problem becomes this: To define the sense of a sentence in which a number word occurs.”28 This is ambiguous between a semantic and a meta-semantic reading. On the semantic reading, our task becomes to specify the meaning or sense of identity statements in which number words occur. But on the meta-semantic reading, our task is to explain what makes it the case that such identity statements have the meanings that they happen to have. Frege’s proposal has traditionally been interpreted along the lines of the semantic reading. It is then natural to assume that what Frege proposes is that the meaning of problematic identity statements be given by a “reductive” truth-condition (T-Red) 27 28

┌

a = b┐ is true iff u ≈ v

See Hale 1997 and Wright 1997. Grundlagen §62; my italics.

275

where a and b are representations associated with referential attempts 〈u, ≈〉 and 〈v, ≈〉 respectively. On this reading there will indeed be a conflict with the principle of compositionality. For according to this principle, the semantic value of an atomic sentence P(a1, … , an) is functionally determined as [P]([a1], … , [an]). Applied to the identity ┌a = b┐, this yields a different, completely trivial truth-condition: (T-Triv)

┌

a = b┐ is true iff [a] = [b]

(And any further or alternative semantic analysis is out of the question, given that the terms a and b are supposed to be semantically simple.) Moreover, the truth-condition (T-Triv) will be of absolutely no use in the project of explaining some problematic form of reference. For the righthand side of (T-Triv)—unlike that of (T-Red)—involves precisely the sort of reference that we are attempting to explain. Faced with this choice between the reductive truth-condition (T-Red), which allows the explanatory project to progress, and the trivial one (TTriv), on which the explanation cannot even get started, it is of course tempting to insist that it is the former that gives the meaning of the identity statement, and that if this conflicts with the compositionality of meaning, then so much the worse for this principle of compositionality. This appears to have been Frege’s view in Grundlagen, where he talks about the righthand side being a “recarving” of the meaning of the left-hand side.29 This “recarving thesis” is explicitly endorsed by prominent contemporary defenders of Fregean ideas about reference, such as Bob Hale and Crispin Wright.30 However, I have insisted throughout this paper that Frege’s proposal is of meta-semantic nature. I am therefore under no pressure to say that (T-Red) gives the meaning of an identity statement. I can instead maintain that the 29 30

See Frege 1884, §64. See Hale 1997 and Wright 1997.

276 only semantically generated truth-condition for an identity statement is the trivial one (T-Triv). What Frege proposes is rather an account of what a subject’s understanding of an identity statement consists in. And as we have seen, this account involves the principle for the identity of their semantic values: (SV)

[a] = [b] ↔ u ≈ v

When this principle (SV) is combined with the trivial truth-condition (TTriv), we do indeed get the reductive one (T-Red), which now emerges as a hybrid of semantic and meta-semantic facts. 3.3. The rigidity thesis A final worry concerns the compatibility of my Fregean theory of reference with the semantic thesis that names and their mental counterparts are rigid designators. Consider a representation a associated with a referential attempt 〈u, ≈〉. Let f≈ be the function determined from ≈ in accordance with the Principle of Individuation (PI). I have argued that a refers, if at all, to the object f≈(u). One may then wonder whether my view doesn’t collapse back into some version of the descriptivist view of names criticized by Kripke and Føllesdal. Specifically, am I not committed to identifying the meaning of a with that of the description “the f≈ of u,” with the result that a isn’t a rigid designator after all? (For instance, one chunk of physical matter may be part of different bodies in different possible worlds.) The above discussion provides the resources needed to respond to this worry. On my proposal, the nature of the function-argument structure f≈(u) is entirely meta-semantic, not semantic. The expression a is semantically simple, and its semantic value, if any, is just the object f≈(u). How this referent is determined is a meta-semantic matter, of no immediate semantic significance. As far as semantics is concerned, a is a simple term or representation whose semantic value is just an object. More generally, not every kind of structure involved in the phenomenon of reference is semantic structure. For instance, reference is often based on perception, and perception is undoubtedly a complicated process that involves all kinds of struc-

277 turing of sensory information. But this structure will generally not be semantic structure. Although perception is often presupposed by the relation of reference and thus also by semantics, perception and its structure aren’t thereby included in semantics. My claim that the function-argument structure f≈(u) isn’t semantic structure enjoys independent evidence as well. Semantic structure is by and large accessible to consciousness; otherwise we wouldn’t know or be rationally responsible for what we say and think. But someone can understand reference to shapes and bodies without having any conscious knowledge of how such reference is structured. Someone’s competence with this structure may be located entirely at a “subpersonal” level, much as the structuring involved in perception is. This is evidence that the function-argument structure f≈(u) isn’t semantic. And if that is right, then my account will be fully compatible with the rigidity thesis and in no danger of collapsing back into descriptivism. My conclusion is thus that my Fregean theory of reference manages to combine the semantic theses of rigidity and compositionality with Føllesdal’s meta-semantic desiderata. Although my theory is much more limited in scope than what Føllesdal ultimately wants, this is at least a first step in that direction. Moreover, this establishes that there is no inherent conflict between a semantic theory that takes singular terms (or their mental counterparts) to be rigid designators and a meta-semantic theory that satisfies Føllesdal’s five Frege-inspired desiderata. Although Føllesdal (like Kripke) takes himself to break radically with Frege on the semantics of singular terms, there is room for extensive agreement with Frege (and disagreement with Kripke) on the meta-semantics of singular terms.31

31

Thanks to Anthony Everett for comments on an earlier version of this paper and to the participants at the Lauener Prize Symposium for discussion. I am particularly grateful to Føllesdal for extensive discussions of the ideas of this paper, both at the symposium and after.

278

References Dummett, Michael. 1981. Frege: Philosophy of Language, 2nd ed. (Cambridge, MA: Harvard University Press) Evans, Gareth. 1982. Varieties of Reference (Oxford: Oxford University Press) Føllesdal, Dagfinn. 1986. “Essentialism and Reference,” in L.E. Hahn and P.A. Schilpp (eds.), The Philosophy of W.V. Quine (La Salle, IL: Open Court), pp. 97-113 ----------------------. 1997. “Conceptual Change and Reference,” in C. Hubig (ed.), Cognitio Humana: Dynamik des Wissens und der Werte (Berlin: Akademie Verlag), pp. 351-367 ----------------------. 2004. Referential Opacity and Modal Logic (New York and London: Routledge) Frege, Gottlob. 1884. Foundations of Arithmetic. Transl. J.L. Austin (Oxford: Blackwell, 1953) Geach, Peter. 1969. “The Perils of Pauline,” Review of Metaphysics 23, pp. 287-300 Hale, Bob. 1997. “Grundlagen §64,” Proceedings of the Aristotelian Society 97(3), pp. 243-261; repr. with a postscript in Hale and Wright 2001 ------------ and Crispin Wright. 2001. Reason’s Proper Study (Oxford: Clarendon) Linnebo, Øystein. 2004. “Frege’s Proof of Referentiality,” Notre Dame Journal of Formal Logic 45(2), pp. 73-98 ---------------------. 2005. “To Be Is to be an F,” Dialectica 59(2), pp. 201-222 ---------------------. 2006. “Sets, Properties, and Unrestricted Quantification,” in A. Rayo and G. Uzquiano (eds.), Absolute Generality (Oxford: Clarendon) ---------------------. 2009. “Frege’s Context Principle and Reference to the Natural Numbers,” in S. Lindström et al. (eds.), Logicism, Intuitionism, and Formalism: What Has Become of Them? (Springer), pp. 47-68

279 McDowell, John. 1977. “On the Sense and Reference of a Proper Name,” Mind 86(342), pp. 159-195 Parsons, Charles. 1971. “Ontology and Mathematics,” Philosophical Review 80, pp. 151-176; repr. in his Mathematics in Philosophy (Ithaca, NY: Cornell University Press) Stalnaker, Robert. 1997. “Reference and Necessity,” in B. Hale and C. Wright (eds.), Blackwell Companion to the Philosophy of Language (Oxford: Blackwell), pp. 534-554; repr. in his Ways of World Might Be (Oxford: Clarendon, 2003) Wright, Crispin. 1997. “The Philosophical Significance of Frege’s Theorem,” in R. Heck (ed.), Language, Thought, and Logic: Essays in Honour of Michael Dummett (Oxford: Clarendon); repr. in Hale and Wright 2001

Comments on the Essays

Comments on the Essays Dagfinn Føllesdal∗

Five of the participants take up issues relating to phenomenology. Rather than commenting on them separately, I will give a brief sketch of Husserl’s phenomenology and bring in and discuss the various papers where they fit into and enrich this sketch. Thereafter I will comment on the remaining essays one by one.

PHENOMENOLOGY

Comments on Suppes, Smith, Beyer, Friedman and De Pierris Husserl’s phenomenology is largely a study of consciousness. Husserl’s key idea is that of intentionality: our consciousness is structured in such a way that it is as of an object. Perception is particularly well suited to illustrate the idea. There are here three factors involved: hyle, noesis and noema. Husserl studied these factors by reflection upon his own consciousness. Suppes in his contribution to this symposium finds that Husserl’s careful reflections fit in quite well with empirical findings in the study of the brain and nervous system. (See also Suppes’s studies in the psychology of perception, some of which are listed in the bibliography at the end of these comments.) It would seem that phenomenology and these empirical studies should go hand in hand. Detailed and careful phenomenological analyses of the kind Husserl carries out may guide brain scientists in what to look for, and brain science may make us aware of phenomena that ought to be traceable in our consciousness. Husserl’s notion of “hyle” plays an important role in perception, but although Husserl often calls the hyle “sense data” (Sinnesdaten), they are ∗

Stanford University and Universitetet i Oslo

284

not data or other kinds of objects of experience. They are experiences that we typically have when our sense organs are “impinged upon,” to use Quine’s expression. We sometimes have hyle also when nothing happens at our sensory organs, for example, when we have fever or are affected by drugs, etc. Suppes has studied in detail the psychology of perception and the sensory experiences that are involved in perception (see the list of references at the end of my comments). In order for us to perceive, the hyletic experiences have to be combined with noetic, or meaning-giving experiences, which structure our experiences so that they are as of a physical object. The noeses are, like the hyle, experiences, they are “anticipations” of there being an object that we are experiencing, The noeses structures our experience around a “determinable X,” That is, as we noted, our act is “as of” an object that has lots of properties, bears relations to other objects and to us, etc. The noeses are thereby “meaning-giving” experiences, according to Husserl; thanks to them, our experiences are integrated into a structure. They make up an act, which is directed towards an object. We perceive an object, which has properties, bears relations to other objects, etc. There are therefore no hyle that can be reidentified from one act to another, independent of the noeses. This is a contested point in Husserl interpretation. Prominent scholars, like Aron Gurwitsch, have interpreted Husserl as if he held that hyle can be reidentified between acts with different noeses, and many have followed him in this.1 However, the textual evidence against that view is strong, and also systematic considerations count in favor of the hyle always being context-dependent. For Husserl there is not something unstructured “given” in perception, no intermediaries of the kind appealed to by sense datum theorists. The pattern of anticipations is tremendously rich, most of the anticipations we never think of. This is important. As Suppes points out in his paper, we could not survive if we were to attend to all our “anticipations.” The word ‘anticipations’ is clearly not appropriate here. ‘Association’ may 1

Gurwitsch, A. (1966), p. 256. That hyle cannot be reidentified from act to act is stressed by Husserl several places, for example in Analysen zur passiven Synthesis Husserl, E. (1918-1926), Husserliana XI, p. 363.

285

be used, likewise ‘setting’. ‘Setting’ has fewer overtones of something we are aware of, and also fits many of the practical and bodily, non-cognitive features that shape what we experience. Whatever we call it, it is largely a result of sedimentations of past experience. The objects of our acts are similarly rich. The features that we anticipate that they have, are so many and varied that they go far beyond what we can attend to or be aware of. And the object itself is anticipated as being even richer than this. We anticipate that the object has very many features that we do not anticipate. Some of them we may come to anticipate, others we will never anticipate. Each object is in this way inexhaustible by our experience. It transcends our experience, Husserl said. This notion of transcendence must be distinguished from the notion of transcendental. The transcendental is a major topic in Husserl. Briefly, the transcendental is what we get aware of when we reflect on our own consciousness and study its structures. We are then not studying the objects, as we do in our natural, object-directed attitude. We study instead how our consciousness “constitutes” objects, to use Husserl’s phrase. One particular notion that is central for understanding the transcendental and the notion of constitution and thereby Husserl’s idealism, is the thetic component of consciousness. This is the component in the noesis that has to do with our conception of the ontological status of the object, whether we experience it as real, as we do in perception, or merely fictional, etc. Since none of the five participants who focus on phenomenology goes into Husserl’s idealism and the thetic component, I refer to my own writings on this, in particular my article “Husserl’s Idealism”2 Another important component in the noesis that I briefly mentioned above is that the consciousness is my consciousness. David Smith in his paper gives an excellent overview and discussion of Husserl’s view on this point and alludes to Sartre’s early article “Conscience de soi et connaissance de soi,”3 which I find so helpful that I have always assigned it for my courses on Sartre from 1962 on, although until 1967 it was available only for students who could read French. 2 3

Follesdal, D. (1998). Sartre, J.-P. (1948).

286

Smith also notes Husserl’s systematic use of quotation marks. Neglecting them has led many commentators astray in their interpretation of Husserl. Most importantly, Smith gives a very fine presentation and discussion of the indexical element in consciousness, which is very central in Husserl and which anticipates contemporary work, particularly by David Kaplan and especially by John Perry, who is here with us, and who has contributed greatly to our understanding of this very interesting and important realm This brings me to Christian Beyer, who like Smith sets forth an approach to the problems of indexicals and reference and also the twin world problem that combines insights from Frege, Husserl and Perry. However, while Smith concentrates on the noesis, Beyer focuses on the noema, the third important notion in Husserl’s discussion of consciousness beside the hyle and the noesis. Briefly, the noema is the abstract structure of the noesis. Everything that I just said about the noesis is therefore reflected in the noema: the act’s thetic character, the structural elements that make the act an act of this ego, and the object-directedness, which Husserl discusses under the label the determinable X, and which is crucially connected with the indexicality of the act. All of this is clearly and precisely presented by Beyer. In particular, his discussion of the determinable X and Husserl’s twin world example illuminates very well these complex issues. Beyer and Smith demonstrate in their contributions how important it is to be familiar with the systematic issues in philosophy when one works in the history of philosophy and when one interprets other philosophers. In the case of Husserl, this is particularly important, since Husserl was so careful and penetrating in his analyses that he anticipated much that has come later. Unfortunately, there has been a tendency to oppose “continental philosophy” to ”analytic philosophy,” which has dissuaded philosophers from reading one another’s work. These labels have tended to close people’s minds. They have deterred many philosophers from reading Husserl, and they have led many who read Husserl to stay ignorant about similar developments in contemporary philosophy of mind, philosophy of language, and epistemology, particularly perception. Fortunately, both Beyer and Smith and many others are open and informed and avoid the use of these polarizing labels. Smith has contributed to make University of Cali-

287

fornia, Irvine, where he serves a chair, to a place I happily recommend to students of Husserl. And Göttingen has become a similar place in Europe. I have strong emotional ties to Göttingen, where I studied in the fifties, and I am very happy that Beyer now has a chair there. In several articles and books he has added greatly to our understanding of Husserl and the philosophical issues he discusses, and he contributes to maintaining the high standards of scholarship and insight that Günther Patzig and others have established in Göttingen. I am very happy that Irvine and Göttingen have overcome the barriers that have been hampering Husserl scholarship. As an example of the bad effect of these barriers Beyer mentions the discussion of my eight page paper on the noema in The Journal of Philosophy 1969, where I bring in the “analytic” philosopher Frege to help understand Husserl, the originator of “continental” philosophy. This article has been followed up by many, but attacked by even more. In 1983, Ethel M. Kersey published a bibliography of the discussion up till then as a whole volume, volume 9 in Philosophy Research Archives.4 Further additions have come in 1982, in Husserl, Intentionality, and Cognitive Science,5 and in Drummond and Embree’s The Phenomenology of the Noema.6 I have read only a very small fraction of these hundreds of papers, the ones that display careful reading of Husserl’s texts and awareness of the philosophical issues and which thereby bring Husserl scholarship forward. Unfortunately, a too large number of these articles attack views that I have never had. The authors of these papers cannot be expected to read Husserl’s texts any better than they read mine, and I therefore quickly turn to other authors. Examples of philosophers who cannot read a text properly are all those commentators who allege that I maintain that Husserl took over the sense-reference distinction from Frege.7 I have never uttered any4

Kersey, E. (1983). Dreyfus, H.L., and H. Hall (1982). 6 Drummond, J. and L. Embree (1992). 7 An example is Guillermo E. Rosado Haddock, who repeatedly has criticized me for this and other views I never had. In a recent article, “Husserl pour les philosophes analytiques” (Philosophiques 37 (2010), 325-348), he presents as his own a view on Husserl very like mine, but without mentioning me. He then writes “Il est donc com5

288

thing like that. On the contrary I have in several places stated that Husserl knew this distinction from several other sources long before 1892, when Frege wrote his famous article. In fact, when I first taught a course on Husserl, at Harvard in 1961-62, I started the course with seven lectures on Bolzano, on this and several other important notions and distinctions in Bolzano that help throw light on Husserl’s phenomenology. The reason I made use of Frege to explain Husserl’s notion of noema in 1969, was that Frege was better known than Bolzano to the readers of Journal of Philosophy. Also, as Beyer brings out well, there are many insights in Frege that point to ways of overcoming difficulties that turn up, for example in connection with indexicals and reference. However, often these subtleties in Frege, and even more in Husserl, go unnoticed by interpreters of Husserl who are not well versed in the so called ”analytic” discussion, to which not only Frege, but also Husserl were important contributors. In his contribution to this symposium Beyer rebuts a number of the criticisms that have been directed against my view. Also David Smith, Ronald McIntyre, Jaakko Hintikka, Charles Parsons, Kevin Mulligan, Denis Fisette, Hubert Dreyfus, Jitendra Mohanty and several others have contributed to set things straight. I will here mention only one objection that comes back in all the attacks I have read. This is the claim that according to Husserl the noema is the object of the act, experienced from some particular view, or without some of its properties. There are serious problems with this claim, which those who put it forth do not seem to be aware of. One is that it goes against a multitude of Husserlian text, texts that are often misread because one does not pay attention to Husserl’s use of quotation marks. However, the simplest refutation of the identification of the noema and the object is to note that according to Husserl there are acts which have no object, although, like all acts they have a noema. How can then the noema be the same as the object? Husserl had several valuable and interesting insights on how to deal with

plètement injustifié et faux de maintenir que Husserl a repris de Frege la distinction entre sens et référence (en la généralisant à la distinction noético-noématique), comme l’ont répété plusieurs philosophes analytiques sans avoir lu Husserl.” (p. 327).

