Knowledge, Language, and Interpretation: On the Philosophy of Donald Davidson 9783110322521, 9783110322200

Citation preview

Maria Cristina Amoretti, Nicla Vassallo (Eds.) Knowledge, Language, and Interpretation On the Philosophy of Donald Davidson

EPISTEMISCHE STUDIEN Schriften zur Erkenntnis- und Wissenschaftstheorie Herausgegeben von / Edited by Michael Esfeld • Stephan Hartmann • Albert Newen Band 14 / Volume 14

Maria Cristina Amoretti • Nicla Vassallo (Eds.)

Knowledge, Language, and Interpretation On the Philosophy of Donald Davidson

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

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

United Kingdom, Ire, 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

2008 ontos verlag P.O. Box 15 41, D-63133 Heusenstamm www.ontosverlag.com ISBN 978-3-86838-000-2 2008 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 ISO-Norm 970-6 FSC-certified (Forest Stewardship Council) This hardcover binding meets the International Library standard Printed in Germany by buch bücher dd ag

CONTENTS

Introduction M. Cristina AMORETTI and Nicla VASSALLO

9

Fly Swatting: Davidsonian Truth Theories and Context R. M. SAINSBURY

33

Frege and Davidson on Predication Eva PICARDI

49

Events and Conservativity: Clues towards Language Evolution Massimo PIATTELLI PALMARINI

81

Davidson and Dummett on the Social Character of Language Jennifer HORNSBY

107

Davidson on Epistemic Norms Pascal ENGEL

123

The Place of Ontology in Davidson’s Theory of Interpretation Andrea C. BOTTANI

147

Language and Conceptual Schemes Michele MARSONET

169

Davidson’s Naturalism Mario DE CARO

183

Davidson, Self-Knowledge, and Skepticism M. Cristina AMORETTI

203

List of Contributors

225

INTRODUCTION M. Cristina AMORETTI (University of Genoa, Italy) Nicla VASSALLO (University of Genoa, Italy)

Donald Davidson, one of the most original and influential contemporary philosophers, has made significant and contentious contributions to many different subjects within the analytic tradition: from decision theory to the philosophy of language, from metaphysics to the philosophy of action, from the philosophy of mind to epistemology. Such a wide range of philosophical interests is quite impressive, especially if we consider the rare unity and the systematic character of his writings. In his essays, arguments and ideas are deeply intertwined: they overlap, mutually refer to one another, and often presuppose earlier outcomes. As a consequence, it is sometimes difficult to understand just one particular thesis without considering the whole of Davidson’s work. The main core of this complex and striking system can be identified with the notion of interpretation. In order rightly to appreciate the leading role of the interpreter, it is necessary to analyze some pivotal themes of Davidson’s philosophy.

1. Theory of meaning Davidson is by no means skeptical about the possibility of carrying forward a theory of meaning for natural languages, namely a theory able to specify a systematic interpretation of all the sentences in a given natural language. More precisely, he proposes employing an axiomatic theory. By way of a finite set of axioms we start defining the meaning of the words of a chosen language. Then, through suitable rules of inference, we derive a potentially infinite number of theorems—of the form “s means that p”— which describe the meaning of all the sentences of the language. The main trouble with this strategy is due to Davidson’s rejection of any intentional entity. He actually endorses a rigid extensional approach, fearing that intentionality will involve just the kinds of problems about meaning that a theory of meaning was supposed to resolve. Moreover, he finds no use for

10 meanings as entities and maintains that they do no useful work in semantics. Despite the difficulties, Davidson does not give up on the project of developing a theory to serve as a theory of meaning for natural languages. He needs an extensional way to pair sentences with the world, and a bold hypothesis is that this can be done by a predicate as “is true iff”. Thus, he thinks that it is possible to read what a sentence of a language means from a theory of truth à la Tarski, whose theorems take the form “s is true in L iff p”. Actually truth is an extensional notion, which obviously does not conflict with his overall extensional approach. Hence it is possible to have a theory of meaning for a given language—namely a theory which helps us to interpret the speakers of that language—without appealing to problematic entities such as intensions. The original Tarskian project, whose aim was to analyze truth by way of translation, and therefore of identity of meaning, has been reversed. Davidson, in fact, assumes the concept of truth as primitive (since he thinks it is simple, clear and irreducible), and then he tries to employ a truth theory as the vehicle for a meaning theory. His goal is not to reduce meaning to truth, but to shed light on the notion of meaning by making use of the concept of truth. Nevertheless, if a truth theory has to be used for interpreting natural languages, it faces several difficulties. To begin with, finding a way to translate all the sentences of a natural language into first-order logic sentences is compulsory so long as Tarskian methods are used. Moreover, indexical and demonstrative terms also need to be dealt with. Finally, some further constraints are necessary for a truth theory to be used for interpreting a speaker: such a theory must be conceived as an empirical theory and, for instance, it must be capable of supporting counterfactual claims.1 1

According to Davidson axioms and theorems must be viewed as laws. For example, instances of “s is true iff p” have to be taken not only as true, but also as capable of supporting counterfactuals. The reason is that a sentence such as (1) “‘snow is white’ is true iff grass is green”—which is true but does not support counterfactuals—does not give us the meaning of the sentence “snow is white”. By contrast, (2) “‘snow is white’ is true iff snow is white” is not merely true but also capable of supporting a counterfactual hypothesis. Lawlikeness can then act as a way to filter out true but uninterpretative truth theories. See for instance (Davidson 1967).

11 2. Radical interpretation The project of radical interpretation aims to determine what is needed to interpret a speaker without the help of any bilinguals. According to Davidson, in order to interpret a speaker, the interpreter must possess a theory of meaning shaped like a Tarskian theory of truth and confirmed by the speaker’s external behaviour and further empirical evidence. The problem is understanding if and how it is possible to ascertain the empirical correctness of such a theory of meaning. A radical interpreter—an interpreter who doesn’t know anything about the speaker’s language—must initially discover the speaker’s attitude towards her own utterances, namely whether she holds a sentence true or not in particular circumstances. In other words, the interpreter must have access to the speaker’s “hold-true” attitudes. Holding a sentence true is already a semantic attitude, but according to Davidson it can precede interpretation. For example, the interpreter may know that a subject holds the sentence “Sta piovendo” to be true without having recognized which specific truth it is. However, even if we take the interpreter to know that the speaker holds a sentence as true, he is still not able to determine what the speaker believes and what her actual sentence means. Let us imagine that during a storm the speaker utters “Sta piovendo”: in order to interpret this utterance and find out that it means “It’s raining”, the interpreter must be able to ascribe to the speaker the belief that it is raining. However, in order to ascribe to the speaker the belief that it is raining, the interpreter must know the meaning of the speaker’s words. To put it more generally, let us assume that the interpreter knows that the speaker holds a certain sentence true: if she knew the meaning of the sentence, she could establish what the speaker believes, and, conversely, if she knew what the speaker believes, she could determine the meaning of the sentence. The interdependence of belief and meaning is apparently inextricable. Hence it is crucial to find a way to break up the circle which binds them together. According to Davidson we can do this by defining a general theory of interpretation which could “deliver simultaneously a theory of belief and a theory of meaning” (Davidson 1974a: 144), which could yield “a method for holding one factor steady while the other is studied” (Davidson 1975: 167). The theory is that of radical interpretation, whereas the method is

12 supplied by the principle of charity (that, with some obvious differences, we can already find in Quine2). As Davidson initially formulates it3, the principle of charity compels the interpreter to assume that, ceteris paribus, the speaker’s beliefs are by and large true. To put it another way, the interpreter must presume that the interpretee is mostly right, and thus she must ascribe a great number of true beliefs to the speaker (beliefs that are true from the point of view of the interpreter). For instance, if the interpreter faced the twofold option of interpreting a speaker’s sentence as if it expressed the belief that reindeer have two humps or, alternatively, the belief that reindeer have antlers, she should opt for the latter alternative. More recently4, the principle of charity has been divided into two different strands: the principles of correspondence and coherence. The result above is achieved by the principle of correspondence, which breaks up the circle between beliefs and meanings: the only way out is to hold belief constant, and then solve for meaning. The first step towards interpretation has finally been taken. Actually, the attribution of attitudes to a subject is governed not only by those constraints imposed by the external world, but also by holistic constraints which connect all the speaker’s beliefs with each other. This is what the principle of coherence prescribes. The interpreter must attribute beliefs and other attitudes to the speaker so as to make her out to be by and large rational. Hence the speaker’s system of beliefs must be consistent, must respect the transitivity of preferences, and so on. However, in order to keep the overall coherence and rationality of the interpretee, sometimes it could be necessary to attribute to the speaker either false beliefs or beliefs very different from our own. As Davidson admits charity “is a confused ideal” (Davidson 1984a: xix), since the interpreter’s goal is to maximize, not the agreement between himself and the interpretee, but rather her understanding of the 2

Actually, in Quine’s work we can already find many other ideas, but in terms of syntax (translation), not of semantics (interpretation). 3 In Davidson’s essays it is easy to find several characterizations of the principle of charity which are slightly different from one another (see for instance (Lepore and Ludwig 2005: § 12)). 4 For more details, see for example (Davidson 1999: 343).

13 interpretee. Finally, the overall system of the speaker’s beliefs must result largely true and coherent.5 The interpretations we should privilege are always those which optimize understanding. Contrary to Quine’s conception, the principle of charity is not merely a pragmatic constraint on translation. Davidson actually takes it not to be an option, but a constitutive element of interpretation, since there could be no interpretation at all without it. The two thesis of indeterminacy of translation (interpretation) and inscrutability of reference are accepted but also revisited. The thesis of indeterminacy—according to which it is always possible to obtain different interpretations of a certain linguistic behaviour from the same observational evidence—must obtain, but loses its relativistic character. Different interpretations can be compared to different measuring systems: as nobody would maintain that temperature is a relative concept just because there are various scales (Celsius, Fahrenheit, etc.), it is also absurd to insist that interpretation is relative. In fact, two interpretations can merely differ in the particular way they track the same empirical and holistic constraints. If “what a speaker means is what is invariant in all correct ways of interpreting him” (Davidson 1999: 81), then it is possible to conclude that there is no such thing as “the meaning”. Obviously, not only meaning, but also reference is a semantic concept featuring in a theory of meaning. According to Davidson, however, there is no independent account of what reference is. On the contrary, the concept of reference is actually subordinate to the notion of truth. The hypothesis of inscrutability is conceived as an ontological thesis that, as such, holds strong anti-relativistic potential. Indeed, since there is nothing to render relative, there could be no ontological relativity. Dismissing relativism, in short, does not lead to a stronger notion of reference and ontology, but rather to their complete dissolution.