289

acts without an object. These all get lost when one identifies the noema with the act’s object. The authors who should be taken seriously are those who seek to find passages in Husserl that go against my view. If my interpretation shall be reliable, it must fit in not only with the passages I quote in the paper, but with all other passages in Husserls 40 000 pages Nachlass. I have spent far more time reading Husserl than reading any other philosopher. Working my way through the Nachlass I have picked out about 30 000 passages that I think are particularly helpful for understanding all the different aspects of Husserl’s philosophy. I can of course never include all these passages in what I write, but I try to include two kinds of passages: those that express particularly clearly the view that I attribute to Husserl, and those that seem to go against my interpretation and therefore have to be defused, for example by noting that Husserl in the margin writes that the passage has to be changed. Now back to the universities that are good places for studying Husserl. I mentioned Göttingen and Irvine. At Stanford, too, the interest in Husserl has been growing. I enjoy having several colleagues working on phenomenology. John Perry has found that Husserl’s ideas about consciousness and subjectivity are very similar to his own, and also that Husserl had thoughts on indexicality that were far ahead of those of Frege and others of his contemporaries. We have taught seminars together both on subjectivity and on philosophy of language. In 2000 Michael Friedman came to Stanford and has engaged himself strongly in the study of Husserl and especially his relation to Kant and the Marburg school of neoKantians, notably Cassirer. So has Thomas Ryckman, who is unfortunately not here. His book The Reign of Relativity,8 which just came out, gives remarkable insights in the contributions of neo-Kantians and of Husserl and Hermann Weyl to our understanding of the foundations of relativity theory and also other developments in contemporary physics. Also, Graciela De Pierris has presented a paper, which I will comment on shortly, connecting Hume and Husserl. Finally, I would like to mention Lanier Anderson, who works on Kant, Nietzsche, Husserl and the existentialists. 8

Ryckman, T. (2005).

290

I am very happy for the stimulation that all these colleagues give me. Friedman’s paper is a fine example. He suggests a way of combining the insights of the neo-Kantian, logical empiricist, and phenomenological approaches to foundational issues in science. He then formulates a direct question to me: Do I agree that Husserlian transcendental phenomenology needs supplementation by ideas characteristic of the neo-Kantian tradition in order adequately to ground the notion of an “essential analysis”? The relation between Husserl and Kant is complex and changed over time. Husserl was in regular contact with several of the neo-Kantians, in particular the five year older Paul Natorp, with whom he corresponded for thirty years, from when Husserl was a young Privatdozent in Halle until Natorp’s death in 1924. There are also a few letters between Husserl and Cassirer. In one of these, from 1925, Husserl wrote: My own development, [was] originally Kant-hostile, but admittedly also unreceptive for the true sense of the Kantian philosophy … However, when the foundational problems in science, which were nearest at hand for me as mathematician, drove me to ever new ones in necessary consequence …and I broke through to the method of eidetic analysis of consciousness, and when the phenomenological reduction opened for me the realm of the primary source of all knowledge, I had to admit that the insight that developed in me by an essentially different method comprised the total Kantian problematic (which first now received a deep and clear sense) and confirmed Kant’s main results in a strict scientific justification and limitation. After having learned to see Kant from my own perspective, I can also – especially during the latest years – be taught richly from Kant and the true Kantians.9 9

„Meine eigene Entwicklung, die ursprünglich kantfeindlich war, aber freilich auch für den eigentlichen Sinn der Kant‘schen Philosophie unempfänglich … Als ich aber von den mir als Mathematiker nächstliegenden wissenschaftstheoretischen Grundproblemen zu immer neuen in notwendiger Konsequenz fortgetrieben … zur Methode einer eidetischen Bewusstseinanalyse durchdrang, und als sich mir mit der phänomenologischen Reduktion das Reich der Urquellen aller Erkenntnis eröffnete, da musste ich erkennen, dass die mir zuwachsende Wissenschaft bei wesentlich andersartiger Methode, die gesamte Kant’sche Problematik umspannte (die nun erst einen tiefen und klaren Sinn empfing), und dass sie Kants Hauptergebnisse in streng wissenschaftlicher Begründung und Begrenzung bestätigte. … Nachdem ich Kant unter eigener Perspektive sehen gelernt habe, kann ich auch – so recht erst in den allerletzten Jahren – von

291

I will not go into a full comparison of Kant and Husserl. This has been excellently done in the book Husserl und Kant10 by Iso Kern, who taught and lives here in Bern. Instead I will give a brief sketch of the elements in Husserl’s phenomenology that are especially pertinent to Friedman’s discussion, in particular the notions of perception, sedimentation, life-world, constitution, transcendental, idealism, objectivity, and a priori. In particular I will discuss from a Husserlian point of view Friedman’s key notion of the relative a priori, and the related need to focus on features that make it possible to test our theory through measurement and observation. The difference between Husserl and most other philosophers, including Kant, begins already with perception. As Friedman points out, “there is no fundamental dualism between sensibility and intellect in Husserl’s picture, and unlike Cassirer … the sensible and intuitive dimensions of experience are placed at the center rather than at the periphery.” There are for Husserl no empirical “givens” with which our abstract theories have to be “coordinated.” As we noted earlier, intentionality and structuring goes “all the way down.” The structure imposed by the noeses is part of an all-comprehensive structure, which we give to the world we live in. The most familiar part of this is the “life-world.” This is a theme in Husserl that particularly engages Friedman. A common misinterpretation of Husserl, which Friedman does not fall prey to, is to regard the life-world as a purely human and social world, a polar opposite to the scientific world. As we could expect from what has already been said about the all-comprehensive structure we give the world, there is an intimate connection between the life-world and science. I have written on this elsewhere11 and quoted numerous passages from Husserl’s writings to back up this interpretation. I will therefore here just mention briefly the three main ways in which the life-world and science are connected: Kant und den echten Kantianern reiche Belehrungen empfangen.“ (1994a), p. 4. 10 Kern, I. (1964). 11 Follesdal, D. (1990).

Husserl, E.

292

1. The sciences are part of the lifeworld. “The scientific world – [the subject matter of] scientific theory – ... like all the worlds of ends ‘belongs’ to the lifeworld.”12 2. Scientific statements get their meaning by being embedded in the lifeworld. This was stressed by Husserl already in 1915, when he was developing what he later called the “lifeworld”: All theory relates to this immediate givenness and can have a legitimate sense only when it forms thoughts which do not offend against the general sense of the immediately given. No theorizing may offend against this sense.”13

3. The sciences are justified through the lifeworld. Our notion of reality originates in our experience of the lifeworld, and when we ascribe reality to the world described by science, this is because the world described by science is regarded as being the same world as the one that we experience as the lifeworld: Though the peculiar accomplishment of our modern objective science may still not be understood, nothing changes the fact that it is a validity for the lifeworld, rising out of particular activities, and that it belongs itself to the concreteness of the life-world.14

Husserl is a typical holist. Not only are the sciences and the lifeworld part of an all-embracing web. The unity of this web gives validity to all parts of the web: A judgment-unity penetrates all the individual judgments ... they have a unity which builds itself up in the progression of judgment, tying together judgmentsense with judgment-sense. This unity confers on all of them an intrinsic, interrelated validity. In this way the multiple statements in a treatise have a com-

12

Husserl, E. (1936), Beilage XVII, Husserliana VI,460.23-27 = Carr, p. 380. Husserl, E. (1905-1920). Husserliana XIII,196.22-34, my emphasis. 14 Husserl, E. (1936), § 34f, Husserliana VI,136.18-22 = Carr, p. 133. 13

293

prehensive judgment-unity, and so has in its way every theory and every entire science.15

The life-world is for Husserl an ultimate court of appeal, behind which there is no point in asking for further justification. The main reason Husserl gives for this, is that most of the life-world consists of acceptances that we have never made thematic to ourselves and which have therefore never been the subject of any explicit judicative decision.16 Husserl calls himself a transcendental philosopher. However, this means for Husserl that he is bracketing our normal concern with objects of acts, and is instead focusing on our consciousness itself. What are, for example, the features of consciousness that distinguish an act of perception from an act of imagining? What in our consciousness marks the difference between experiencing something as real rather than illusory? What is involved in the notion of an object? How do we experience others? In what way is perception “communal”? Through these analyses Husserl seeks to answer the skeptic and the solipsist by arguing that in trying to express their position they are undercutting it, they are sawing off the branch they are sitting on. “Transcendental” philosophy is a study of consciousness in all its various forms. Husserl is not claiming any kind of infallibility for his transcendental philosophy. He is a fallibilist. He asks: “Can I begin with a truth — a definitive truth? ... If I already had such 'immediately selfevident' truths, then I could perhaps mediately derive new ones from them. But where do I have them?”17 In Formal and Transcendental Logic he writes: “Even an ostensibly apodictic evidence can become disclosed as deception.”18 Husserl’s transcendental philosophy is therefore quite unlike philosophies that claim certainty. Also, he is not searching for necessary conditions for something to be possible. He is describing consciousness, an enterprise that is fallible like all other attempts to get insight. 15

Husserl, E. (1923/4), 3. Vorlesung, Husserliana VII,19.24-36. Husserl, E. (1938) § 67, p. 330 = Churchill & Ameriks, p. 275. 17 Husserl, E. (1936) § 73, Husserliana VI, 269.24-29 = Carr, 335. 18 Husserl, E. (1929) § 58, Husserliana XVII, 164.32-34 = Cairns, 156. 16

294

Husserl uses the term ‘a priori’. But he uses it in much the same way it is used by C.I. Lewis, as something that is anticipated, but may go wrong. To use Lewis’ example: I expect to open the door by turning the doorknob, but somebody has replaced it by a piece of soft rubber that just yields when I twist it. My a priori expectation breaks down. We always concentrate on some issue or task and take a background for given. Most of this background we are not even aware of, but we may become aware of it when something breaks down. What is given as background and what we are trying to do or find out may vary from situation to situation. What is background in one situation may be attended to in another, and conversely. As I have argued in an article where I compare Husserl and Wittgenstein, one important difference between the two is that Husserl held that there is no individual statement that is held fixed and is the basis for certainty.19 Wittgenstein, however, often appears to think that there are such fix-points. Here is one among the many passages of his that point in this direction: That is to say, the questions that we raise and our doubts depend on the fact that some propositions are exempt from doubt, as if they were like hinges on which they turn.20

Are the hinges always the same, or are the hinges in one case doubted in another? That is to say, it belongs to the logic of our scientific investigations that certain things are in deed not doubted.21

Again, are there some things that are never doubted, or is Wittgenstein just making the point that at any time some thing or other is not doubted? The order of quantifiers is important. Husserl is very clear on this: Whichever issue we are investigating, some other issues are presupposed but not con-

19 20 21

Follesdal, D, (2005). Wittgenstein, L. (1969), # 341. Wittgenstein, L. (1969), # 342

295

sidered in that investigation, but there is no fixed set of issues that are always presupposed. Friedman’s idea of the relative a priori seems to come down somewhere between Husserl and Wittgenstein. In Newton’s physics inertial frames of reference play a special privileged role. They provide us with situations where we can measure and relate our sensible perceptual experiences to Newton’s theory. By moving the frames out from the surface of the earth to the center of the solar system we can get steadily better fit between our experiences and the theory. Friedman then suggest an extension of these ideas to general relativity, where inertial reference frames no longer exist, but where we can get local inertial frames which approximate, in small regions and under appropriate conditions, the global inertial frames of our earlier theories. To return now to Friedman’s question: Do I agree that Husserlian transcendental phenomenology needs supplementation by ideas characteristic of the neo-Kantian tradition in order adequately to ground the notion of an “essential analysis”? My answer is that many of the salient features of the neo-Kantian tradition are found in Husserl. Certainly, Friedman’s idea of inertial frames and the relative a priori can fit in well with Husserl’s transcendental phenomenology, where, as Friedman mentions, one moves from our perceptual experience through idealizations, approximations and layers of sedimentation to abstract mathematical scientific theories. However, Husserl is very clear about the order of the quantifiers: in every inquiry we take something for given – often without noticing it – and depend on it when we explore with the rest. However, there is not something that is always taken for given and can never be overturned. I am in full sympathy with Husserl, Quine and many others who are holists and also fallibilists. There is no belief of ours that is guaranteed to be true. However, this whole web of beliefs and their background is structured. As Quine and Husserl both point out, some beliefs serve as grounds for others; many of them are pivotal. They may be pivotal in the sense that if they are given up this will require comprehensive revisions. Others are pivotal for our being able to make measurements or in other ways test our theory. However, getting back to the difference between Husserl and Witt-

296

genstein, let us be open for the view that as scientific theories develop, the crucial features that make it possible for us to relate our sensible perceptual experiences to the shifting theories need not always be the same. Friedman has pointed out how the inertial frames in Newton’s theory have no direct counterpart in general relativity. So if he agrees with Husserl on the order of quantifiers, as I think Friedman does when he talks about the “relative a priori” then we all agree: for every theory there will be some features that enable us to connect it with our sensory experience. Finally, there is Graciela De Pierris’ contribution. I agree with De Pierris that it may be helpful to use phenomenology as a help to interpreting Hume. Brentano encouraged his students to study the British empiricists, Locke, Berkeley, Hume and Mill, and Husserl came to appreciate Hume greatly. According to Husserl, Hume’s skepticism, correctly understood, is nothing short of a “transcendental phenomenology.”22 And in the “Nachwort zu meinen ‘Ideen ...’ ” (1930), Husserl describes Hume’s Treatise as “der erste Entwurf einer geschlossenen Phänomenologie”23 As De Pierris points out, Hume’s view that neither space nor time can be infinitely divisible has been widely criticized. Antony Flew and Norman Kemp Smith are two notable critics. It is always recommendable when one interprets a person’s views to try to make it as plausible as possible. This is what De Pierris does here. I find her phenomenological approach both sensible and interesting. In some of his first lectures, in November 188724, Husserl praises Hume for distinguishing between empirical science and mathematics, as opposed to, for example Mill, who regarded mathematics as based upon and continuous with empirical science. As Husserl’s view developed, especially after 1913, he came to see more connections between our sensory experiences and mathematics, and he was no longer so satisfied with Hume’s position. However, the distinction that De Pierris emphasizes, between our acts and their components – which Hume was concerned with – and the objects and their components, was one that

22

Husserl, E. (1929), Cairn’s translation, pp. 226-227. Husserl, E. (1930), Husserliana V, 155. 24 Husserl, E. (1886-1901), Ergänzende Texte, Husserliana XXI, pp.228-233. 23

297

Husserl always held on to, although he later saw them as more connected with one another than he did in 1887. By the way, Quine in his lectures on David Hume’s philosophy in 25 1946 has a discussion of Hume’s view that neither space nor time can be infinitely divisible. Quine does his best to interpret Hume in a sympathetic way, and his discussion of this point in Hume is therefore more in the spirit of De Pierris, and in my view better and more detailed than the later discussions of Flew and Kemp Smith which De Pierris takes issue with.

SCIENCE

Comment on Essler Essler takes up a cluster of issues connected with the conception of space. These are central in the philosophy of physics, especially in connection with relativity theory. However, they first came up in connection with the discovery of non-Euclidean geometry and its relation to the Kantian view on space. These issues continue to be a challenge to Kantians, but are of concern to every philosopher – understanding space and time has always been central to philosophy. The particular issues that Essler brings up in his comments on my paper “Relativity, Rotation and Rigidity” are concentrated on Riemann, Helmholtz and Lie’s work on geometry and space. Their ideas are important and have been very much discussed. Michael Friedman, who is here, is a main contributor to this discussion, and another Stanford colleague, Thomas Ryckman, gives an excellent overview and analysis of the situation in his recent book The Reign of Relativity26, which I mentioned earlier and will strongly recommend.

25

Included in Quine, W.V. (2008a), pp. 36-136. Quine’s discussion of this point is found on pp. 80-88. 26 Ryckman, T. (2005).

298

In his contribution Essler asks about Husserl’s view on these matters. Husserl took up these issues quite early. In 1889-90 he gave a series of lectures “Geschichtlicher Überblick über die Entwicklung der Geometrie” where he discusses Riemann’s geometry and raises objections against this and all other attempts to clarify the logical foundations of geometry through a general theory of curvature. This includes Gauß and Helmholtz. Later, he took a more positive attitude to these approaches. He discusses them briefly in Formal and Transcendental Logic (1929), particularly in §30. Hermann Weyl, who was strongly influenced by Husserl, partly through his wife, who studied with Husserl, seems to have been an eye opener for Husserl. In a letter to Husserl on March 26, 1921, Weyl writes: Recently I have endeavored to understand the essence of space from the last grounds that are accessible through mathematical analysis. This has to do with group theoretical considerations similar to those which Helmholtz and Lie carried out in relation to the problem of space. However, today one must take into account that the situation has been changed through the theory of relativity, which has made it possible to deepen the foundations. Through this, the a priori character of space, which comes to expression through the uniform nature of the metric in every spatial position, distinguishes itself from the a posteriori, which in itself changes freely, depending on the reciprocal ordering in nature of these metrics in the different points. Just thereby, that one presupposes that this varies freely and not assumes that the Euclidean distance geometry is rigidly given, does it become understandable why the nature of the metric is what it really is and not one of the infinitely many others that are mathematically possible.27

Husserl did not respond immediately to this letter, but a year later he writes to Weyl that …Einstein’s theories, but only when they are supplemented and founded on your infinitesimal-geometrical research, depict that form of “structurelawfulness” of nature (as opposed to the specific “causal” lawfulness) which is required as necessary by the deepest transcendental-philosophical grounds …28 27 28

Husserl, E. (1994b), p. 291. Ibid., pp. 293-294.