5

It is possible, then, that the interpreter attributes to the speaker few false beliefs or even limited contradictions. On the contrary, it is impossible that the interpreter attributes to the speaker a great number of false and/or incoherent beliefs. If this happened, the interpreter would be forced to review her own interpretation, and not to opt for the speaker’s irrationality. If the interpreter were faced with a fully irrational speaker, in fact, the interpretation could not begin. Rather, the interpreter could not recognize the speaker as a speaker.

14 According to Davidson, intersubjectivity and interpretability are two pivotal concepts that we must take for granted. In fact, if interpretation were not to be possible, there would be no possibility of analyzing meanings ex parte interpretis, and thus the whole Davidsonian project would simply fail (of course it is not necessary that the subject could always be interpreted in every single aspect and situation). It is worth stressing that the principle of charity leads us to assume that the speaker has a coherent set of beliefs which is largely true, given certain conditions and rational constraints discovered by the interpreter. However, this principle does not by itself guarantee that those very conditions also specify the content of the speaker’s beliefs.6

3. Ontological relativism Davidson is well known for having firmly criticized ontological relativism and the dualism which, according to him, characterize it, that is dualism between conceptual scheme and empirical content (Davidson 1974b). More precisely, on the one hand there is an organizational scheme (which has been identified with the mind, a language, a natural language, a conceptual system and so on), on the other hand there is a neutral content (sense data, impressions, ideas, sensorial stimulations, propositions, mental representations etc.) waiting to be organized and carved up by the conceptual scheme. The crucial point is that different conceptual schemes could organize the very same content in incommensurable ways. And that would imply ontological relativism. Arguing against the very notion of a conceptual scheme, Davidson identifies a conceptual scheme with a set of intertranslatable languages; thus, given two different languages, if we are not able to translate one into the other, then we can recognize two incommensurable conceptual schemes and come up with an example of ontological relativism. Accepting this identification, the notion of incommensurable schemes can be analysed through the notion of non-translatable languages. However, is it really possible to find a language which cannot be translated into our 6

For more details, see for instance (Lepore and Ludwig 2005: § 12).

15 own language? If our attempts to interpret a speaker totally fail (that is, if we are faced with a case of complete incommensurability), Davidson maintains that we have no reason to deem that the speaker actually has a language, nor to suppose that she is a rational creature with beliefs and other propositional attitudes. And, if we are not able to attribute beliefs and other propositional attitudes to the speaker, then we cannot even think that she has a conceptual scheme. Now, let us suppose that our attempts to interpret a speaker only partially fail: have we then found two different conceptual schemes? Even if radical interpretation should be only partially successful, that would still imply that the interpreter has attributed a coherent and largely true set of beliefs to the speaker (coherent and largely true from the interpreter’s own point of view). Thus the speaker’s beliefs are the same as those entertained by the interpreter. In this way the interpreter can determine the meaning of the speaker’s words and create the right space for the emergence of diversities and errors. In order to claim that there are some differences between two conceptual schemes, it is then necessary to extend their common base as far as we can. Given Davidson’s interpretative methodology, it is never certain nor clear whether a particular contrast is due to a divergence between conceptual schemes or more simply to a difference of opinions. To put it another way, there are no elements to establish unambiguously and beyond any doubt whether two conceptual schemes are different or are one and the same conceptual scheme. As a consequence, the very notion of a conceptual scheme loses most of its consistency and plausibility. To sum up, we face two alternatives. On the one hand, the mere fact that interpretation succeeds implies that there is no untranslatability as such; it means that speaker and interpreter share a common ontology with concepts which refer to the same objects in the world. On the other hand, if interpretation fails and the speaker cannot be interpreted, then there is no reason to consider the speaker as a rational creature endowed with propositional thought and language. Given this dilemma, Davidson maintains that the very idea of a conceptual scheme is basically empty and senseless, and that we would be better off without it. Hence, both theories of ontological relativism and ontological absolutism are clearly untenable.

16 4. Language and normativity One of the most controversial theses held by Davidson is that “there is no such thing as a language”, followed by the immediate qualification: “not if a language is anything like many philosophers and linguists have supposed” (Davidson 1986: 107). More precisely, he argues against the idea that it is an essential part of a language that speaker and interpreter on occasions of communication bring to bear syntactic and semantic conventional rules which they have gained in advance and share with each other. In other words, handling a system of prior shared conventional rules is neither necessary nor sufficient for interpretation. From a theoretical point of view, in fact, we can imagine two speakers who do not share the same language (and maybe are not even able to produce the same sounds), and nevertheless they understand each other.7 Moreover, there are countless situations where, in order to interpret a speaker, knowledge of prior shared conventional syntactic and semantic rules is by no means sufficient. The correct understanding of malapropisms, slips of the tongue, and metaphors can only be achieved through appeal to our knowledge of the world, of human interests and attitudes, of the particular speaker and context, and so on.8 The above considerations, however, do not rule out the existence and actual importance of prior shared conventional rules. They simply lead us to refute the pretence of considering such rules necessary and sufficient for interpretation, and to recognize that it is not the sharing of the same meaning which makes communication possible, but rather successful communication which guarantees the sharing of the same meaning.