299

These are important philosophical problems in physics, and I am happy to have so many friends who are doing stimulating work on these issues, three of them here: Essler, Suppes and Friedman.

Comment on Roll-Hansen Nils Roll-Hansen takes up a complex issue: the mutual dependence of science and democracy. This has been the subject of two books by outstanding philosophers of science: my Stanford colleague Helen Longino’s The Fate of Knowledge29 and Philip Kitcher, Science, Truth and Democracy.30 The authors were invited by the journal Philosophy of Science to review one another’s books and respond to one another’s reviews. Roll-Hansen starts out from their contributions and carries this discussion further. Roll-Hansen has a very good background for dealing with these issues. He was doing research in biology before he turned to philosophy of science, and he has also worked for years on questions of science policy in what is now called Nordic Institute for Studies of Innovation, Research and Education. In addition, he has also done important research in the history of biology, with books and articles both on the history of eugenics31 and on reductionism in biological research. His most notable book, The Lysenko Effect: The Politics Of Science32, was praised in Nature (25 August 2005), and for good reasons. Roll-Hansen combines the competencies I have mentioned with fluency in Russian, and he was one of the first to work in the Russian archives when they were opened. What sets Roll-Hansen’s discussion apart from Kitcher’s and most other science studies is that Roll-Hansen carries out careful studies in the history of science to test hypotheses about the interplay between science and society. Kitcher asserts that modern biological science has supported discrimination and social repression and that this is particularly true of ge29

Longino, H. (2002). Kitcher, P. (2002). 31 Roll-Hansen, N. (2005). 32 Roll-Hansen, N. (2004). 30

300 netics since its formation around the turn of the 19th century. Roll-Hansen quotes him as claiming that “a wealth of historical studies hammers home the same moral.” Here Roll-Hansen, the professional historian comes in. He points to his own study of eugenics in Scandinavia which shows that eugenics here was motivated by social concerns over women’s emancipation and improved living conditions for children rather than by eugenic visions of a genetically better race. This shows that one should be wary of sweeping generalizations, especially in a field where the development depends on an interplay of very many factors. In such cases it is advisable to look at the details in order to see better how the various factors function. Since this is not only a philosophical meeting, but also a gathering of friends, let me report an episode from Stanford many years ago, which in particular may interest my many colleagues here from Stanford. Our colleague Georg Kreisel knew well William Shockley, who was awarded the 1956 Nobel Prize in Physics for his role in the invention of the transistor. Kreisel liked social experiments, and he invited me for lunch with Shockley. The first Shockley did when we sat down at the table, was to take up a tape recorder, saying that he hoped that I would not mind his recording our conversation – he had many sad experiences with people claiming that he had said things that he had absolutely not said. Like many Nobel Prize winners, Shockley wanted to branch out to other fields. In particular, he wanted to teach a course and carry out a scientific study of blacks and IQ. He was convinced that blacks have a lower IQ than other races, and at the annual convention of the American Psychological Association in 1979 he proposed a plan for raising the average IQ in the United States by paying out a bonus to blacks who let themselves be sterilized. Departments do not normally interfere with the contents of courses their faculty wants to teach. However, the physics department solved its problem by deciding that this course belonged in the psychology department, and the psychology department found Shockley’s research and teaching projects methodologically deficient and therefore declined to invite him to teach such a course.

301

What then was wrong with Shockley’s project? I will mention only some features, which are relevant to the issues Roll-Hansen discusses. First: why did Shockley single out skin color as one of the factors he was interested in? One main reason why both the physics department and the psychology department as well as Stanford as a whole was unhappy with Shockley’s proposal, was that skin color has been and still is a major factor of discrimination not just in the United States, but all over the world. Research focusing on skin color would probably intensify discrimination. However, as Roll-Hansen asks: why should potential bad effects of research stop us from carrying it out? Is not increased insight and understanding always of value? However, from a purely methodological point of view, skin color seems to be a very unpromising factor to focus on. One knows that skin color, like eye color, depends on a very minute part of the human genome, that does not seem to have anything to do with genes that affect important human traits. Why then, was skin color emphasized so strongly by Shockley and his allies, such as Arthur Jensen and Richard Herrnstein? One knows from studies of discrimination that visible features play a particularly important role in prejudices. They are easy to spot and distract attention from other features. A second reason for Shockley’s prejudice against blacks is that simple statistics actually shows that blacks actually score less than whites on IQ tests. Isn’t this exactly what he wanted to show even more convincingly through his research project? There may be other explanations for the low score of many blacks, that Shockley seems not to have thought about, or at least not mentioned. Most blacks in the US grow up in poverty and go to bad schools – given that schools in the US are not centrally financed, but financed by the local school district, something that contributes strongly to the status quo. There is a very strong correlation between the kind of schools one goes to and how one scores on IQ tests. If Shockley had decided to study the connection between school quality and IQ he would have come up with a much higher correlation. However, this would not call for sterilization, but rather for improving schools, especially schools in slum areas. And this would be a task that would probably not appeal much to Shockley.

302

One further point, having to do with the other factor Shockley emphasized: IQ. Shockley was apparently unaware of how strongly IQ tests depend on cultural factors. They favor abilities that relate to the daily life, challenges and dominant interests of the groups they are made for. Shockley seemed ignorant of the fact that many questions of the type used in IQ tests favor men over women, others favor women. Such questions are not used, one does not want IQ tests that favor one sex over the other. Likewise, some questions would favor those who grow up in slums. Why should such questions be eliminated? Finally, and this seems to me to be the most important: Why emphasize IQ? What about all the other human traits that might be much more important than IQ? What about warm-heartedness, artistic abilities, craftsmanship and so many other qualities that we would like to see in ourselves and in our fellow human beings? The Shockley example might seem to support Kitcher’s negative assessment of the biological sciences and in particular genetics because they have supported discrimination and social repression. However, things are not that simple. The Shockley case shows how important it is when one evaluates science to look carefully at the quality of the projects. It also illustrates several points that are important in this evaluation of quality. Here I come down on the side of Roll-Hansen and Longino against Kitcher. The quality evaluation should be strict, in order to avoid waste of resources and in extreme cases bad science. However, this does not mean that acceptable projects must be compatible with our best current theories, as Kitcher proposes. Many of the major breakthroughs in science come from putting forth and testing theories that are incompatible with the current best ones. However, one should require that the researchers should know and understand the theories with which this new project has to compete, and also be aware of how prejudices, both cultural and scientific, may prevent one from thinking about hypotheses that may be more promising than the ones that first came to one’s mind.33 33

For many years Roll-Hansen and I have shared an interest in the interplay of science and society. I have included among the references two articles where I discuss the question of social control of science and take up some of the issues Roll-Hansen discusses here: Follesdal, D. (1979) and (1986a).

303

LOGIC AND RATIONALITY

Comment on Parsons Charles Parsons raises an interesting, but difficult question: Where should we draw the line between logic and mathematics? Should we include second-order logic in logic “in the basic sense,” or is it so entangled in mathematics that we should include only first-order logic or perhaps even less? Both Parsons and I have an interest in the history of philosophy, thinking that we can often learn from it. In this case, starting from the historical beginning and following the historical development can help us to see what should definitely be included in logic. Aristotle’s syllogistic logic is certainly a good candidate. However the medievals discovered arguments that turn on relations and were therefore not syllogistic although they certainly ought to qualify as logic. Certainly anything we can think of, regardless of field, can stand in relation to other things, and certainly in the identity relation to itself. These relational arguments therefore clearly belong to what Parsons calls “basic logic,” which he characterizes as follows: “a normative canon of reasoning about objects and subject matters very generally, perhaps even for anything that can properly be called thought.” I agree with Parsons that this canon is what we are after when we study logic. This counts strongly in favor of including first-order logic in logic. It also makes it natural to offer courses in first-order logic to everybody who wants to improve their reasoning in their own field and in daily life. Some universities require a course in truth functional logic for everybody. However, truth functional logic does not get us very far. Hardly any reasoning proceeds by truth functional logic, most reasoning requires predicate logic. Courses in truth functional logic may confirm many students’ opinion that logic is good for nothing. What students need, is an inspiring and well taught course in first-order logic, full of examples from their field and from daily life. First-order logic is a very useful tool, both for carrying out our

304

arguments and for clarifying thought. The difference between Wittgenstein and Husserl in their view on ultimate justification, which I discuss in my comment on Friedman, is an example. There, a difference in the order of quantifiers is so important that it will automatically be made explicit by a person who is familiar with first-order logic. Parsons mentions that first-order logic has no decision procedure and that for this reason maybe only monadic logic should be included in logic. He gives some reasons for this alternative. However, he is clearly uncomfortable with it, and I think that a decisive consideration is that such a logic would not have room for the relational arguments which made medieval philosophers dissatisfied with the syllogism. This counts in favor of identifying logic with first-order logic. Also, a further feature of first-order logic that makes it a natural cut-off point, was stated by Quine: first-order logic differs from various other stronger candidates by being complete. So the case for first-order logic is strong. What about second-order logic? An argument in favor of second-order logic has been that it captures mathematical structures up to isomorphism while set theory permits an overabundance of non-standard models. However, Jouko Väänänen has argued convincingly that this depends on perspective.34 When put into their proper contexts, one sees that both second order logic and set theory either manifest a high degree of categoricity or alternatively permit a plethora of non-standard models. Väänänen’s argument gives me one more reason for agreeing with Parsons that the entanglement of second-order logic with mathematics is too strong to make it part of basic logic. This entanglement makes secondorder logic dependent on mathematical insights in a way that brings us beyond the realm of logic. I concur with Parson’s approval of Quine’s terse statement that second-order logic is “set theory in sheep’s clothing.” It counts in favor of first-order logic that it makes the relationship between logic and mathematics perspicuous by treating the epsilon of membership as a special two-place predicate that needs its own axioms and brings us from basic logic into a special application of logic, to set theory and there-

34

Väänänen, J. (2012). See also Väänänen, J. (2001).

305

by mathematics. We thereby avoid the unsatisfactory entanglement of second-order logic with mathematics. What about other extensions of first-order logic, for example to different types of modal logic? Here I will follow Bolzano, who did not deal with this particular problem but gave us a way of demarcating logic, or what he called “Ableitbarkeit.” My Stanford colleague John Etchemendy discusses him in work that Parsons criticizes extensively in his paper. I would always like to defend a colleague. However, in this case, I agree with Parsons’ objections. In order to place modal logic in relation to basic logic, I will use one key idea of Bolzano, which, although it has its problems, which are discussed by Etchemendy and by Parsons, may be of help here. Bolzano’s approach to logical consequence gives, roughly, the extension of this notion, but does not address the question Prawitz discusses in his contribution to this conference: in what way does a logical inference justify the conclusion. Anticipating Ajdukiewicz, Tarski and Quine by a hundred years, Bolzano defined what it is for two expressions to have the same logical form. They have to satisfy three conditions: (1) they can be got from one another by keeping “logical terms” constant and substituting non-logical terms for one another, (2) the substitutions must be uniform – the term in question has to be substituted at all occurrences of the term for which it is substituted, and (3) the term that is substituted must belong to the same syntactic category as the term for which it is substituted. That is, ‘Plato taught Socrates’ does not have the same logical form as ‘Plato taught well’, since ‘Socrates’ and ‘well’ do not belong to the same syntactic category. Bolzano then defines logical truth as follows: a sentence is logically true if and only if all sentences of that logical form are true. He maintained that what terms we regard as logical is arbitrary. The more terms we include in our logical vocabulary, the more will we include in logic. If, for example, we include terms like ‘possible’ and ‘necessary’ we get alethic modal logic. If we include ‘knows’ we get epistemic logic, and so on.35 These various kinds of logic are usually called the logic of the terms that are kept constant: the logic of knowledge, the logic of belief, etc. The35

Follesdal, D. (1967).

306

se are appropriate labels, they indicate that we are here leaving basic logic and go into special fields. This is also reflected in the need for special assumptions for the terms that we keep constant. I will say more about Bolzano in my comments on Prawitz, which come next. Parsons brings out very well the complicated entanglement of second-order logic with mathematics, an entanglement which leaves very many questions in second-order logic unsettled. He thereby makes a good case that basic logic does not include second-order logic. Both considerations from below and from above hence indicate very strongly that firstorder logic is the best candidate we have for basic logic.

Comment on Prawitz Dag Prawitz takes up a topic that has engaged him ever since he wrote his dissertation, Natural Deduction. A Proof-Theoretic Study36 How do valid inferences justify assertions and beliefs? This question is part of a wide cluster of interrelated central questions in the philosophy of logic and language that Prawitz has worked on, among them: What is a valid inference? What is justification? How can the meaning of a statement be determined by what would establish the statement as true? Concerning the latter question Prawitz has developed in interesting ways an approach to meaning originating with Gentzen and Dummett. In this paper, Prawitz concentrates on one specific question in this cluster: “What does it mean that an inference is logically valid?” There are therefore many points of contact between his paper and that of Charles Parsons, in particular a critical evaluation of the Bolzano-Tarski approach to logical validity. Prawitz includes Etchemendy among his allies in his criticism of this approach, while Parsons was sharply critical of Etchemendy. The difference between Parsons and Prawitz on this point is not due to a difference in their views. Prawitz concentrates, approvingly, on Etchemendy’s criticism of the Bolzano-Tarski approach, while Parsons is 36

Prawitz, D. (1965).

307

sharply critical of Etchemendy’s way of handling the entanglement of logic and mathematics. Prawitz is interested in two notions of inference: one broad notion, which includes all inferences that provide grounds for their conclusions, and one narrower notion, which includes only that subset of the former that qualify as logically valid inferences. Examples of the former are easy to come by. Prawitz gives as an example the inference from the two premisses ‘Adam is longer than Beatrice’ and ‘Beatrice is longer than Carlo’ to ’Adam is longer than Carlo’. This inference clearly does not qualify as a logically valid inference, it depends on a special feature of the relation “longer than,” that it is transitive. This is not a feature that all relations share. Epistemic logic, which I mention in my comment on Parsons, similarly depends on special features of the predicates involved. This may be a place to say more about Bolzano, who is discussed both by Parsons and by Prawitz and whose account of logical consequence I outlined briefly in my comment on Parsons. Bolzano is one of my heroes, and he has a special connection to Prawitz’s teacher and thesis advisor, Anders Wedberg (1913-1978). Wedberg, who is hardly known outside Scandinavia, was a remarkable philosopher who was of great importance for philosophy in Sweden. Among his doctoral students are, in addition to Prawitz, also Stig Kanger and Jan Berg. Wedberg was reluctant to publish, but what he did publish had very high standards. His careful and innovative little book, Plato’s Philosophy of Mathematics, came out in 1955.37 His most influential work, the three volume History of Philosophy, came out in Swedish in 1958-1966. However, this original, argumentoriented survey of the history of philosophy did not become available in English until Prawitz arranged for an English edition at Oxford University Press in 1982-84.38 The third volume of this work, which is basically on philosophy in the twentieth century, begins with a chapter on Bolzano, who was born in 1781 and died in 1848, the year Frege was born. Wedberg was right. Bolzano should be regarded as the first philosopher of the twentieth century. Through his choice of topics, his way of 37 38

Wedberg, A. (1955). Wedberg, A. (1982-84).

308

dealing with them and the insights he arrived at, he anticipated the central figures in twentieth philosophy, such as Frege, Carnap and many others. Inspired and guided by Wedberg, Jan Berg wrote a dissertation, Bolzano’s Logic, which appeared in English in 1962.39 This book contributed much to making Bolzano known in the English-speaking world. At that time, Bolzano was almost unknown in the English-speaking world. His Paradoxes of the Infinite had been translated into English in 1950.40 However, as late as in 1961-62, when I first gave a series of lectures on Bolzano, there was not even an entry on him in Encyclopedia Britannica and nothing from his main philosophical works was available in English. The later filmmaker Terry Malick, whose term papers impressed me by their beautiful English and who also knew German, translated a few pages that were needed for my course. Malick hence has the honor of being the first to translate anything from Bolzano’s Wissenschaftslehre41 into English. Thanks to Berg, Morscher, George, Künne and his students Textor and Beyer, Rusnock, Russ, Sebestik, Simons, Sundholm, Winter, and many others, Bolzano is now widely studied. There is still much to be learned from him within the many fields in which he was far ahead of his time. However, there is one important difference between Bolzano’s and Prawitz’s approaches to validity: Bolzano’s approach, like Tarski’s, is model theoretic, while that of Prawitz is proof theoretical. According to Bolzano-Tarski, an argument is valid when it has no counter-model. Prawitz wants to understand why, if an argument is valid and one has grounds for the premisses, then the inference gives us a ground for the conclusion. This is a question that is not taken up by the Bolzano-Tarski approach. There is a tendency for model theorists to be realists: there is a structure that provides counter-examples even if we do not always succeed in finding them. Proof theorists are often non-realists or anti-realists. However, regardless of one’s ontological views, the proof theorist’s aim seems attractive: we do not just want to know that some statement is valid, we 39

Berg, J. (1962). Bolzano, B. (1851). 41 Bolzano, B. (1837). 40

309

want to know why it is valid. This is the enterprise to which Prawitz has made a number of important contributions. Prawitz shows convincingly how a number of attempts to explain how proofs can give us grounds for their conclusions are circular; they presuppose what they set out to show. Prawitz ends up with a “modal” notion: a valid inference has a “compelling force,” when one “acts” and carries out the inference, the act gives one a ground for the conclusion. We are compelled to hold on to the conclusion, we have become aware that it would be irrational not to hold the conclusion true. As Prawitz point out, “the necessity is not particularly connected with logically valid inferences; simply valid inferences are equally compelling.” The “act” aspect of Prawitz’s proposal brings to mind Jaakko Hintikka’s discussion of Descartes’s Cogito: Inference or Performance.42 Although there are important differences there are also similarities. Also Frege struggled with related problems when he introduced the Urteilsstrich, or judgment-stroke, to “convey assertoric force.” It seems to me that what Prawitz has in mind is the conviction that we get when we have struggled to understand something and get an insight that makes everything clear. An example from my own experience when I studied mathematics may perhaps help to illustrate this. I was taking a course in functional analysis based on a German book where everything was very complicated. Proofs of some theorems were often split into a large number of cases that were proved one by one, often in very convoluted ways. I could follow the proofs step by step and could see that they were correct. However, I got no grasp of the structure wherein all this was supposed to fit. Then, in the middle of the term, our professor came enthusiastically and told us that a new book had just been published that was far better and should replace the one we had been reading. This was RieszNagy’s Leçons d’analyse fonctionelle (1953).43 This book opened my eyes. Everything “fell into place.” My experience fits exactly what Prawitz describes as the difference between checking a proof without insight and seeing why it is valid. 42 43

Hintikka, J. (1962). Riesz, F. & B.V.Sz. Nagy (1953).

310

Now to my own view, which resembles that of Prawitz, although it stems from a quite different tradition, with Husserl, Duhem, Goodman, Quine, Israel Scheffler and Rawls as main figures. Rawls called it the method of reflective equilibrium. It is explained clearly and concisely by Nelson Goodman, in his Fact, Fiction and Forecast (1955): How do we justify a deduction? Plainly by showing that it conforms to the general rules of deductive inference. ..... But how is the validity of rules to be determined? Here again we encounter philosophers who insist that these rules follow from some self-evident axiom, and others who try to show that the rules are grounded in the very nature of the human mind. I think the answer lies much nearer the surface. Principles of deductive inference are justified by their conformity with accepted deductive practice. Their validity depends upon accordance with the particular deductive inferences we actually make and sanction. If a rule yields inacceptable inferences, we drop it as invalid. Justification of general rules thus derives from judgments rejecting or accepting particular deductive inferences. This looks flagrantly circular. I have said that deductive inferences are justified by their conformity to valid general rules, and that the general rules are justified by their conformity to valid inferences. But the circle is a virtuous one. The point is that rules and particular inferences alike are justified by being brought into agreement with each other. A rule is amended if it yields an inference we are unwilling to accept; an inference is rejected if it violates a rule we are unwilling to amend. The process of justification is the delicate one of making mutual adjustments between rules and accepted inferences; and in the agreement achieved lies the only justification needed for either.44

Four main features of the method of reflective equilibrium are prominent here: coherence, prereflective intuitive acceptances, justification and total corrigibility. Coherence together with prereflective intuitive acceptances yields justification, but this does not mean infallibility. We must always be prepared to revise our views. The idea of prereflective, intuitive acceptance as the basic source of evidence is very prominent in Husserl and I discuss this in Follesdal (1988) and various other articles. 44

Goodman, N. (1955). Here quoted from the second edition, Bobbs-Merrill Company, New York, 1965, pp. 62-64. The italics are Goodman’s.

311

Through reflection, systematization and observation we seek to gradually modify our acceptances, strengthen some of them and weaken others, but we do not attempt a whole-sale rejection of all of them in order to replace them with something radically new. Husserl emphasizes that there is no source of evidence upon which such a new edifice could be built, all the evidence there is, is imparted through these intuitive acceptances. This kind of pre-reflective acceptances that via the systematic interconnections in a theory spread to every part of the theory is also central in Pierre Duhem’s philosophy of science. In The Aim and Structure of Physical Theory, he writes: “These certainties and truths of common sense are in the last analysis the source of all truth and all scientific certainty.”45 Also what Duhem says about the physicist applies to the logician: “But while the physicist is powerless to justify this conviction, he is nonetheless powerless to rid his reason of it.” (Ibid., p. 27) This seems to me very close to what Prawitz has in mind when he ends his article by the following words: This kind of necessity makes it quite appropriate to place the word ‘must’ in front of the conclusion. However, when placed in that position, saying ‘hence B must be true’ should be understood to say, not that it must be the case that the B is true, but rather that we must hold B true.

Comment on Elster Elster distinguishes rationality from reason. I agree with him on what he says about reason. However, as he points out, we disagree on the notion of rationality. I regard rationality as consisting of three components: rationality of beliefs, rationality of values, and rationality of action. The latter starts from a person’s beliefs and values and uses these as input to explain the person’s actions. This latter component has been the center of attention in 45

Duhem, P. (1906), p. 264.

312

the theory of rational choice, to which Elster has made very interesting contributions. In the theory of rational choice one takes the person’s beliefs and values for given and explains the person’s actions from these. Explanation of action has to be based on the beliefs and values of the person. That is, such explanation is radically subjective. It starts from this person’s perspective, from the beliefs and values that the person has internalized. Here Elster and I agree. We also agree in warning against taking this to mean that the theory of rational choice presupposes egoism. One starts from the agent’s values, but these may very well be altruistic. This is a rather common misunderstanding, especially by people who are critical of the theory of rational choice.46 Where Elster and I disagree is where we go beyond rational choice theory and ask to what extent a person’s beliefs and values are rational. With regard to beliefs, Elster holds that while reason requires that the beliefs should be well-grounded in the agent’s evidence, rational choice presumes only a very modest consistency demand on the agent’s beliefs. “The rational person is one who does as well as possible by his or her own lights, whatever these may be.” I use the word rational in such a way that a person who abides with beliefs that seem to be in obvious discordance with evidence available to anybody in that situation, is not rational. In my view, the same holds for values. Elster maintains that “The desires are the unmoved movers of choice, which are not themselves chosen. At least this is true for a rational agent. If an agent desires to change his desires, it must reflect some form of irrationality.” My view is that values can be supported or rejected by rational arguments. I have argued for this in several writings, for example in "The Emergence of Justification in Ethics."47 There are many philosophers who, like Elster, hold that rationality considerations do not apply to values. Many of the members of the Vienna Circle had this view, as well as many of the existentialists. The fact that the main development in rational choice theory took place in a period when 46 47

For more details see Follesdal, D. (1986b) and Follesdal, D. (1980). Follesdal, D. (2005).

313

there was widespread skepticism concerning the role of argument in ethics may explain why rational choice theory takes values as ultimate posits and does not include a study of how values are acquired and justified. Unlike many rational choice theorists Elster does not reject the argumentative approach to values.48 He just maintains that this is not part of the study of rationality, but that it belongs to the realm of reason. One may find it odd that the notion of rationality shall not be applied to these arguments about values. However, the disagreement between Elster and me is, it seems, just a matter of terminology. Terminological disagreements can be settled in different ways. One may ask people how they use the word. In this case people who work in rational choice theory would be likely to say, like Elster, that the notion of rationality applies only to the explanation of action, and perhaps to some low degree to some consistency considerations concerning belief. However, persons working in other fields, for example in ethics, might find it odd that the notion of rationality should not apply to the arguments used in ethics. I do not regard terminological issues as important, especially not when, as in the case of Elster and myself, we are clear about our distinctions and therefore can communicate without difficulties. However, there may be some strategic reasons for using a broader notion of rationality than one that is confined to the realm of rational choice. The economic crises that have been succeeding one another the last few years, one more destructive than the other, may be an indication that something is missing in the study and application of economics: one is totally neglecting the study of value. Adam Smith saw this. It might be conducive to a more responsible economics and a better world if economists were required to broaden their study and their practical activities beyond the narrow realm of rational choice. Rational choice theory is only part of a broader field. Using the word “rational” for this whole broad field might help to remind us of this.49 48

See, for example Elster, J. and J. Roemer, eds. (1991). Elster discusses the economic crises and the shortcomings of economics in Elster, J. (2009). However, he blames defects in economic theory and the economist’s ”excessive ambitions” and does not criticize economists’ neglect of values. 49

314

MEANING

Comment on Gjelsvik Olav Gjelsvik goes straight to the central point: what kind of theory of perception is needed for an adequate theory of meaning? Meaning and communication is a theme that engages me a lot, and since Gjelsvik is the only speaker who addresses this, I will make a rather full comment. Philosophers of language and linguists have traditionally developed semantic theories without bringing in perception. Frege’s theory of sense and reference is an example. Frege took it for granted that different users of a language would connect the same sense with the same words, but he never made clear how this is supposed to happen. Even philosophers who accept that language is a social phenomenon seem to forget this when they discuss meaning and communication. They all too easily fall back on a Frege-like idea of meanings that we all share. One of Quine’s main contributions to philosophy was that he took the social nature of language seriously and asked how our experience of one another and the world around us can enable us to communicate. Perception is, of course, a most important factor in this process. However, perception involves a lot of intensional notions: structuring the world and individuating the objects in it, etc. One of the purposes of communication is to find out how others conceive of the world and its objects. Quine was aware of this, and in order to avoid begging the question by assuming from the beginning that we know the other’s world picture, he wanted to start from something that we can describe without knowing how the other structures the world. On the first page of Word and Object he states clearly that language is geared to the things around us that we perceive and deal with in our daily lives: Linguistically, and hence conceptually, the things in sharpest focus are the things that are public enough to be talked of publicly, common and conspicuous

315

enough to be talked of often, and near enough to sense to be quickly identified and learned by name; it is to these that words apply first and foremost.50

However, in the next chapter Quine focuses instead on stimuli, which he sometimes describes as evolving patterns of triggered nerve endings, other times as patterns of irradiation on the eye. These are two rather different notions, the former is not publicly accessible in our normal lives, in the situations where we normally learn and use language. The second notion, patterns of irradiation on the eye, is not one that we normally think about and are aware of, but at least we can put ourselves in the other’s place and decide what we would see or not see if we were to have our head in the same place, with our eyes in the same direction, etc. To take the rabbit behind the tree example that Gjelsvik mentions, we could at least decide that the other does not see a rabbit when it is hidden behind a tree. Similarly for the other senses. However, in 1960, when I asked Quine about what he meant by stimuli, he said that he had the first notion in mind, triggered nerve endings. Both notions are unsatisfactory. Although stimuli play a causal role in perception, we need much more to get full-fledged perception. Quine never held that the objects we perceive are stimuli. We perceive physical objects. Quine was fully aware of this, but he was set on seeing how far one can get in dealing with perception and the interplay between perception and meaning without bringing in intensional notions. In his later work he made important improvements and found, for example, that the whole notion of intersubjective comparison of stimuli that is so central in Word and Object could be left out. His final view, which he stated already in Word and Object, was that when one exhaust all the empirical evidence, without bringing in intensional notions, one will end up with a variety of alternative translation manuals, all of which are compatible with the empirical findings. Patrick Suppes, in his contribution to this symposium and in very many other articles and books, stresses the richness and complexity of the mechanics in our brain that are involved in perception, language learning 50

Quine, W.V. (1960), p. 1.

316

and language use. This is information we must make use of when we seek to understand communication. However, this is not evidence available to us in the normal process of language learning and language use. Our syntax will be corrected and adjusted towards perfection, since all our sentences, from those where we report what we see, hear, etc., to those where we report our most abstract beliefs, all wear their syntax on their sleeves and are constantly open to correction by fellow users of our language. Our semantics and our beliefs, however, and everything that we want to express through language, are not immediately available to our conversation partners. They have to be picked up on the basis of what is said. Our own beliefs and conception of the world are acquired and adjusted in terms of what we hear others say. And the beliefs and world view we take the other to have, is a product of an interplay between our own beliefs and the meaning we attribute to the expressions we hear and use. One way of explicating the difference between translation manuals is to say that each translation manual gives a different way of drawing the line between genuine disagreement and verbal disagreement. Each such correlation is a correct translation. There is nothing further to be right or wrong about. Brentano and following him, Husserl, developed a view on intentional states as states that we are in and can know through reflective analysis. If one accepts this view, one might think that among the various alternative translation manuals there is one that is the correct one, namely the one that correlates one person’s intentional states with corresponding intentional states in other persons. There are, however, several problems facing such a view. First, assuming that we can analyze and get clear about our own intentional states, what does it mean that two persons are in the same intentional state? If it means that the two persons give descriptions of their intentional states that translate into one another, we might ask: according to which translation manual? So we are back to Quine’s position again. One might think that at least when the two people speak the same language, this problem does not arise. However, as Quine pointed out repeatedly, his argument concerning indeterminacy of translation applies at home.

317

The underlying predicament is that, according to empiricism, communication has to take place via our senses. Telepathy or other forms of non-empirical transfer of mental states does not take place. Quine is sometimes taken to reject mental states for ontological reasons. In Word and Object there is some vacillation on this point. However, from 1990 on Quine is clear: “… physicalism is irrelevant, monism is irrelevant. One can wallow in the rankest mentalistic ontology without affecting the indeterminacy of translation.”51 Indeterminacy follows from the empirical nature of communication: one communicates by help of one’s senses, and where they leave a slack there is nothing more to be right or wrong about. In Word and Object Quine states his view as follows: Brentano’s thesis of the irreducibility of intentional idioms is of a piece with the thesis of indeterminacy of translation. ... To accept intentional usage at face value is, we saw, to postulate translation relations as somehow objectively valid though indeterminate in principle relative to the totality of speech dispositions.52

Now on to Davidson. Gjelsvik wrote his dissertation in Oxford with Davidson as one of his advisors during Davidson’s tenure as Eastman Professor there in 1984-85. Both he and I regard Davidson highly. I think we agree that Davidson made two great improvements on Quine: 1. Davidson enhanced the procedure Quine uses in constructing translation manuals. Quine splits up sentences into parts that are then combined to form new sentences. Davidson pointed out that these constructions could not be merely syntactic, but had to be semantic, the constructions had to show how the semantic features of whole sentences depend on the semantic features of their part, à la Tarski’s theory of truth. This led among other things to “the Davidsonian program” of extending Tarski’s theory of truth to comprise adverbs and other kinds of linguistic constructions that we find in natural languages. 51 52

Barrett, R. and Roger Gibson, eds. (1990), p. 110. Quine, W.V. (1960), p. 221.

318

2. Davidson extended the range of evidence that can help us narrow down the assortment of translation manuals. In addition to assent and dissent, he also took into account our choices between different alternatives for action. Gjelsvik and I also agree that Davidson made two wrong starts in his attempts to overcome the difficulties Quine got into: First, Davidson’s attempted to get by with just maximizing agreement, without bringing in perception. This attempt failed, as shown by the rabbit behind the tree example. Secondly his next attempt, by way of triangulation, takes it for granted that the other structures the world and individualizes its objects the way we do, and thereby presupposes what we want to find out. During a meeting at Stanford in 1986, sponsored by Stanford’s Center for the Study of Language and Information (with John Perry as its Director), this came out very clearly. Quine, Davidson, Dreben and I sat together for five days, and the main topic was the proximal versus the distal (stimuli versus physical objects). Quine then recognized that his appeal to stimuli was fundamentally flawed. However, he did not accept Davidson’s idea of triangulation. Davidson could not understand why. Quine had a special awareness of problems. He saw that the notions of object and individuation are very complex. In Word and Object he tried to get around the problems by turning from objects to stimuli, He now saw that this was a blind alley. However, he did not regard Davidson’s triangulation as getting around these problems. On the contrary, Quine found that triangulation presupposes what we want to find out, namely how the other structures the world and individualizes its objects. Gjelsvik points out that even as late as in 1990 there was a difference between Davidson and me in regard to the role of causality in language learning. He is right. In my view Davidson’s appeal to causal chains in connection with language learning is illusory. Davidson argued that in a language learning situation the expression learned relates to the object, event, or situation which is the last common cause in the two infinite causal chains that lead to the sense organs of the teacher and learner in the learning situation.

319

Against this I objected that the events with which I am familiar have a multitude of causes. One should rather talk about causal trees. One then sees immediately that the notion of the last common cause makes little sense. Which of the many junctions between our two causal trees is the last one? Similar problems arise for many causal theories of reference. Gjelsvik brings up a number of the difficulties that confront us when we try to work out a theory of the perception of objects. Any satisfactory theory of perception must be able to deal with these issues. Gjelsvik ends his paper by stating that I built more on Quine than on Davidson, and that the direction I have pointed to springs from my study of Husserl. He is right on both points. I have already described how I side with Quine against Davidson on many key issues. As for Husserl, Quine never gave up his insight that meaning and communication are intimately connected with perception, and Husserl gave a masterly analysis of perception. Some years ago I taught a course on perception where we considered the main theories of perception that have been proposed and the various difficulties they face, which have led to important modifications and often abandonment of the theories. Remarkably, Husserl’s theory of perception, which was set forth more than a hundred years ago and antedated all the other theories we were considering, was able to handle all these difficulties which made the other theories bog down. This testifies to Husserl’s good philosophical intuitions and the very careful way in which he worked them out. Quine was impressed by this. In an interview with Giovanna Borradori in 1994, he said: I recognize that Husserl and I, in very different ways, addressed some of the same things.53

53

“Twentieth-Century Logic,” p. 64 in Quine, W.V. (2008b).