7

As we shall see, Davidson believes that his theory of triangulation has no need to use concepts like rule or convention (Davidson 1994). 8 Lepore and Ludwig (Lepore and Ludwig 2005), moreover, stress that the interpreter must consider the speaker fallible. Thus, the interpreter cannot presuppose that the speaker knew the conventional meaning of the words he is actually employing. If Lepore and Ludwig are right, then Davidson is justified in maintaining that on any occasion of communication it may always be necessary to enrich knowledge of conventional meaning with further general knowledge and knowledge of the current speaker.

17 Another question is whether on particular occasions of communication we should think that rules governing a language are both prior and conventional, and also shared by speaker and interpreter. Davidson’s answer is no: what is prior and conventional is not shared, and what is shared is not prior and conventional. On the one hand, what speakers bring to a single occasion of communication (their prior theory) is not only different from subject to subject, but also changes with regard to one single subject: it can vary on a future occasion of exchange and according to the particular audience the speaker is addressing. On the other hand, what speaker and interpreter share (the passing theory) is not the precondition, but the result of communication. In short, the idiolect is conceptually primary, whereas public natural languages are only secondary. Obviously, Davidson does not mean to deny the existence of natural languages (conceived as abstractions built up from various different idiolects), but merely to question their conceptual priority. In Davidson’s overall project, there is something prior to all occasions of communication which must be shared by both interlocutors at the very same time for communication to be successful. This essentially social element is the speaker’s intention to be interpreted as she actually intends and expects to be interpreted. This basic intention—forcing the speaker to make herself interpretable as much as possible—explains why those speakers belonging to the same community are apt to make their idiolects uniform, and, more importantly, it creates the right space for the emergence of error. Indeed, error actually consists, not of violating a particular prior shared conventional rule, but of the failure of the speaker’s intention to be interpreted as she actually expected and intended to be interpreted.

5. Events and anomalous monism According to Davidson, adopting events9 in our ontology proves to be very useful to solve some controversial aspects of natural language semantics. 9

Here, the word “event” is used to indicate both a change and a state. In particular, the expression “mental event” is meant to denote also a mental state such as a belief.

18 We looked at the project of using a Tarski-style theory of truth to serve as a theory of meaning. A prerequisite to do that is formalizing all the sentences of the natural language in first-order logic. But, without events, translating sentences containing certain types of adverbial modifiers into first-order logic sentences seems to be an unfeasible task. Accepting events entails admitting the existence of the entities to which some particular expressions refer, expressions like: “her buttering toast”, “his reading a book”, “their passing an examination”, “your holding a belief” and so on. It is worth stressing that sentences do not refer to events: reference to events is the job of some singular terms can. This is a consequence of Davidson’s extensional semantics, which brings us to accept that coreferential expressions should be intersubstitutable salva veritate. Thus, since the reference of a sentence is its truth value, all true sentences would refer to the same event. That is to say that all events are identical, which is clearly unacceptable. Moreover, he believes that events are unrepeatable, concrete particulars located in the space and time. However, it is quite difficult to find a suitable and general principle for their individuation. Pushed by Quine’s effective criticisms, Davidson has definitely given up hope of finding a principle of individuation for events based on causality, and he has finally embraced the view which Quine also advanced: two events are the very same event if they occur in exactly the same space-time zone.10 In spite of these difficulties, Davidson’s theory of events can also help with avoiding some problems encountered when trying to explain actions and causality. For instance, thinking of events as unrepeatable, concrete particulars allows him to distinguish actions (conceived as events) from their descriptions, and causal relations (which connect two single events) from causal explanations (which instead involve descriptions of events). Furthermore, events have a pivotal role in Davidson’s argument for anomalous monism. The thesis of anomalous monism attempts to offer a solution to the mind-body problem which denies the existence of any kind of real entities besides physical ones. At the very same time, though, it aims to preserve the autonomy of the mental and to eschew reductionism, namely the possibility of connecting mental properties to physical 10

See (Quine 1985) and (Davidson 1985).

19 properties through strict psychophysical laws. Therefore, such a theory is a form of monism from the ontological point of view, but (given the mental’s anomalousness) it is a dualistic theory from the conceptual point of view. Davidson is actually looking for a solution to the problem of reconciling the existence of causal connections between the mental and the physical with the hypothesis of the anomalousness of the mental, which denies the possibility of strict laws connecting the mental and the physical. In other words, the problem is how to accommodate three apparently conflicting principles: (1) the principle of causal interaction; (2) the principle of the nomological character of causality; and (3) the principle of the anomalousness of the mental. According to (1), some mental events interact causally with physical events (since on the one hand mental events can be caused by physical events, and on the other hand mental events can also cause physical events, such as actions). In (2), events related as cause and effect must be covered by strict laws. Then, also mental events should be subsumed by such laws. However, (3) states that there can be no strict laws covering the interaction between physical and mental events. More precisely, there can be neither strict psychophysical laws nor strict psychological laws. Notwithstanding the above tension, Davidson believes that all three principles can be held simultaneously if we simply recognize that every particular causally interacting mental event must be tokenidentical to some physical event. It is actually the very same event (hence monism), which can be described using both mental and physical vocabulary. Therefore, (1) and (2) can be held to mean: every mental event can be subsumed by a strict law as it is token-identical to some physical event, and it has a physical description in terms suitable for this subsumption. But even (3) can be held true at the same time, because the mental (the event described in mental terms) is nomologically irreducible to the physical. The relationship which better describes the dependency of the mental on the physical is that of supervenience: the mental supervenes on the physical if a difference at the mental level always implies a difference at the physical level. The notion of supervenience seems to be able to reconcile ontological monism and nomological antireductionism of the mental to the physical.

20 6. Triangular externalism So far, we have left unexplained how the content of our mental states can actually be determined. On the issue, Davidson endorses a kind of externalism—that is “triangular externalism”—and maintains that the content of our mental states constitutively depends on external factors, both causal and social. To introduce triangular externalism, we can examine the nature of beliefs. Davidson claims that (some exceptions apart) the content of a belief11 is objective, namely it is true or false independently of the actual existence of that particular belief and of its subject.12 One can actually have beliefs if one is aware of the objectivity of their content. To put it another way, one must understand that what one believes could be either true or false, or rather that one could have made a mistake.13 Moreover, even though for many beliefs (including abstract beliefs) this additional problem does not arise, it is necessary to explain how experience contributes to determining the empirical content of those beliefs concerning the external world.14 The theory of triangulation has actually been held to deliver both results.

11

Here, the word “content” is used to refer to what is traditionally known as “proposition” or “thought”. Davidson, on the contrary, prefers to identify the content of a belief with an utterance (see Davidson 1989). 12 For instance, the content of the belief “The cat is on the mat” is objective because it is true or false independently of the existence of the subject and of such a particular belief. Even the content of a belief like “I am happy” is objective because it is true or false independently of the fact that I am actually entertaining such a belief and whether I am deceiving myself. 13 According to Davidson, understanding what it is to fall into error means having the concept of objectivity or objective truth. Understanding the idea that we could be wrong, in fact, requires that we are aware of the contrast between truth and falsehood, of the existence of things which are not dependent on us, which can be true or false independently of our thoughts, which are objective. 14 Obviously those beliefs do not exhaust our doxastic system. However, Davidson maintains that they are particularly important because they anchor our propositional thought and language to the outside world.

21 In the most basic cases, what determines—partially at least—the content of a belief about the external world is its “typical” cause. However, there is the problem of narrowing down the cause15 which partially determines the content of a speaker’s belief. Davidson maintains that a single subject is not able to isolate the “relevant” stimulus, i.e. the stimulus she is truly reacting to, the actual cause of the content of her particular belief. The individuation of the relevant stimulus basically depends on the presence of a second subject (namely the interpreter) who shares the same external world with the first, and perceives it in a similar way. Let us consider, for example, the ringing of a bell: a single subject is not able to determine whether the relevant stimulus is at the level of his sensorial receptors, or of the sound waves, or of the bell. When a second “person” comes into the picture, the relevant stimulus can be narrowed down and identified with the external object, which is common to interpreter and interpretee and, as Davidson would say, is where the imaginary lines connecting the interpretee to the world and the interpreter to the world intersect.16 The triangle has finally been shaped: the first side links the external object to the interpretee; the second side links the external object to the interpreter; the third side links the interpreter and the interpretee, and it is represented by linguistic communication. Through the theory of triangulation, Davidson wants not only to explain how beliefs about the external world acquire their empirical content, but also to account for our understanding of the idea that we could be wrong, that things could be different from how we believe them to be. In fact, without interpersonal communication we wouldn’t be able to understand what it is to fall into error; and Davidson’s triangle is just the basic paradigm of interpersonal communication. In the process of triangulation, the interpreter connects the speaker’s responses to stimuli in 15

The great importance of the notion of causality is worth stressing, but also its problematic character. For instance, on the theory of triangulation, many critics stress the difficulty in narrowing down one single cause. See, among others, (Føllesdal 1999), (Lepore and Ludwig 2005), (Pagin 2001). 16 The problem is that of guaranteeing that speaker and interpreter are actually responding to the very same stimuli and then are able to identify the same similarities. What determines the similarity of their responses must depend on the biological similarity of our sensory apparatuses (see, for instance, (Davidson 2001a)).

22 their common external environment. Once several correlations have been established, given a certain response of the speaker, the interpreter will expect the corresponding external stimulus. But, when the interpreter’s expectation fails, when the external stimulus and the typical response are no longer correlated as before, the space for the emergence of error finally comes out.17 Although triangulation is necessary, it is not sufficient in itself to determine the content of our mental states and to create the right space for the emergence of error. In fact, Davidson maintains that the line linking speaker and interpreter must be “thick” enough to amount to a language: when both speaker and interpreter linguistically communicate their own propositional contents to each other, they can finally make judgments about the external world and become aware of the concept of objectivity (i.e. they understand that things could be different from what one believes they are). From an analysis of the theory of triangulation, it is easy to appreciate how triangular externalism aims to reconcile two different kinds of external factors: the causal, environmental element and the social one. The contents of one’s beliefs constitutively depend on one’s causal history, namely on the history of all particular interactions had with objects and events in the outside world. However, causal interactions with the external world can actually contribute to the determination of one’s mental contents only if they are embedded in an intersubjective space, and if one knows how to share such a space with other similar creatures. The “second person” (necessary in determining the relevant cause and explaining the emergence of the concept of objectivity) is thus the social factor that Davidson aims to integrate with the causal, environmental one. The role the interpreter plays within the process for the individuation of one’s mental contents sheds light on another critical element, i.e. holism. Contrary to other brands of externalism, triangular externalism does not allow external factors to totally determine the contents of one’s 17

Obviously, mere divergence between speaker and interpreter is not sufficient to establish who is actually making a mistake. However, it creates the right space where such a target could then be realized (through further linguistic exchanges or the intervention of other subjects).