320

REFERENCE

Comment on Perry John Perry, with his usual clarity and simplicity, first gives a sympathetic reading of my 1961 dissertation and then raises some questions concerning issues that I do not take up in the dissertation, but that are relevant both to the work he and Barwise did on “the slingshot” and to work that he and also Kaplan have done on indexicals and demonstratives. First, the dissertation. Perry regrets that he had not read my dissertation when he and Barwise wrote their paper on the slingshot (1981).54 He is excused. In the summer of 1961 Max Black contacted me and said that he wanted my dissertation for the blue series he edited, which was devoted to the first book by various philosophers, including, for example, Hintikka’s Knowledge and Belief.55 However, Black wanted me to write an introduction which would make the book more accessible to non-logicians. I was then teaching my first course and had also promised to translate several articles for van Heijenoort’s Source Book in Mathematical Logic, and I told Black that I would not have time to write an introduction. I now regret this, I should rather have postponed the translations. So the book was not published. Widener Library at Harvard has a list of those to whom the volume was sent on interlibrary loan, and they were quite many. Also the dissertation was published in a mimeographed edition by Oslo University Press in 1966. However, it was not until 2004 that Robert Nozick brought it out in the series of Harvard dissertations he edited for Routledge. So Perry is definitely excused. Perry is right that a main point of the dissertation is to block Quine’s argument that modal distinctions are doomed to collapse. However, I also wanted to show that we have to give up the traditional Frege-inspired view on reference, according to which sentences, general terms and singular terms all have a sense that determines their reference or extension. All 54 55

Barwise, J. and J. Perry (1981). Hintikka, J. (1962).

321

types of semantics that had been proposed at that time, including many that were partly in opposition to Frege, as for example Carnap’s, shared this feature: they are “one-sorted” and treat sentences, general terms and singular terms on a par. For Frege, Carnap, Church and others, this simple “onesortedness” is an argument in their favor. I argued that names and other “genuine singular terms,” which Kripke in 1971 aptly called “rigid designators,” have to stick to their objects. In the jargon of modal logic, they have to keep the same reference in “all possible worlds.” (As Perry notes, the point can be made without appeal to modality and “possible worlds.” What matters is individuation and keeping track of one and the same individual through various kinds of vicissitudes.) My main argument for this “two-sorted” semantics was that it was needed in order to block Quine’s argument and save the modal distinctions from collapse. However, I also pointed out that the way we normally use names counts in favor of this view; we use names to indicate that we want to keep on talking about the same object, and not philander from one object to another. Quine was my thesis advisor, and although my thesis showed that he had to give up his favorite argument against the modalities, he was very positive. He agreed with me and immediately rewrote the pertinent passages in a new edition of From a Logical Point of View.56 Quine was always more interested in finding out what was right than in being right, to echo Kaplan’s characterization of Carnap.57 Perry mentions that Quine earlier had argued that the modal logician has to accept what Quine called “Aristotelian essentialism,” a metaphysical position about which Quine was unenthusiastic. I found this argument of Quine’s sound, but concluded that this was a reason for accepting this metaphysical doctrine, since without it, one could not deal with causality and other notions that are needed in science, for which Quine had great respect. Later on, Ruth Marcus, Terrence Parsons and others have argued that one can do modal logic without committing oneself to essentialism. However, they have mixed up two notions of essentialism, they criticize Kripke’s talk of essences, which has nothing to do with “Aristotelian essentialism,” and 56 57

Quine, W.V. (1953), 2nd edition, 1961. Kaplan, D. (1975).

322 which can and should be avoided.58 So here I agree with Marcus, although she thought, unfortunately, that she had refuted Quine’s view on essence. While a main point in my dissertation is the insistence on a twosorted semantics, where names and other “genuine singular terms” keep their reference, I left open the question how this can be achieved. Here, several options might be considered, but I find them all unsatisfactory. The idea of there being a causal tie between the term and the object, which provides a stable reference for the term, must be rejected for two reasons: first, and most importantly, it goes against the social nature of language, according to which meaning and reference is established and maintained through social interaction. And secondly, it gets into trouble with reference shifts, which actually occur. (I use ‘refer’ and ‘reference’ throughout this comment. If one wants to distinguish ‘naming’, ‘denoting’, ’referring’ and ’describing’ the way Perry does, one might use the neutral ‘designate’ where I have ‘refer’, and designation’ for my ‘reference’ ) The idea that the semantics of a language as well as the syntax are established and learned through a transmission process seems a natural one. This was, I believe, first proposed by Peter Geach, in “The Perils of Pauline” in 1969.59 However, both his view and Kripke’s very similar later view do not account well for reference shift. I have proposed a somewhat different view, but I will not go into this here.60 And much more work remains to be done. Now to Perry’s three questions, concerning sequestering, intrusion and redistribution. First sequestering. This seems to me to be a very important and common feature of genuine singular terms. These terms carry information, but they belong to a syntactical category that signals to users of the language that this information just fixes the reference, but does not call for new evaluation of what the reference is every time the name is used. A type of example I often use, is 58

See Follesdal, D. (1986d). Geach, P. (1969). 60 Follesdal, D. (1986a). 59

323

when we make a listener attend to a particular object or person by for example saying ‘that boy with the glasses’ and then introduce a name: ‘… is Paul’. By introducing the name we signal that now we want to speak about that person regardless of whether he takes off the glasses and no longer satisfies the original description. The same happens when we introduce a personal pronoun for this person, like ‘he’. This is also how we use indexicals, demonstratives and variables, variables are simply the pronouns of formal languages. So names, pronouns, indexicals, demonstrative and variables are all expressions that signal sequestering. I hence differ somewhat from Perry in that I think that names have more in common with pronouns, indexicals, demonstratives, and variables than Perry does (at least in his formulation of this question). They are all introduced in a particular situation by some descriptive and/or demonstrative means, and signal that we want to stick to the reference. However, names are more permanently fixed to their references than these other terms. These terms are not permanently attached to one object, the way names are, but they all differ from definite descriptions; descriptions normally philander from object to object, and this is why they create so much trouble in modal and other non-extensional contexts. I say “normally”; there are cases where definite description are used like names. This requires, however, some special signaling. This does not mean that all these kinds of expression, indexicals, demonstratives, pronouns, and variables are semantically alike. There are important differences among them – this is why we sort them into different semantic groups. However, they are genuine singular terms, they are the reason why in my thesis I talk about “names and other genuine singular terms.” Perry ends this point by asking “Could the Orthodox Modal Logician retain a one-sorted semantics for all singular terms other than variables, supposing that all singular terms denote, while still evading the slingshot by appeal to sequestering of this sort?” My answer is no. Definite descriptions, class abstracts and various other expressions would still create havoc. It is just this kind of expressions that Quine uses in his argument to derive the collapse. So we would

324

need a semantics where these expressions are not treated on a par with “genuine singular terms.” Now intrusion. This is a very interesting phenomenon. As Perry puts it, there are cases where an operator can unlock extra information in a demonstrative. His example is ‘I think that city is Pittsburgh’. Here although what I am looking at is Buffalo, the sentence is true. However, when we substitute ‘Buffalo’ for ‘that city’ we get a false sentence. How can that be, are not ‘that city’ and ‘Buffalo’ in this case co-referring? Clearly, if I said only ‘That city is Pittsburgh’, what I said would be false, but appending ‘I think’ changes the reference of ‘that city’. So, operators like ‘I think’, ‘I believe’ and various others make it clear that the descriptive element in the demonstrative is supposed to refer to what fits the description in the world of my beliefs, or whatever world the operator brings in. This often works well, in this case I just mix up two cities that are part of the universe we share. However, we get problems when my world of beliefs is quite different from that of the listener and we have problems correlating our universes. I will not go into this here. In such cases we will have trouble understanding what the other person refers to, and communication will break down, Perry then turns to Hintikka’s apparently similar example: Seeing the man on the beach handing classified documents to obvious Bolsheviks, I say ‘I know that man passed the secrets’. I would treat this on a par with the previous example: The context makes it clear that the demonstrative clause refers to the man I am seeing. However, as Perry suspects, I would not like to follow Hintikka in distinguishing different types of individuation. Among other difficulties this gets us into troubles with Leibniz’s law, which I regard as fundamental for identity and individuation. And as I stress in my dissertation, clarity about individuation and individuals is crucial to reference and quantification. Finally there is redistribution. Here my answer can be short and give Perry solace. In my thesis I showed that Quine’s argument for the collapse of modal distinctions applies to all

325

non-extensional constructions, not just the logical modalities, but also the epistemic and deontic modalities, to causality and probability statements, etc., and that it made no assumptions that were not generally accepted at that time. The only assumption that could feasibly be questioned, was the one-sorted semantics that was current at that time. I showed that a twosorted semantics where names and other genuine terms stick to their reference would invalidate the argument and also was plausible for other reasons. However, I did not show that there are no further problems. When it comes to the logical modalities, I agreed with Quine that the basic notions of logical necessity and possibility were unclear, and that appeal to “possible worlds” did not get us out of this circle. I gave a brief presentation of the semantics of modal logic at its best, based on Kanger and Hintikka’s idea of worlds being possible relative to one another.61 Their idea gave a natural semantics for the various modal systems – what has later been called “Kripke semantics.” These structures give rise to algebraic problems but they do not tell us which worlds are possible and which not. However, two-sorted semantics does not unravel all the problems of semantics, nor undercut all problematic arguments. Barwise and Perry showed that arguments based on logical equivalence lead to undesirable results and to deal with them they proposed various forms of partiality built into their situation theory. These arguments are not blocked by a twosorted semantics, and I welcome Barwise and Perry’s work as an interesting way of getting around these problems. So, Perry: take solace, there was part of the story left to tell when you and Barwise wrote your article in 1981.

Comment on Linnebo It has been a great pleasure to listen to Øystein Linnebo’s paper. First, he is a careful reader and good interpreter. Secondly, he comes up with original ideas and develops them in interesting ways. 61

Presented in Stig Kanger’s dissertation, Kanger, S. (1957), and in Jaakko Hintikka’s two papers Hintikka, J. (1957a) and (1957b).

326

A key idea of Linnebo’s proposal is the distinction between semantics and meta-semantics. That, speaking modally, names refer to the same object in all possible worlds, is a semantic view in the usual sense of ‘semantic’: The semantic value of an expression is the contribution that this expression makes to the truth value of the sentences in which it occurs. This semantic view is the crucial idea in the “new theory of reference.” If one agrees with this semantic view, that is, that names stick to their objects in all possible worlds, then the next question is ”How does this relation between name and object come about?” Linnebo calls this a meta-semantic question. As he mentions, several answers to this question have been proposed, the two main types are the historical chain theory, first put forward by Geach and followed up by Kripke, and various causal theories, put forward by Evans and later developed further in different directions. Linnebo and I both think that the Fregean notion of sense may still have something to contribute. But what? Clearly, a sense could fit uniquely one object in one situation and another object in another situation (or “possible world” if you will). What role do then senses play if our semantics requires that names should never change their reference? I try to give part of an answer to this in the discussion of sequestering in my response to Perry: names signal to users of the language that the sense used to fix the reference when the name is introduced, does just this; it fixes the reference, but there is no call for new evaluation of what the reference is every time the name is used. Linnebo sets forth several new ideas. The first is to “translate the problem of explaining how a singular term comes to refer into the problem of how certain complete sentences involving the term come to be meaningful.” This is a very reasonable step. If we use names in order to signal that we want to keep track of the same object through its varying stages, shapes, etc. (through “all possible worlds”), then the sense we attach to the name should be designed to do that, This leads Linnebo to the next step: In order to use a name properly, one has to have a conception of what it is to be “the same object.” That is, one has to rely upon a theory of individuation. Linnebo gives some examples: physical objects, directions, shapes, syntactic types, and natural num-

327

bers. I consider this the right way to go. I have, however, two further suggestions: First, we do not start from unique basic criteria, for example the notion of parallel in order to arrive at direction, and the notion of spatiotemporal continuity in order to arrive at physical bodies. I think that our various criteria and the correlated notions of identity of objects come as a package: Theoretical and practical considerations make us favor certain notions of objects rather than others, and the criteria of individuation that we make use of are adjusted accordingly. Secondly, while most of the examples Linnebo mentions are fine, for example the notion of direction seems to be neatly connected with the notion of parallel, the notion of a physical object is not so neatly correlated with the notion of spatiotemporal continuity. Spatiotemporal continuity is certainly an important factor, but it has to be balanced with other factors. Both spatiotemporal continuity and these other factors are, as I see it, a result of balancing various theoretical and practical considerations. I assume that it is this kind of considerations Linnebo has in mind when he says that the different chunks of stuff “are spatiotemporally connected in a suitable way” (my emphasis). Also his example of pointing to the floor and saying “this body” makes a related point. Finally, there is one important step that Linnebo mentions, but does not take in this paper. He focuses on thoughts rather than on language. This allows him to concentrate on one person rather than on a whole language community. However, it is crucial that we take this further step to language and language community. In fact, the idea we started from, that names stick to their objects, is in my view intimately connected with the fact that we use names to indicate to one another that we want to refer to the same object whenever we use that name. Definite descriptions in most of their use do not signal this. What linguistic expressions refer to, and what they mean, is a product of social interaction. A study of this interaction is crucial for our understanding of how names function. However, Linnebo states explicitly that “for the purposes of this paper I will have nothing to say about how linguistic expressions come to refer.” When we turn from thought to language, one of the five theses that Linnebo ascribes to me, becomes doubtful, namely what he calls “Russell’s

328

Principle”: “In order to refer to an object, a thinker must know which object he is referring to; that is, he must possess some discriminating conception of the referent.” He notes that “this principle isn't explicitly endorsed in Føllesdal's writings. But the underlying idea that the relation of reference must be epistemically constrained is frequently endorsed.” Linnebo is right that I maintain that the relation of reference must be epistemically constrained. However, the constraints may be weak, we may refer without possessing any discriminating conception of the referent. This is particularly clear when we pass from thought to language: Listening in on a conversation I may pick up a proper name, say ‘Baldy’ and then use it without having an inkling of whom is referred to. Nevertheless, one refers to this person and says something true or false about him. Linnebo manages to combine the semantic theses of rigidity and compositionality with a Frege-inspired account of reference. I think this is the right way to go.

References Barrett, R. and R. Gibson, eds. (1990). Perspectives on Quine. Oxford: Blackwell. Barwise, J. and J. Perry (1981) "Semantic Innocence and Uncompromising Situations." Midwest Studies in Philosophy 6, pp. 387-403. Berg, J. (1962). Bolzano’s Logic. Stockholm: Almqvist & Wiksell. Beyer, C. (2006). Subjektivität, Intersubjektivität, Personalität – Ein Beitrag zur Philosophie der Person. (Reihe: Ideen und Argumente) Berlin/New York: de Gruyter. Bolzano, B. (1837). Wissenschaftslehre. Versuch einer ausführlichen und grösstentheils neuen Darstellung der Logik mit steter Rücksicht auf deren bisherige Bearbeiter, 4 volumes, Sulzbach: J. E. v. Seidel; 2nd improved edition: Leipsic: Meiner, 1929, 1929, 1930, and 1931. English translation of selected parts: Theory of Science, Rolf George (ed.), Oxford: Oxford University Press, and Berkeley-Los Angeles: University of California Press, 1972; and Theory of Science, Jan Berg (ed.), Dordrecht: Reidel, 1973.

329

-“- (1851). Paradoxien des Unendlichen, Franz Přihonský, (ed.), Leipsic: C. H. Reclam sen.. Darmstadt: Wissenschaftliche Buchgesellschaft, 1964. English translation: Paradoxes of the Infinite, Donald A. Steele, ed., London: Routledge and Kegan Paul, and New Haven: Yale University Press, 1950. Daniel, Mano (1992). “A bibliography of the noema.” Drummond, J. and Embree, L. (1992), pp. 227-248. Also: Center for Advanced Research in Phenomenology: http://phenomenologycenter.org/about/specialized-bibliographies/noema/ Dreyfus, H.L. and H. Hall, eds. (1982). Husserl, Intentionality, and Cognitive Science, Cambridge, Mass.: The MIT Press. Drummond, J. and L. Embree, eds. (1992). The Phenomenology of the Noema. Dordrecht: Kluwer. Duhem, P. (1906). La théorie physique. son objet, sa structure. Paris: Chevalier et Rivière. English translation: The Aim and Structure of Physical Theory, Philip P. Wiener, transl. Princeton: Princeton University Press, 1954. Elster, J. (2009). ”Excessive Ambitions.” Capitalism and Society 4, Issue 2, Article 1. Pp. 1-30. Elster, J. and J. Roemer, eds. (1991). Interpersonal Comparisons of Well-Being. Cambridge: Cambridge University Press. Etchemendy, J. (1990). The Concept of Logical Consequence. Cambridge, Mass.: Harvard University Press. Reissued Stanford: CSLI, 1999. Follesdal, D. (1965). “Review of Jan Berg: Bolzano’s logic.” Isis 56, pp. 390-91. -“- (1967). ”Knowledge, identity and existence.” Theoria 33, pp. 1-27. -“- (1979). ”Some ethical aspects of recombinant DNA research.” Social Science Information 18, pp. 401-19. -“- (1980). ”Explanation of action.” In: Risto Hilpinen, ed., Rationality in Science. Dordrecht: Reidel, pp. 231-247.