23 beliefs. Davidson thinks that mental content actually depends on two distinct kinds of relation: the first between content and external world, the second among different contents. To put it another way, the external cause of our beliefs must be determined by taking into account the holistic constraints which bring together our system of beliefs. Therefore, within the theory of triangulation the causal element and the holistic one cannot be separated.

7. Truth In Davidson’s mind the concept of truth is fundamental. As such, it can neither be reduced to other simpler and more basic concepts, nor considered philosophically irrelevant. The notion of truth cannot be defined, but at the most illuminated by showing how it is systematically related to other concepts. Davidson’s own theory is thus quite different from the main traditional theories of truth (the correspondence theory, the coherence theory, pragmatist theories, epistemic theories, the deflationary theory, the redundancy theory and so on). Initially, Davidson develops a coherence theory of truth18, but then, in a second phase, he also recognizes its limits. The main problem is that coherence is not enough to guarantee truth. In fact, many critics have observed that it is possible to imagine many systems of internally coherent beliefs which, however, are different and incompatible with one another. If this is right, we have difficulty in deciding which the true system is just by looking at their coherence. In order to set the limits of coherence theories and to fully understand the notion of truth, Davidson maintains that the best one can do is to analyze how the concept of truth is related to the concept of belief. On the one hand, he thinks that the notion of truth can only emerge in an interpretative space (where beliefs are obviously involved). Actually, through the theory of triangulation one becomes aware that what one believes could be either true or false, that how things are outside oneself is independent of what one believes. On the other hand, having beliefs and 18

See (Davidson 1983).

24 other propositional attitudes with a specific content depends on the fact that one possesses the concept of objective truth. In Davidson’s perspective, in fact, if one does not grasp the concept of objective truth, one cannot even possess beliefs or any other propositional attitude. Given these relations, and then reflecting on the “nature” of belief, Davidson argues that a coherent set of beliefs must also be largely true. What has been said about radical interpretation implies that a coherent set of beliefs must be largely true from the interpreter’s point of view. In order to interpret a particular subject, in fact, one must apply the principle of charity, attributing to the speaker a system of beliefs which is by and large coherent, and simultaneously interpreting the sentences the speaker holds true as actually true (exceptions apart). The mere fact that interpretation succeeds guarantees that—from the interpreter’s perspective—the speaker’s beliefs are largely true. If they were not, interpretation could not even begin. From the interpreter’s point of view, therefore, a mostly coherent set of beliefs must be largely true. Only after a common base has been established can one also attribute errors to the speaker. Since radical interpretation does not guarantee that a mostly coherent set of beliefs must also be largely true tout court, Davidson considers the theory of triangular externalism. According to this theory, what determines the content of one’s beliefs, at least in the most basic cases, is their “typical” cause. At this basic level, when a subject is acquiring her first language there is no room for error, because the concept of objective truth emerges together with propositional thought. The very fact that we have beliefs and other propositional attitudes is therefore sufficient to assure us that there is an external world, that there are other creatures sufficiently similar to us, and that the world cannot be very different from how we believe it to be. Now Davidson has finally demonstrated what he wished: given the externalist nature of mental content, a mostly coherent set of beliefs must be largely true.

25 8. Self-Knowledge Davidson recognizes three distinct but interconnected kinds of propositional knowledge: knowledge of the propositional contents of one’s own mind (“I know that I desire to go on holiday”), knowledge of the external world (“I know that the cat is on the mat”) and finally, knowledge of the propositional contents of other minds (“I know that Emily desires to go on holiday”). In his terminology they are called, respectively, subjective, objective and intersubjective knowledge. Subjective knowledge is “not based on inference or evidence” (Davidson 1990: 203), i.e. we do not need proofs, observations or any further particular knowledge in order to know what we believe.19 Objective knowledge, by contrast, causally depends on our sensory apparatus, namely on evidence and observations. Finally, intersubjective knowledge almost completely depends, not only on observation, but also on linguistic communication. Davidson firmly believes that none of these kinds of propositional knowledge is fundamental, and thus he refuses any attempt at derivation. There is not one kind of knowledge the others can be derived from, nor a pair from which one can arrive at the third. Not even subjective knowledge is independent or has priority over the other two kinds of knowledge. Assuming the truth of triangular externalism, in fact, we must recognize that we could not know the propositional contents of our own mind if we did not communicate with other human beings in a public external world (and thus if we did not know the external world and the propositional contents of other minds). That is to say, objective and intersubjective knowledge cannot be derived from subjective knowledge.20 19

It is worth mentioning that there is a tendency to characterize knowledge of the contents of one’s own mind as a kind of “direct” knowledge to emphasize that it is not based either on inferences or on causal interactions with the external world. This terminology, however, can be misleading and create confusion between subjective knowledge and direct non propositional knowledge (“I know To the light house”, “I know Emily”, “I know my faults”). 20 A similar argument can be developed to demonstrate that each kind of knowledge needs the others. The reason is that, accepting triangular externalism, the very possibility of having propositional thoughts depends on linguistic interactions with other human beings in a common external world.