330

-“- (1986a). ”Reference and sense.” In Venant Cauchy, ed., Philosophy and Culture: Proceedings of the XVIIth World Congress of Philosophy. Montreal: Editions du Beffroi, Editions Montmorency, 1986, pp. 229-239. -“- (1986b). Rationality and Irrationality. Proceedings of a conference in Bellagio, 1982. Epistemologia 9 (1986). Edited and with an introduction by DF: ”What is rationality?” (pp. 5-13). -“- (1986c). ”Risk: philosophical and ethical aspects.” In Anton Brøgger and Per Oftedal, eds, Risk and Reason: Risk Assessment in Relation to Environmental Mutagens and Carcinogens. (Progress in Clinical and Biological Research, Vol. 208), New York: Liss Publications, 1986, pp. 41-52. -“- (1986d). ”Essentialism and reference.” In Lewis E. Hahn and Paul Arthur Schilpp, eds., The Philosophy of W.V. Quine (The Library of Living Philosophers). La Salle, Ill.: Open Court, pp. 97-113. (Reply by Quine, pp. 114-115). -“- (1988). ”Husserl on evidence and justification.” In Robert Sokolowski, ed., Edmund Husserl and the Phenomenological Tradition: Essays in Phenomenology (Studies in Philosophy and the History of Philosophy, Vol. 18). Washington: The Catholic University of America Press, pp. 107-129. -“- (1990). ”The Lebenswelt in Husserl.” In Leila Haaparanta, Martin Kusch, and Ilkka Niiniluoto, eds., Language, Knowledge, and Intentionality: Perspectives on the Philosophy of Jaakko Hintikka (Acta Philosophica Fennica, Vol. 49). Helsinki, pp. 123-143. -“- (1998). ”Husserl’s idealism.” In Marcelo Stamm, ed., Philosophie in synthetischer Hinsicht (Festschrift für Dieter Henrich). Stuttgart: Klett-Cotta, pp. 189202. -“- (2005). "Ultimate Justification in Husserl and Wittgenstein." In Johann Christian Marek and Maria Elisabeth Reicher, eds., Experience and Analysis. Erfahrung und Analyse. Wien: hpt et öbv, pp. 127-142. -“- (2005). “The emergence of Justification in Ethics.” European Review 13, pp. 169-182. This essay overlaps with my essay for Frauchiger. M. and Essler, W.K. (2008), pp. 51-65.

331

Frauchiger, M. and W.K. Essler, eds. (2008). Representation, Evidence, and Justification: Themes from Suppes. Ontos Verlag, 2008, pp. 51-65. Geach, P. (1969), “The Perils of Pauline.” Review of Metaphysics 23, pp. 287-300. Reprinted in Geach, Logic Matters, Oxford: Blackwell, 1972, pp. 153-165. Goodman, N. (1955). Fact, Fiction and Forecast, Harvard University Press, Cambridge, Mass., 1955 and later editions. Gurwitsch, A. (1966). Studies in Phenomenology and Psychology, Evanston, Ill.: Northwestern University Press. Hintikka, J. (1957a). “Modality as Referential Multiplicity.” Ajatus 20 (1957), pp. 4964. -“- (1957b) “Quantifiers in Deontic Logic.” Societas Scientiarum Fennica, Commentationes Humanarum Literarum 23, no. 4. Helsinki, 1957. -“- (1962a). Knowledge and Belief. Ithaca, N.Y.: Cornell University Press. -“- (1962b). “Cogito, Ergo Sum: Inference or performance?” Philosophical Review 71, pp. 3-32. Hintikka, J., ed. (1975). Rudolf Carnap, Logical Empiricist. Dordrecht: Reidel. Husserl, E. (1886-1901). Studien zur Arithmetik und Geometrie. Texte aus dem Nachlass (1886-1901), ed. Ingeborg Strohmeyer. The Hague, Netherlands: Martinus Nijhoff, 1983 (Husserliana XXI). -“- (1905-1920). Zur Phänomenologie der Intersubjektivität. Texte aus dem Nachlass. Erster Teil. 1905-1920. [On the phenomenology of intersubjectivity. Texts from the estate. Part 1. 1905-1920.] Edited by Iso Kern. The Hague, Netherlands: Martinus Nijhoff, 1973 (Husserliana XIII). -“- (1918-1926). Analysen zur passiven Synthesis. Aus Vorlesungs- und Forschungsmanuskripten, 1918-1926, ed. Margot Fleischer. The Hague, Netherlands: Martinus Nijhoff, 1966 (Husserliana XI).

332

-“- (1923/4). Erste Philosophie Erster Teil: Kritische Ideengeschichte. [First philosophy. First part: the critical history of ideas.] ed. Rudolf Boehm. The Hague, Netherlands: Martinus Nijhoff, 1956 (Husserliana VII). -“- (1929). „Formale und transzendentale Logik. Versuch einer Kritik der logischen Vernunft.“ Jahrbuch für Philosophie und phänomenologische Forschung, 10. Halle a.d.S.: Niemeyer. v-xiii; 1-298. Husserliana edition, ed. Paul Jansen, The Hague, Netherlands: Martinus Nijhoff 1974 (Husserliana XVII). English translation by Dorion Cairns: Formal and transcendental logic. Attempt at a critique of logical reason. The Hague, Netherlands: Martinus Nijhoff 1969. -“- (1930). “Nachwort zu meinen ’Ideen I‘.“ In Jahrbuch für Philosophie und phänomenologische Forschung, 11. Halle a.d.S.: Niemeyer. 1930. Reprinted in Ideen zu einer reinen Phänomenologie und phänomenologischen Philosophie. Drittes Buch: Die Phänomenologie und die Fundamente der Wissenschaften [Ideas pertaining to a pure phenomenology and to a phenomenological philosophy. Third book: Phenomenology and the foundations of the sciences.], ed. Marly Biemel. The Hague, Netherlands: Martinus Nijhoff, 1971 (Husserliana V), pp. 138-162. -“- (1936). Die Krisis der europäischen Wissenschaften und die transzendentale Phänomenologie. Eine Einleitung in die phänomenologische Philosophie. Ed. Walter Biemel. The Hague, Netherlands: Martinus Nijhoff, 1976 (Husserliana VI). English translation by David Carr: The crisis of European sciences and transcendental philosophy. An introduction to phenomenology. Evanston, Ill.: Northwestern University Press, 1970. -“- (1938). Erfahrung und Urteil. Praha: Akademia-Verlag. New edition by Ludwig Landgrebe, Hamburg: Claassen Verlag, 1964. English translation by James S. Churchill and Karl Ameriks, Evanston: Northwestern University Press, 1973. -“- (1994a). Briefwechsel, Band V, Die Neukantianer. In Verbindung mit Elisabeth Schuhmann herausgegeben von Karl Schuhmann. Dordrecht: Kluwer. -“- (1994b). Briefwechsel, Band VII, Wissenschaftlicher Korrespondenz. In Verbindung mit Elisabeth Schuhmann herausgegeben von Karl Schuhmann. Dordrecht: Kluwer. Kanger, Stig (1957). Provability in Logic. Stockholm: Almqvist & Wiksell.

333

Kaplan, D. (1975). “Hommage to Rudolf Carnap.” In Hintikka, J., ed. (1975), pp. xlvii-xlix. Kern, I. (1964). Husserl und Kant: Eine Untersuchung über Husserls Verhältnis zu Kant und zum Neukantianismus. Haag: Nijhoff. Kersey, E. (1983). “The Noema, Husserlian and Beyond: An Annotated Bibliography of English Language Sources,” Philosophy Research Archives 9. Kitcher, P. (2002). Science, Truth and Democracy. Oxford: Oxford University Press. Longino, H. (2002). The Fate of Knowledge. Princeton: Princeton University Press. Prawitz, D. (1965). Natural Deduction. A Proof-Theoretic Study. Stockholm: Almqvist and Wiksell. Quine, W.V. (1953). From a Logical Point of View. Nine Logico-Philosophical Essays. Cambridge, Mass.: Harvard University Press. 2. ed., 1961, 2. ed., revised, 1996. -“- (1960). Word and Object. New York: John Wiley and Sons, Cambridge, Mass: MIT Press. -“- (1986). Lewis E. Hahn and Paul Arthur Schilpp, eds., The Philosophy of W.V. Quine (The Library of Living Philosophers). La Salle, Ill.: Open Court. -“- (2008a). Confessions of a Confirmed Extensionalist and Other Essays. Cambridge, Mass: Harvard University Press. -“- (2008b). Quine in Dialogue. Cambridge, Mass: Harvard University Press. Riesz, F. & B.V.Sz. Nagy (1953). Leçons d’analyse fonctionnelle. Budapest: Akadémiai Kiadó. Roll-Hansen, N. (2004). The Lysenko Effect. The Politics of Science. Amherst, NY: Humanity Books. -“- & G. Broberg, ed. (2005). Eugenics and the Welfare State. Norway, Sweden, Denmark and Finland. East Lansing, MI: Michigan State University Press.

334

Ryckman, T. (2005). The Reign of Relativity: Philosophy in Physics 1915-1925. Oxford Studies in the Philosophy of Science. Oxford: Oxford University Press. Sartre, J.-P. (1948). “Conscience de soi et connaissance de soi.” Bulletin de la Société Française de Philosophie, 42, pp. 49–91; English translation: “Consciousness of self and knowledge of self,” in N. Lawrence and D. O'Connor, eds., Readings in Existential Phenomenology, Englewood Cliffs, NJ: Prentice-Hall, pp. 113–42. Smith, D. (2007). Husserl. (The Routledge Philosophers). London: Routledge. Suppes, P. (1991). “The principle of invariance with special reference to perception.” In J. Doignon & J. Falmagne (eds.), Mathematical Psychology: Current Developments. New York: Springer, 35-53. -“- (1995). “Some foundational problems in the theory of visual space.” In R. Duncan Luce, M. D'Zmura, D. Hoffman, G. J. Iverson & A. Kimbal Romney (eds.), Geometric Representations of Perceptual Phenomena: Papers in honor of Tarow Indow on his 70th birthday. Mahwah, NJ: Lawrence Erlbaum, 37-45. -“- (1999). With B. Han, J. Epelboim and Z.-L. Lu. “Invariance of brain-wave representations of simple visual images and their names.” Proceedings of the National Academy of Sciences, 96, 14658-14663. Väänänen, J. (2001). “Second-Order Logic and Foundations of Mathematics.” Bulletin of Symbolic Logic 7 (4), 504-520. -“-

(2012). “Second Order Logic Or Set Theory?” Bulletin of Symbolic Logic, 18 (1), 91-121.

Wedberg, A. (1955). Plato’s Philosophy of Mathematics. Stockholm: Almqvist and Wiksell. -“- (1982-84). History of Philosophy, Vols. 1-3. Oxford: Oxford University Press. Wittgenstein, L. (1969). On Certainty (Über Gewissheit), ed. G.E.M. Anscombe and G.H. von Wright. Translated by D. Paul and G.E.M. Anscombe. Oxford: Basil Blackwell, 1969-1975.

Interview with Dagfinn Føllesdal by Michael Frauchiger∗

Frauchiger: Professor Føllesdal, how do you see the prospects for philosophy maintaining its discreteness as an academic field? Today philosophy gives the impression of getting increasingly differentiated into diverse specialist areas, losing its cohesion as a distinctive abstract discipline, which is clearly distinguishable from the variegated specialized branches of empirical science. Is academic philosophy currently disintegrating into a loose association of inter- and transdisciplinary projects? Dagfinn Føllesdal: This is a very important issue. I think that philosophy more than any other subject depends on being an integrated subject. While the separate sciences often can be split into parts that can be dealt with independently, the different parts of philosophy are so dependent on one another that it is normally impossible to concentrate on one area and just focus on that without having studied the others. I think that a very important feature of a good philosophy program is that it gives the students a broad background in the different parts of philosophy. Because of this interdependence between the different fields of philosophy, I believe that philosophy is not likely to diverge into many separate disciplines that have very little to do with one another, although of course there will always be tendencies in that direction. To have a broad familiarity with the different branches of philosophy becomes more and more demanding. There is intensive activity within all these different branches. As in many other fields of knowledge, I think more work has been done in philosophy during the last twenty years than during the whole earlier history of philosophy. In order to get a satisfactory grasp of the contemporary situation it is tempting to become a specialist of just one field, for example the philosophy of mind. ∗

Lauener-Stiftung / Lauener Foundation for Analytical Philosophy

336 However, since the various philosophical issues are so interdependent, I think philosophy has to remain a unified discipline although it will cost a lot of effort.

Frauchiger: Yet opposite the Scylla of philosophy as a disintegrated discipline we have the Charybdis of philosophy as a self-absorbed, selfcontained armchair discipline, which takes place in the isolated ivory tower and has no relevance to practical aims. Philosophers of course are geared towards exploding this popular myth, pointing out that their discipline does indeed have practical significance and tangible results. Still, what concrete function can philosophy fulfil in reference to specialized scientific disciplines? Dagfinn Føllesdal: First, the fact that philosophy relates to so many other disciplines means that it will be even harder to keep on having some contact with all parts of philosophy. That makes it even more tempting to specialize in just one field. However, there is a second side to the relation between philosophy and other disciplines. Philosophy has to relate properly to what is going on in other neighbouring disciplines. If for example you want to study philosophy of science of course you have to know the science or sciences you work on. But this applies not only to the philosophy of science. If, for example, you want to study perception, you have to study what is going on in neurophysiology, in experimental psychology, etc. And that means that the scope of science and knowledge that a philosopher should know about increases even more. But again it is very important that philosophers study the sciences that have a bearing on their fields. If we sink into armchair philosophy what we do in philosophy will be unsatisfactory. And of course, nobody else will care much about what we are doing. But by relating in the proper way to the various other scientific disciplines I think we can do a much better job in philosophy and also in many cases contribute to the understanding of the issues in these disciplines. Not only for the other disciplines, but for philosophy itself this is very important. I regard the findings of empirical science as a kind of boundary condition that every proper philosophical theory has to fit in with. You cannot

337 develop a philosophy that simply clashes with what we know in the different sciences. Of course, sciences can go wrong. It may very well happen that many of the things scientists believe in today will be given up. But science is the best approximation we have at a given time to knowledge about what the world is like and philosophers should take this into account.

Frauchiger: But can we, by taking the findings of different sciences into account, turn empirical scientific results into philosophical results? Do you think that philosophical subject matters are open to empirical methods in much the same way as fields of scientific research are? Dagfinn Føllesdal: No, I do not think that empirical science can solve philosophical problems. There will always be problems belonging properly to philosophy, and the sciences will not give us answers to these problems. However, the answers we try to give as philosophers should be compatible with and informed by what goes on in the sciences. So this is why I say that the sciences provide boundary conditions for philosophy, but they will never replace philosophy. Many issues discussed in philosophy are highly relevant to sciences but they are not always topics dealt with properly in the sciences. For example, time, space, freedom of the will, etc. – clearly physics, psychology and other disciplines have something to tell us about this, but they do not give us an answer to most of the fundamental philosophical problems.

Frauchiger: So, in your view, the philosophical enterprise is not an empirical, explanatory one? Dagfinn Føllesdal: No. It is not what we’d call an empirical enterprise. It might in some ways be slightly explanatory, because it would try to put the findings of the particular sciences into a broader framework. And that often also helps us understand better what is going on and in that sense it will be explanatory. But, of course, the regular sciences have as a basic purpose to be explanatory, while philosophy is attempting to deal with problems that

338 are more general, common to many sciences and often not properly dealt with in each science.

Frauchiger: Let me recapitulate: Philosophy provides clarification of empirical findings by putting them into a broader practical and methodological framework and that, moreover, facilitates the investigation of foundational issues concerning scientific theories. Dagfinn Føllesdal: Yes. That is one task of philosophy. It is clearly to provide clarification. Partly clarification of how the results in a particular science relate to our ordinary way of experience of our lives. But partly also to throw light on some fundamental issues in the science. And this is mostly when these fundamental issues also relate to other sciences. There it can be necessary to look across sciences in order to see how these issues should be dealt with. However, of course most sciences will themselves also deal with their fundamental issues. And there will not be a replacement but a cooperation between scientists and philosophers on those fundamental issues.

Frauchiger: Yes, certainly, and a scientist may in personal union do some philosophical work or a philosopher may act as a scientist. Dagfinn Føllesdal: I think most scientists are interested in understanding better the world in which we live, and they will often in connection with their scientific activity come out with more general philosophical statements. And of course, you have philosophers who also as part of their doing philosophy come out with utterances about things that are properly scientific. In that way it is of course important that the philosophers know their science. And it is also important that the scientists know their philosophy. It is fairly easy to find out whether somebody who talks about science really knows his scientific field. But it is not so easy, especially for non-philosophers, to see, when a scientist talks about philosophy, whether this has substance to it. And this is unfortunate because there are some scientists who have a tendency to speak very firmly and almost dogmatically

339 about philosophical issues without having read what philosophers have been doing on them. They would never do that in science. They would not go around in their own science and utter dogmatic statements about issues that they have not studied, especially when these already have been studied by other scientists in their field. We have to learn from one another and always look at what has been done in a field before we utter general statements about that field.

Frauchiger: How do you explain that academic philosophy has become much less visible than it once was while at the same time the 20th century has been a period of most intensive and fruitful development in philosophy? Dagfinn Føllesdal: There are clearly a lot of interesting advances that have been made in philosophy in the 20th century. And philosophers know about it. But you wonder about the general public. I think there might be two reasons why philosophy is perhaps not as dominant in our society as it was e.g. in the time of Plato and Aristotle. One is that the sciences and also so much else that is going on occupy people’s minds. And the parts of science and scholarship that get the most attention in the media will of course be achievements within the natural science and technological field that contribute strongly to shaping our daily lives. Sometimes also new insights in the social sciences arouse people’s interest, but for the humanities it is usually difficult to make headlines in that way. A second factor might also be that more and more of the discussion in philosophy has become such that you have to know the issues quite well in order to follow the discussion. And that of course means that for laymen it is harder to follow. This again puts a responsibility on us philosophers to at least spend some of our time communicating what we are doing to non-philosophers. But although the attention given to philosophers is diminishing, there are areas where more attention is given to philosophy now than earlier. One such area is of course the mind/brain problem. As we learn more about the brain and its functioning, we pay more attention to the relation between what we learn about the brain and what philosophers think about the mind – I think that’s an area where laymen are interested in hearing

340 what philosophers have to say. It is also a field where many nonphilosophers come out with opinions that are not well thought through and where philosophers have a great responsibility. Another field where much attention is given to philosophy lately is ethics. There are several reasons for this. One is the many rapid changes in our present society, the developments of new kinds of technology that often change our life situation quite a lot, and also changes in our society, our culture, our economic life, etc. These changes make us face problems that have not been dealt with in the earlier history of mankind and where traditional ethical views that have been transmitted from earlier generations do not give us answers. Earlier, people often did what they had been taught to do and what one had been accustomed to do through generations. They registered that this is what one does, and did not reflect much about these normative issues. However, now one is again and again in situations unlike anything people have encountered before, and one has to reflect and take a stand. What, for example, should one do with atomic power, cloning, genetically modified crops, etc.? This forces one to reflect. Fortunately, these are issues where philosophers are quite active and where many people expect philosophers to be of some help. So mind/brain research and ethics are two fields at least where more attention is given to philosophy than what was traditionally the case.

Frauchiger: So, returning to our previous topic, philosophy provides clarification e.g. of scientific, technological, and moral issues by contextualizing relevant problems and findings and, beyond that, by contrasting and integrating different fields. Such rational reconstruction and systematization contributes to our understanding of science and society in general. It is a characteristic philosophical task, which (notwithstanding “naturalistic” views to the contrary) does not amount to a wholly descriptive, explanatory activity, but rather to a partly normative and relatively a priori undertaking of rational justification (as certain present-day Kantian “transcendentalists” and “criticists” would insist). Do you agree with such a statement?

341 Dagfinn Føllesdal: Yes. Of course philosophy will deal with lots of normative issues. In addition to the ones I mentioned from ethics, there are methodological questions, for example, what are good explanations, what are bad? What is clear, what is unclear, etc. However, I am not seeing such tremendous tensions here as many people see in contemporary philosophy. I think, for example, that what is called “naturalism” contains much that is interesting and positive. But one problem is of course the use of labels: “naturalism”, “transcendentalism”, etc. Like most labels these are vague, and it is very unclear what is meant by them. Such lack of clarity is unfortunately used very widely for polemic purposes. One puts a label on a view that one does not like, often without having even studied exactly what the view is, and then one transfers that label very quickly to lots of other philosophers whom one has not read at all. This technique may give comfort in our time when so much is written and we get time to read only a very small part of it. If we can then write off large chunks of it and say that is such and such an “–ism,” then we feel that we need not worry about it. We do not need to read it, it is all wrong, irrelevant or the like. Labels close people’s mind, and we should get away from using such labels and rather look directly at the issues at hand. Philosophers, in particular, should feel a responsibility here. Use of labels is one of the most widespread tools for spreading prejudices, hatred and violence. Calling somebody a Jew or a Muslim have been much used tools in some of the most tragic events in history. Philosophers have a special responsibility for warning against this kind of label-throwing, and we should certainly avoid use of labels in our own philosophical discussions. The label “transcendendental philosophy” also lumps together many different views and issues. Kant and Husserl, for example, use this label differently. One ingredient in many uses of this label is that one tries to bring to light matters that go on without our knowing about them. This, I think, is something that is important in philosophy. And it is quite clear that, even in as daily an activity as perception lots of processes go on without our being aware of them. Perhaps is this typical of what we are not aware of, that it is something that occurs so often that it becomes habitual – we would not survive if we steadily were aware of it. This side of what we could call “transcendental” is one that I find interesting and important. It is connected with one use of “a priori”: what is transcendental is anticipated

342 in our experience of the world, without our being aware of it or reflecting on it. However, this is of course only one of many notions that many have in mind when they talk about what is transcendental. Transcendental philosophy is often contrasted with “naturalism.” Naturalism is then sometimes taken as a kind of rough materialism. There is nothing in the world except physical matter. That view, I think, is very hard to defend and I certainly do not have that view. But naturalism can also express the view that I hinted at earlier, namely, that the sciences, the study of nature may have something to teach philosophers, e.g. about how we perceive, how we find out something about the world around us. And we should definitely, for example when we study perception, look carefully at what neurophysiologists and experimental psychologists have found out about features of the human perceptual system. In that sense, if naturalism simply means that we should pay attention to the evidence that is provided by the natural sciences, then I would be a naturalist as well.

Frauchiger: But then again, you would for sure want to distinguish between actively positing new norms and describing already existing norms. As is well known, many “naturalists” do not want to make such a distinction. Dagfinn Føllesdal: Yes. Of course some “naturalists” ─ some of those who think that the label applies to them ─ would hold that all you can do, in a scientific way or in an argumentative way, is to describe what the world is like. There is no way of giving arguments for anything normative, they would maintain. You find this view in Hume, to some extent, and you find it in many of the logical empiricists, but you also find it in existentialists. Sartre and Heidegger make normative statements. However, they do not support these statements by arguments and seem to think that there can be no arguments for or against such statements. This becomes like merely stating something and assuming that people will accept it. I am very negative towards this. I think one main feature of good philosophy is that you are not presenting your views like a preacher. You present evidence for your views so that the people you are talking to, can make up their mind in view of this and other evidence. This is an important part of dealing with

343 the other person as a full-fledged human being. We should not attempt to get others to accept a view just because we accept it ourselves. Instead we should try to get them to reflect on evidence and arguments and encourage them to seek for more information when the decision is important and there is time to explore the issue more thoroughly.

Frauchiger: I take it that our having the ability to present evidence in support of our views does not presuppose our proper capability to collect empirical data by doing experiments. Dagfinn Føllesdal: I agree. Normally, collecting data and doing experiments requires considerable competence in that field. It is very easy to go completely wrong in collecting empirical data, so anybody who wants to do that in a reasonable way will have to learn how to do it. One has to study and train and practice in order to become a good experimentalist. If philosophers do that, they should normally be able to carry out experiments. But usually we have to look at what people who are experienced in experimental work find out and base our arguments on that. However, I do know philosophers who carry out experiments. One good example is my Stanford colleague, the first Lauener Prize winner, Pat Suppes, who actually carries out interesting experiments on how the brain works.

Frauchiger: Moving on to another topic, you hold in a way similar to Husserl that the process of philosophical justification in both theoretical and practical subjects ultimately needs to be anchored in sets of intersubjective intuitive acceptances, which are largely pre-reflective and unthematized. Could you elaborate a bit on this view of justification? Dagfinn Føllesdal: Yes, this is a rather complicated thing to explain because through the whole history of philosophy from its beginning until recently the dominant view on how we can provide evidence for a view has been to argue that we need some point which we can be absolutely certain about, and then we have to build from there. Aristotle, for example, held that we could give arguments for something in three different ways:

344 We give an argument for it by showing that it follows from something else and then we ask where did that come from and we go back and back. One possibility would be that we keep on going back ad infinitum, there is an infinite regress. Aristotle did not find this satisfactory. A second way he discusses is that we go back to something, then continue to something beyond that again and eventually get back to where we started, so that we have moved in a circle. He also rejected that. Then he said that a third alternative is to go back for a while until we get to a point where we can be absolutely certain. Typically, in geometry we have axioms, which seem to be perfectly evident. One then progresses from there, proving various theorems from the axioms. Aristotle regarded this as the right way of justifying our beliefs. This might seem to fit geometry and other branches of mathematics. It was also attempted in ethics, for example by Spinoza, because in ethics as well as in mathematics it is very hard to point to anything empirical on which we could base our views. However, in other fields than mathematics it is difficult to get back to a basis that is absolutely certain. Gradually, it has dawned on philosophers that maybe the alternatives are not just certainty or nothing, but there might be standpoints that we can be pretty confident about, but they might not be certain. We are ready to accept that maybe some evidence will come in which will make us change our view. I call this view ‘fallibilism’, using a word from Peirce. And I think the right attitude is to be a fallibilist. That does not mean that we are skeptics, not willing to believe in anything. We have to accept very many positions when we live and act. This is compatible with our being open to the fact that there could be evidence coming in that will make us change our view. Of course, in many areas we are very confident of what we believe in, but we should just never be absolutely certain. One reason I like Husserl is that he reflected carefully on this kind of justification. He is the philosopher I know who has most thoroughly worked out this view. According to Husserl, the basis that we tend to fall back on, is what he called “doxa”, the very simple common views that we make so much use of in our daily life that we take them for granted and hardly reflect on them. Some other philosophers, like G. E. Moore in England, have had similar views. Wittgenstein, too, had very comparable views. However, I

345 find Husserl interesting because he thought through some important problems that neither Moore nor Wittgenstein seem to have thought through properly. This is one reason why it pays to study very carefully what Husserl has to say about justification.

Frauchiger: Could you highlight the part intersubjectivity plays (from a Husserlian point of view) in our assumption of an objective reality, i.e. of a world independent from us, which comprises objectively existing subjects, such as ourselves, who experience inexhaustible spatio-temporal objects transcending their experience? Dagfinn Føllesdal: Husserl’s analysis is extremely good when it comes to the subjective. No philosopher I know of has been so good at analyzing the subjective perspective. And for that reason very many who were impressed by Husserl’s analysis of subjectivity stopped there. They took it over and went into developing extremely subjectivist philosophies. Many of the French philosophers did that. What they have not noticed is that Husserl, in his very first phenomenological work, the Prolegomena, the first volume of the Logical Investigations, which came out in 1900, criticizes that kind of subjectivism and relativism. Husserl shows quite well that it is simply untenable. In his later writings Husserl shows, as you said, that intersubjectivity is very important for our world view. We do not experience ourselves or other selves as isolated subjects who each lives in a world of their own. We are experiencing one another as living together in a world that we all share. Others experience this world from their perspectives and I experience it from my perspective. By learning about the other person’s perspective I can adjust my own. You could say that the more perspectives you get to know the better an idea you get of what this world might be like. Husserl has 6000 pages of very detailed discussions of these intersubjective adjustments, how we get from subjectivity to intersubjectivity. Through this process we are getting an idea of living together in a shared objective world that we are all experiencing from different perspectives. There is a kind of three-step procedure in Husserl: subjectivity – intersubjectivity – objectivity. Here we have one of many similarities be-

346 tween him and other philosophers who have had other labels put on them, e.g. “analytic”. Donald Davidson used as a title of one of his books, written of course long after Husserl: Subjectivity, Intersubjectivity, Objectivity. Just this three step process is so typical in Husserl, and he worked it out in great detail.

Frauchiger: Intersubjectivity thus presupposes empathy, the conscious attribution of a perceptual or other intentional viewpoint – which we are able to simulate – to other persons. The phenomenological method of apprehending other minds regards empathy primarily as a cognitive experiential act rather than as the emotional phenomenon which modern psychologists often focus on in relation to issues of moral motivation and development. Could you clarify a bit further the use of the notion of empathy within the phenomenological tradition? Dagfinn Føllesdal: Yes, I did not use the word empathy, but that is clearly, as you say, a very central notion in Husserl. It is actually the topic of the dissertation of one of his students, Edith Stein. The German word they use is “Einfühlung”. There are related words in other languages: ‘empathy’, but also ‘sympathy’, David Hume, for example, in the English of his time talked about sympathy in very much the same way that Husserl some times talked about “Einfühlung”. As you note, “Einfühlung” does not primarily have to do with feelings. It has much more to do with ability to get into the other person’s perspective. To take it step by step: when we are experiencing the world, then we are experiencing full-fledged things in the world. We are not experiencing sense data or surfaces and then infer that there are things, and we are not experiencing bodies and infer that there are persons. Husserl has good and interesting arguments to show that we are directly experiencing one another as persons. What does it mean to experience a person? It means to experience somebody who doesn’t just have a body but is also experiencing the world. That is, each of us has a perspective on the world. We all experience the world from our different points of view. Since it is the same world we all live in, but which we experience from different perspectives, we can, through communicating with others

347 and finding out what their perspectives are like, get a better idea of what the world is like, which we all experience from different points of view. I may, for example, be colourblind, but by trying to grasp another’s perspective, listening to what others say, noticing how they behave, etc. I can come to see that the world in which they live and I live is a world with colours, but that my sensory organs are not suitable for helping me to make those distinctions. So that is how intersubjectivity is very intimately connected with our experience of other people as persons. And once you get to that, you also come to see that there is one world, which Husserl regarded as the objective world. We all come to this world from our different perspectives, and it is only by assisting one another, by sharing one another‘s perspectives, etc., that we can get closer and closer to understanding more about that objective world in which we live.

Frauchiger: Since the 1950s you have been making pioneering efforts to bring phenomenology and analytical philosophy closer together. Especially you have made clear that Frege and Husserl started from similar problems and proposed parallel solutions. This concerns especially propositional attitudes or rather intentional states, mental states which have content. Similar to Frege’s distinction between sense and reference, Husserl advocates that intentional content, i.e. noema or noematic sense, uniquely determines the intentional object of the mental state, i.e. its reference. However, where does the analogy between Frege’s and Husserl’s respective theories of sense and reference end? Dagfinn Føllesdal: First a little bit about the similarities. You are quite right that in 1954, which is quite a long time ago, I did write a little thing on “Husserl und Frege” in German and I was struck by many similarities between Husserl and Frege. This has turned off some phenomenologists because they kept the old labels and the idea that analytic and continental philosophy are worlds apart, they cannot meet. The idea that there should be similarities just was repugnant to them. But again, my view is: Drop the labels! Sometimes one says that what I do is to try to bridge the gap. Rather I tried to make people see that when one concentrates on the philosophical issues there is not such a gap. The idea of the gap started coming

348 in the 1920s, and partly it is due to Heidegger and his approach to philosophy, partly to his critics, who brought in the political side of it and argued that the kind of philosophy Heidegger was doing was closely related to Nazism. I do not think that the latter point is right, but clearly there was a kind of gap arising then, which unfortunately has been kept alive also through the war and after the war. When I was looking at these things in the mid50s, it was generally assumed that the philosophy on the continent and the philosophy in England and the United States were so different that they had nothing to learn from one another. In this sense there was a big gap. Philosophers from these two camps hardly talked to one another. They did not even read one another, both sides held that this would be wasted time. If there was an appointment to be made they wouldn’t dream of appointing somebody from the other side of the gap. This meant that for a long time there was no communication. What I wanted to show was that philosophers on both sides were really dealing with very similar issues and could learn from another. Not only can one find parallels, but one can also find differences. One such difference that your question brings out in connection with Husserl and Frege, is that Frege was focusing very much on language and logic, while Husserl was concerned with human activities beyond just linguistic activities. Husserl explored these ideas within a much broader field. Husserl was also a little later than Frege, born 11 years later, but they had contact. They corresponded ─ they lived before the gap. They understood that they were dealing with related issues. And this is reflected in their correspondence. What I find interesting in Husserl is that he goes into these broader issues. One sees how there are structures there that might resemble the ones in Frege. But they get a much wider application than they had in Frege. Frege on his side had the great virtue that he wrote extremely clearly. Frege is really a model for philosophers when it comes to trying to express your thoughts and communicate with others. Husserl unfortunately did not write well and that is another reason that the gap developed. It was hard to read him and many decided that they could spend their time better by reading other philosophers. The label-throwing furthered this attitude. He was labeled a “continental” philosopher, after all he was the source of some of

349 the basic ideas in “continental” philosophy. So here we have a good example of how the use of labels closes people’s minds.

Frauchiger: Let us just go a little into the particular question of the parallels between Husserl’s and Frege’s respective theories of meaning: Are Husserl’s noemata, just as Frege’s senses, platonic objects, which are absolutely abstract, atemporal, and invariable? Do such meanings exist independently of their actual or potential role as intentional contents, thus independently of their being grasped by human minds? And if so, how can they nevertheless become objects of phenomenological description and be discovered? Dagfinn Føllesdal: This is a very good question. It relates to one of the main points of disagreement between me and many people who have worked on Husserl and who come from the “anti-analytic” side and who think that Husserl cannot be so similar to an arch-analytic philosopher like Frege. My view is that this noema which Husserl talks about, is a structure very similar to Frege’s “Sinn” or the meaning of linguistic expressions. Very much of what Frege says about the “Sinn” applies also to the noemata including what you mention, that they are abstract entities and can be grasped by different people. They are not private, but can be grasped by others. Many would say: But we cannot grasp abstract entities. However, it was part of Husserl’s philosophy to develop a theory of how we can experience abstract entities. He devotes much attention to how we can experience mathematical objects, numbers etc., and also concepts, that are all abstract. Both Frege and Husserl held that all these abstract objects can be experienced by us. But Frege never worked out a theory of exactly how we can grasp them. Husserl devoted much time and thought to developing such a theory. He held that when we think that we can perceive only physical objects, we fail to note that perception has an intricate structure that it shares with a more general notion of “intuition,” “Anschauung,” whose object need not be a physical object, but can also be an abstract entity. We can be empiricists, holding that all knowledge about objects in the world gets to us

350 through our senses. But it is illegitimate to go from there to concluding that the objects we experience therefore have to be material objects. Husserl’s view on perception as a special case of intuition is in my view very interesting and convincing once one sits down and studies it. Kurt Gödel studied Husserl and was very impressed by this view. He took it over as his own, and in his own philosophical writings he talks in the same way as Husserl about our experience of a mathematical object being very much like a kind of perception. Many people find this view incomprehensible. But if they read Husserl they will see how it is very carefully worked out, and that the burden of argument actually lies on the side of those who have an alternative, more limited view on perception. Frege was a remarkably original and clear thinker. Husserl did not have his ability to communicate clearly and engagingly. However, Husserl had a great problem-sensitivity and on many issues where they both worked Husserl saw problems that Frege had not noticed. This is the case in connection with indexicals, where Husserl anticipated part of the work done later by Kaplan and Perry. Husserl also discussed what he called the “twin world” example, which Putnam two generations later called the “twin earth” problem. (Putnam’s work was independent of Husserl’s, Husserl’s work on this remained unpublished until recently.) In his discussion of the noema Husserl devotes much attention to “the determinable X” as an important component of the noema, intended in part to get around what has later been called “internalism.”