26 Nevertheless, Davidson maintains that there is an evident asymmetry between, on the one hand knowledge of one’s own mental states, and on the other knowledge of the external world and knowledge of the contents of other minds. As we have said, what one knows about one’s own mental contents depends neither on inferences nor on evidence. But, more importantly, knowledge of the contents of one’s own mind has a special presumption of truth, an authority which is not shared by the other two kinds of propositional knowledge: if a speaker sincerely affirms that she believes that p, there is a presumption that she knows she believes that p. Apart from exceptional cases, it would be absurd to question one’s selfattributions. Obviously, this does not mean that one is omniscient or infallible regarding the knowledge of the content of one’s own mind. Davidson believes that first person authority can be explained without appealing to the notions of introspection, privileged access or a priority21, but is to be explained rather by reflecting on the very nature of interpretation. The subject knows what she believes, not because she has privileged access to her own beliefs or she knows them by introspection or a priori, but because it is an essential presumption of interpretation. In fact, if we could admit that the speaker can systematically mistake what she believes, then interpretation could not even start. That is a consequence of the interpretative procedure regarding radical interpretation: the interpreter knows that the speaker holds a particular sentence true, and that holding a sentence true depends both on what the speaker believes and on the meaning of the sentence. In order to start interpretation, the interpreter must presume that the speaker knows she is holding a particular sentence true and—according to what we have just said—that the speaker knows what she believes. First person authority is “saved” thanks to interpretation. In short, if we accept Davidson’s interpretative procedure we must also admit first person authority, because if we rejected first person authority we would be forced to refute the interpretative procedure too.

21

Here, the notion of a priori knowledge does not imply any metaphysical necessity. A priori simply means “independently of empirical investigation”.

27 It follows that first person authority does not derive from the special characteristics of subjective knowledge. Rather, it depends on interpretation and on the particular character of self-attributions.

9. Summaries of essays In “Fly swatting: Davidsonian truth theories and context”, Mark Sainsbury analyzes some difficulties raised by context-dependence which may worry a Davidsonian truth-theorist. Sainsbury concentrates on two kinds of difficulties. On the one hand, indexicality compromises Davidson’s project raising doubts about the possibility of isolating specific semantic knowledge from any other knowledge. It actually forces a departure from the ideal of homophony, and thus from the hope of deriving the truth conditions of any particular utterance on the mere basis of axioms whose ambition is to characterize solely semantic knowledge. On the other hand, context-dependence raises some technical questions about the analysis of specific idioms like, for instance, “grunt”. In the first case, Sainsbury shows how exactly semantic and non-semantic knowledge work together to determine the true conditions of the familiar indexical expressions. In the second, he argues that there is no “general method” for bringing all forms of context-dependence within Davidson’s project, and that consequently a piecemeal examination of various particular cases is actually required. In “Davidson and Frege on Predication”, Eva Picardi confronts Davidson’s and Frege’s theses on predication. According to Davidson, in order to understand the unity of the proposition it is necessary to link the role of predicate expressions to the concept of truth, not to ontology, nor to the characteristics of the attitude of judgment. Following Tarski and Quine, Davidson points out that the function of predicates is to be true of objects, and in order to explain the unity of the proposition, no entity can correspond to a predicate. Even though this point is something he has in common with Frege, Davidson does not accept the Fregean approach to predication. Picardi shows that—notwithstanding Davidson’s criticisms— Frege’s thesis, according to which concepts are particular kinds of functions, is not only correct, but also the only possible solution to the

28 problems of the regress of copulas. At the same time, moreover, Frege’s thesis offers a unique explanation for both first and second level predication. Picardi maintains that appealing to functions as rules of correspondence between arguments and values is essential in explaining how quantifiers (generalized) work. Frege was one of the first to recognize that the so-called “paradox” of the concept “horse” that he actually invented, is spurious and not present in Ideography. Those who currently refer to the paradox try to demonstrate the questionable thesis of semantic ineffability in the framework of Frege’s (and Wittgenstein’s) theories. Affirming that certain fundamental notions we apply do not fit into an authentic definition, but are only an elucidation, is perfectly compatible with the attempt to give a systematic explanation of how the truthconditions of the sentences of a language systematically depend on the semantic value of the expressions which compose them. In “Events and Conservativity: Clues towards Language Evolution”, Massimo Piattelli Palmarini shows how Davidson’s original theses have deeply influenced and enriched, not only philosophy, but also other important disciplines, like for instance cognitive sciences. More specifically, Piattelli Palmarini analyzes the crucial progress made by the contemporary semantics of natural language thanks to the deployment of the concept of event, i.e. thanks to Davidson’s classical idea that words and expressions of natural languages refer to and quantify events. Within this general, and mainly Davidsonian framework, Piattelli Palmarini then focuses on the critical puzzle of the conservativity of natural language determiners (among which quantifiers are a remarkably interesting subcase), and on the specific consequences that conservativity could have for the reconstruction of language evolution. In “Davidson and Dummett on the Social Character of Language”, Jennifer Hornsby critically discusses what the social character of language could be. As argued by Dummett, and contrary to Davidson’s main thesis, Hornsby believes that to fully appreciate the social character of language it is necessary to admit that when a language is used, a shared language is used, whose norms determine the meaning of the language’s sentences. Notwithstanding the fact that Hornsby defends a Dummettian view of language, she doubts whether Dummett’s own arguments against Davidson are effective. Moreover, she also claims that dispensing with the