Frauchiger: Let’s turn from the contents to the objects of intentional states. Earlier we touched on Husserl’s account of the crucial role intersubjectivity plays in our experience of an objective spatio-temporal world. We also mentioned his view that perceptual objects are transcendent and feature inexhaustible unobserved properties. Does Husserl’s assumption of the existence of particular reference objects that are perfectly objective additionally require the adoption of a fairly strong naïve realism? Dagfinn Føllesdal: Well, he is a realist in one sense of this label, although his realism is a rather interesting one because he has thought through very carefully some of the arguments against realism. One kind of opposite to

351 realism is idealism, and there was a long period from 1906 on when Husserl said that he was an idealist. However, he then seems to have discovered that he used the word ‘idealist’ in a quite different way from what others did, and in 1934, four years before he died, he wrote in a letter: "No ordinary 'realist' has ever been as realistic and concrete as I, the phenomenological 'idealist' (a word which by the way I no longer use)." 1 The point is that Husserl is not a naïve realist. He is not a realist in the sense that he just takes it for granted that we are experiencing real objects, but he gives a very interesting and careful analysis of what it is to be real. And on that basis he then develops his realism. It is the same with Gödel, who also studied Husserl very carefully and also is a realist in that same sense.

Frauchiger: Another truly groundbreaking contribution of yours to contemporary accurate philosophy has been your development of a firm philosophical basis for quantified intensional logics. At the time of your dissertation, which dates back to the beginning of the 1960s, the metalogical work in this area, especially formal relational semantics, did not suffice as philosophical foundation for modal logics. In addition, above all else, an enhanced theory of reference was needed. Quine’s various objections to quantified modal logics culminated in what was later termed his slingshot argument, aimed at showing that quantification into modal contexts forces all modal distinctions to collapse. You revealed that Quine’s argument, albeit valid, was not sound. Yet, in a nutshell, what kind of insights made for the dismantlement of Quine’s slingshot? And what further pragmaticsemantic problems have presented themselves in consequence of this refutation? Dagfinn Føllesdal: I started to write my thesis in the fall of 1960, when Quine had just published his Word and Object. Quine had for 20 years criticized modal logic because he found that there was a basic lack of clarity concerning the notions of necessity, possibility etc. that it was based on. 1

I owe this reference to Iso Kern: Husserl und Kant: Eine Untersuchung über Husserls Verhältnis zu Kant und zum Neukantianismus. (Phenomenologica 16), The Hague: Martinus Nijhoff, 1964, p. 276, n.

352 He presented one argument after the other, ever stronger, against modal logic. Finally, in Word and Object he had found an argument that showed that, when combined with quantifiers, the notions of necessity and possibility etc. would lead us to a total collapse of modal distinctions. That argument has later been called the slingshot. The word ‘slingshot’ was introduced by Barwise and Perry. When I read the argument in Word and Object I found that it was a very remarkable argument. However, when I started reflecting on it, I saw that there must be something wrong about it because it applies not just to the modalities, but also to basic epistemic notions like knowledge and belief, and basic ethical notions like obligation, permission etc. And it applies even to basic notions in science like counterfactuals, the notion of probability, etc. All of these should collapse according to this argument, so there must be something wrong. Quine always wrote his books and articles in such a way that he did not introduce formalisms unless it was absolutely necessary. Thereby he could communicate with a wide range of people. What one can do if one thinks there is something wrong in an argument is to formalize it. That means that one puts in all the assumptions etc. that are needed for the formal proofs and that are normally not mentioned because they are assumed to be obvious. When one puts in all the assumptions needed for a formal proof, one normally discovers, if the conclusion is wrong, that there is at least one assumption with which something is wrong. The problem with Quine’s argument was that all the assumptions he had made use of were generally accepted philosophical views at the time. Given that the conclusion was unacceptable, this meant that there must be something wrong with at least one of these accepted views. The only one that seemed to be a possible candidate at all was a view that had been dominant in philosophy since Frege. Namely, that all the basic terms in our language, sentences, general terms and singular terms, including descriptions, class abstracts and proper names, have the same kind of semantics, in particular that both sense and reference were preserved when such expressions were substituted for one another. Frege was very satisfied with this unified semantics, and it had been taken over by pretty much every philosopher of language in the 20th century. However, by analyzing

353 Quine’s argument one could see that this unified semantics had to be given up. So I proposed a different way of looking at the semantics of names: Names in language are not behaving like descriptive terms. They behave quite differently. In my thesis I therefore argued that first there must be something wrong with Quine’s argument and secondly – and that was the most important thing – the way to block the argument was to have a new view on how names relate to things in the world. I called names and a few other terms “genuine singular terms,” and argued that they refer to the same object in “every possible world.” Ten years later, Kripke proposed what he called “rigid designators,” defined in the same way. I think that is a good label. ‘Rigidity’ is a good word for the defining feature of these expressions, that they stick to their object. Quine was my thesis advisor, and one might think that he would be opposed to my criticism of his final, clinching argument against the modalities. However, Quine was always very eager to find out what is right and wrong. Much more eager in fact than in being right. So he accepted the thesis and actually changed some of his publications in view of this result. There was never any problem with Quine. I owe my whole attitude to philosophy, and even my decision to go into philosophy to him. If it were not for him, I would have stayed in mathematics and science.

Frauchiger: So the core of your counter-argument against Quine’s slingshot amounts to introducing a new paradigm of semantics where genuine singular terms are to be interpreted in a radically different way from predicates and sentences. Since then you have been developing your two-sorted theory of reference (and meaning) further. Dagfinn Føllesdal: Yes, there have been proposed many different accounts of how these genuine proper terms relate to their objects. Saul Kripke proposed one view which I think is interesting, but not accounting well for reference change. There have been various others. One has proposals like the causal theory of reference and many other views on how these terms relate to their object if it is not via some kind of description. My own view is different from these. I think that our use of genuine singu-

354 lar terms has a normative function. When we use a name or other genuine singular term, we signal to the ones we communicate with that now we want to keep track of that object that the name refers to regardless of how that object could change and how we could be mistaken about it, etc. I may discover myself that I had wrong views of an object but still my name keeps on referring to it because that’s what it does in our language community. The traditional description approach to names would of course tell us that a description is associated with the name, and when the original object changes and does not satisfy the description any longer, while some other object does, then the name begins referring to this second object. According to many views on reference, for example the causal view, sense plays no role in the determination of reference. My own view is that sense and features of objects do play a role but they are not the crucial factor. What is important for us is to keep the reference, and the sense we attached to a name is designed to help us keep the reference. I do not think that I got the final solution, but I think that there are big problems with some of the solutions that have been proposed and that one has to take a look at the exact details of theories that are proposed and see whether and how they are able to overcome the various problems. One such problem that I mentioned is the problem of how to deal with reference change, given that names should, in principle, keep their reference. Gareth Evans found it to be a problem for his causal theory that there are proper names, for example ‘Madagascar’ that up through history have changed their reference. How can this be possible if names are supposed to keep their reference? According to the normative view this would be possible because although we may feel committed not to change the reference, sometimes we and our whole language community get so confused and wrong that a name changes its reference in spite of our efforts. So this is one reason why I call it normative, we can go wrong, but we try not to.

Frauchiger: Modal semantics have been given renewed attention in recent years. What do you personally consider to be the most significant new ideas and developments in present-day modal semantics and metaphysics?

355 Dagfinn Føllesdal: I tend to be skeptical to much of the work that is done in modal semantics because it tends to leave the basic fundamental notions, the ones that Quine complained about, obscure. I think that although Quine’s argument against modal logic in Word and Object was fallacious, his dissatisfaction with the basic modal notions was justified. I think that philosophers up till then had not been able to clarify very well what they meant by the modal notions. This unfortunately also seems to be the case for later work on modalities, so that often one gets a proposal for a formal system which is then supposed to throw light on some issue, but the question is begged because the basic notions are not clarified. One just gets a formal structure, but it does not seem to be anchored properly. There are many fields of philosophy that make use of notions that are in a certain way modal. And it would be very valuable to be able to throw some fundamentally new light on these areas. This can often be done by help of a structure that indicates how these things are interconnected. So, it is not that I’m against that kind of structural work, but I would suggest that people who do it should be very conscientious about trying to explain the starting points.

Frauchiger: As a last point let us turn to ethics. Realist forms of moral naturalism and moral rationalism both entail that objective moral truths can be known, a posteriori and a priori respectively. On the other hand, noncognitivism implies that moral knowledge is impossible because our moral judgements are not capable of being objectively true, since they do not express propositions or describe facts, but are, for instance, expressions of purely subjective emotional attitudes. According to such non-cognitivist views, ethics is thus beyond rationality and objectivity. In contrast, you have been emphasizing the importance of rational arguments that support or reject normative statements in ethics. Could you elucidate this point, particularly as regards “reflective equilibrium”, which, following Husserl and Rawls, you propound as the appropriate method for providing justification of the rationality and objectivity of values in ethics? Dagfinn Føllesdal: There are very many non-cognitivist views. Through most of the 20th century this non-cognitivism was dominant. I mentioned

356 earlier that many of the logical empiricists had such views, but that one also finds it on the opposite end of the spectrum in the existentialists etc. One of the great contributions of Rawls was that he argued very well for the view that normative statements in ethics can be supported by arguments. And he used the label “reflective equilibrium” for his position. The method is similar to what one has been using in the sciences. Actually, Rawls showed that also many of the great ethicists all through the history of ethics have had similar modes of reasoning. Rawls made this kind of reasoning very explicit and he uses it to support his views in A Theory of Justice, that big book of 1971, which was a watershed in the development of ethics. He there proposed a theory of justice which was supported by arguments. He gives, for example, arguments to show that his theory is better than the utilitarian theory. Rawls extends the method that we use in the sciences, the hypothetico-deductive method: one puts forward an hypothesis and sees how it fits in with the material at hand and how it fits in with other hypotheses. Emphasis is on the simplicity of our scientific theory etc. This method was very well known quite early in the history of philosophy, for example in Plato and many of the neo-platonists. And it became very popular through Nelson Goodman and Quine – both of them were studied by Rawls, and he gives credit to them. What is interesting in Rawls, is that he transferred this kind of method to ethics and to normative science.2 Now Husserl also did the same thing, but he did not do it in the clear and lucid way as Rawls. If one reads carefully what Husserl wrote, one sees that he, too, thought that normative judgements can be supported by arguments that are based on evidence in the life-world. Rawls similarly goes back to what he called “considered judgements”, that is for him the ultimate source of evidence. But Rawls himself sometimes noticed that this might not be quite satisfactory as evidence. I think that Husserl also here is good because, again, for him the basis for ethics is intersubjectivity, our ability to experience what is going on in other people, to see things from their perspective. 2

So did, by the way, also Israel Scheffler, in ”On justification and commitment.” Journal of Philosophy, 51 (1954), 180–190, at about the same time as Rawls published his first article on this issue.

357 There is a parallel here between what we do in the theoretical field and what we do in ethics: by seeing things from other perspectives we are led to modify our own original opinions. In ethics our likes and dislikes become transformed to preferences that are shared by the community and ultimately would be regarded as objective. It is by bringing in these perspectives of other people from our own culture and from other cultures and seeing what survives that we can see what is right and what is wrong. I regard this as a rather interesting development of the standard method of justification, which was first used in the sciences, to the field of ethics.

Frauchiger: What kind of objective moral knowledge can be attained by means of this holistic method of justification of rational values? Dagfinn Føllesdal: Husserl maintains, and I agree, that the idea of getting a priori knowledge, in the sense of infallible knowledge, is misguided. Husserl certainly uses the word ‘a priori’, but he uses it for the features that are anticipated in our experience. Those anticipations can go wrong, so therefore what Husserl calls a priori might turn out to be wrong. And that is not how, for example, Kant uses the word ‘a priori’. But in any case, the process in ethics is very similar to what we talked about in the theoretical sciences. In ethics we start with the subjective experience of what matters and what does not matter to us. Even as isolated human beings there are things that matter: We need food, we need shelter, etc., and there are lots of other things that we might prefer. Then by getting to know others and by developing our ability for intersubjectivity we come to see that what matters for us and what we prefer is not always the same as what other people prefer. And we may sometimes come to adjust our preferences accordingly. For example, if I am offered a big piece of cake at a party and I like the cake, I may decide to take a huge piece, leaving little to the others. In the beginning I might not have thought about it, but gradually I come to see that others who experience the same situation have a different view on what the preferences should be, and I may come to see I should change my preferences. Thereby at least we who are in the group, where we tend to be more aware of the feelings of others, may come to change our preferences. The next step is that we are now ex-

358 panding the group. We try to get a grasp of the perspectives of more and more people in very different kinds of circumstances, in very different parts of the world, with different problems, different cultures, etc., and try to incorporate this in our ethical views. In that way we go through the same three steps that I mentioned earlier: we start with subjectivity, go through a process of intersubjectivity and get an idea of there being something objective to be right or wrong about. So, in that sense ethics will develop towards objectivity in a way similar to the way sciences develop.

Frauchiger: You appear rather optimistic about the resolution of conflicts between different social and cultural groups which have mutually incompatible moral core values. Do you think indeed that such conflicts can be resolved by balancing the basic values in contest out, aspiring after total equilibrium? – And if so, can empathy effectively play an important role in the resolution of such omnipresent conflicts between the fundamental moral concerns of different communities? – Or else, if not, could we adopt some kind of intersubjective pluralism whereby fundamental ethical values hold relative to particular communities, which have different outlooks on life, yet might still be able to flourish simultaneously, remaining at variance with each other, but in honourable disagreement? Dagfinn Føllesdal: At first two kinds of problems: First, is it possible to find out what will be right or wrong? And secondly, even if one found it out, would there be any way of implementing it? The last question is mostly a question for not just political philosophy, but also even more for political science, for the study of the political processes, etc. One might study and learn something about what kind of social institutions might be conducive to getting a society that is more in accordance with our norms. Therefore, let us concentrate on the first basic philosophical question, where we philosophers should have a special competence and responsibility: namely is it possible to arrive through good philosophical arguments to one universal, common set of norms, or is all we can and should aim for norms that apply to one group, one culture, sub-culture and not to others?

359 I think that, for all practical purposes, what we should do when we are making important decisions, is to put ourselves in the positions of everybody who is affected by that decision. This need not usually be all human beings (and in some cases some animals). There will be decisions that affect, let us say, only my spouse and myself. There can be decisions that affect only people in my immediate community etc. In such cases there is, of course, no need to bring in a solution that should satisfy everybody, including people far away. So, often the considerations one has to carry out, the aspects one has to bring in, might be those that involve only a few people. What is important is that no aspect should be left out concerning the experiences of anybody who is affected by what happens. This is in my opinion the general criterion. In making a practical decision the first question should be: who will be affected by this decision? We have to be inclusive when we answer this question. It is easy, and convenient, but often severely immoral, to concentrate on our own immediate group and leave out of the picture some people who are actually affected. Not only those directly affected by the consequences, but maybe others would suffer in an indirect way; they all have to be included when we ask who is affected. The second step is to ask how they experience what is happening, all these people who are affected. Here we have to try to put ourselves in the situations of everybody affected and try to grasp their perspectives. Thereby we will get a better and better awareness of what we should do. This is an old idea, it is actually one of the oldest ideas in ethics. It is the Golden Rule: Do to others what you want others to do to yourself. This short formula is not quite satisfactory, it is only a first step on the way to perform this kind of empathy that we talked about, namely trying to grasp the perspective of the other person, how she or he experiences what is happening. We should not just ask: How would I experience this, if it happened to me? We have to ask: How is this likely to be experienced by the persons who are affected by it? We will then act, I hope, on the basis of our answers to these two questions. The main thing is never to leave out of sight anybody who is affected. This unfortunately happens far too often in our society, both on the small, individual level and on the broad political level. We some times do things that affect somebody very badly, because we do not take into ac-

360 count that person or because we simply are not as empathic or as goodwilled as we should be. However, I doubt that two groups, which we assume to be perfectly similar in their ways of experiencing what is happening, will end up with the same decision in all situations, even if they work hard and conscientiously to find out who will be affected by a decision and how they will experience what is happening. In simple and perspicuous situations this may happen, but I regard it as likely that in more complex situations there will be differences between the decisions in the two groups. But I think that if one reflects carefully, brings in a rich variety of perspectives, etc., one might in many situations converge on the same decision. Again there is a parallel with empirical science here: Careful and rich and varied observations increase the likelihood that we will arrive at an outcome that those who are affected will tend to accept. In the natural sciences one sometimes gets theories that fit all the data very well, but when more data come one finds that the theories are unsatisfactory. The same may happen in ethics, here more data usually means the inclusion of some people who are affected by the decision, but who were not taken into account when the decision was made. Such inclusion of more people who are affected often leads to revisions in our ethical judgments. A final little point. In the case of scientific theories one has the much discussed phenomenon of different, but empirically equivalent theories. Given the parallel between scientific theory and ethics, one will have the same situation in ethics. However, if the data are the same, which in ethics would mean that one has taken into full consideration all who are affected by a decision and how they experience it, it seems that differences in the more general, abstract parts of ethical theories should not matter, as long as they prescribe the same actions. Just as in science one may have different, but empirically equivalent theories, so in ethics one may have different ethical theories that prescribe the same practical decisions. In practice, it is the decisions that matter. However, among these equivalent ethical theories, there may be one which unifies the field in a simpler and more satisfactory way. Such a theory might be found more convincing than more complicated theories which prescribe the same actions. They might thereby be better tools for improving our society and get

361 people from different cultural and social groups to work together towards a better and more inclusive society. However, in ethics, as in science, to find just one satisfactory ethical theory is such a big challenge that we philosophers have a lot of work set out for us.

Frauchiger: Thank you very much indeed, Professor Føllesdal.

About the Editor Michael Frauchiger, Dr. phil., has been lecturing in the fields of Philosophy and Methodology of Science with the University of Applied Sciences of Zurich, the Open University as well as the University of Bern and has additionally been an Advanced Research Fellow of the Swiss National Science Foundation. Since its establishment in 2003, he has been the Managing member of the board of trustees of the Lauener Foundation for Analytical Philosophy. His areas of research are Epistemology (incl. Methodology), Philosophy of Language and Logics, Ontology as well as questions at the intersection of Philosophy of Psychology and Philosophy of Morality.