29 whole Davidsonian project cannot be a successful strategy to clarify and understand what constitutes the essentially social character of language. Hornsby’s argument is that we need to take material from both Dummett and Davidson to determine what constitutes the essentially social character of language correctly. In particular, we need to use the Davidsonian idea of a communicative intention alongside the idea of a shared language. Pascal Engel, in “Davidson on Epistemic Norms”, explains why Davidson’s overall philosophy can be characterized as a kind of normativism: the mind has essentially normative properties which can be identified with the norms of rationality in the process of interpretation. Firstly, Engel analyzes how Davidson investigates the question of epistemic normativity, and then he also outlines a more precise picture of Davidson’s own normativism. This close examination highlights two major problems: first of all, Davidsonian norms do not have any normative force, that is they do not have any power of prescribing how we ought to act or believe; secondly, they lack specificity too, and thus they do not allow us to make all the particular distinctions we actually need. Engel concludes that Davidsonian norms are not norms at all, but rather idealized principles of interpretation. Therefore he suggests an alternative conception of epistemic norms, namely a stronger and more fine-grained conception. In “The Place of Ontology in Davidson’s Theory of Interpretation”, Andrea Bottani explores some ontological consequences of Davidson’s theory of radical interpretation. Bottani observes that in Davidson’s approach there is a deep connection between matters of language and matters of fact, between semantics and ontology. As a result, on the one hand the interpreter’s ontology shapes radical interpretation, and on the other successful interpretation becomes a condition of ontological acceptability. After having clarified how semantics and ontology are actually interconnected within Davidson’s theory of radical interpretation, Bottani shows that such a connection does not depend solely on formal constraints as the assumption of a Tarskian theory of truth, but is due also to the regulative principles of the theory of interpretation, basically to the principle of charity. On these premises, Bottani aims to demonstrate that the alleged role of ontology in radical interpretation conflicts with another central theoretical aspect of Davidson’s theory, that is with the thesis of the inscrutability of reference.

30 Michele Marsonet, in “Language and conceptual schemes”, examines Davidson’s critique of the notion of a conceptual scheme. Following the analytical tradition and its linguistic approach, Davidson identifies a conceptual scheme with a set of intertranslatable languages, and then he tries to demonstrate that a language, for it to be considered as such, must be translatable into our own language and also associated with our own ontology. Given those premises, it is quite straightforward to conclude that there are no incommensurable conceptual schemes, and that the very idea of a conceptual scheme is completely meaningless. Marsonet firstly compares the Davidsonian view of a conceptual scheme with the notion of a “conceptual apparatus”, and then he shows the necessity to modify and enrich the criteria through which conceptual schemes are traditionally identified. More precisely, Marsonet claims that the supremacy of language should be abandoned, in order to adopt a less rigid and more articulated notion of a conceptual scheme. As the title of the article, “Davidson’s Naturalism”, already tells us, Mario De Caro analyzes the original kind of naturalism held by Davidson. This naturalism is chiefly expressed in the theory of anomalous monism, a conception of the mental which aims to reconcile ontological monism and antireductionism. More specifically, De Caro examines whether it is possible to accommodate the token identity thesis, the supervenience of the mental on the physical and semantic externalism within the theory of anomalous monism. First of all, he analyzes the various notions of supervenience which can be found in Davidson’s work, to then show that there has been a major shift away from the hypothesis of individual supervenience (according to which mental states supervene solely on brain states) to that of global supervenience (which requires, at the subvenient level, physical states different from brain states). On the one hand, De Caro emphasizes that, given Davidson’s triangular externalism, such a shift is necessary; on the other, he also points out that numerous difficulties stem from it. In “Davidson, Self-Knowledge, and Skepticism”, M. Cristina Amoretti aims to evaluate whether Davidson’s triangular externalism is or is not affected by one of the most famous objections raised against externalist theories on mental content: the Reductio Argument. According to these critics, combining externalism with certain characteristics of self-

31 knowledge leads to overly strong and highly counterintuitive anti-skeptical consequences, and thus to the refutation of externalism. First of all, Amoretti argues that triangular externalism is not affected by the objection raised by the Reductio Argument; secondly, she outlines some consequences for the anti-skeptical argument that Davidson hopes to derive from his own theory. Her conclusion is that triangular externalism is not an authentic answer to the skeptic about the external world. Thanks to their heterogeneity, the nine essays in this volume offer a clear testimony of Donald Davidson’s authority, and they undoubtedly show how much his work—even if it has raised many doubts and criticisms—has been, and still is, highly influential and significant in contemporary analytical philosophy for a wide range of subjects. Moreover, the various articles not only critically and carefully analyze Davidson’s theses and arguments (in particular those concerning language and knowledge), but they also illustrate how such theories and ideas, despite their unavoidable difficulties, are still alive and potentially fruitful. Davidson’s work is indeed an important and provocative starting point for discussing the future progress of philosophy.*

References Davidson D. (1967), “Truth and meaning”, reprinted in (Davidson 1984b), 17-42. Davidson D. (1974a), “Belief and the Basis of Meaning”, reprinted in (Davidson 1984b), 141-154. Davidson D. (1974b), “On the Very Idea of a Conceptual Scheme”, reprinted in (Davidson 1984b), 183-198. Davidson D. (1975), “Thought and Talk”, reprinted in (Davidson 1984b), 155-170. Davidson D. (1983), “A Coherence Theory of Truth and Knowledge”, reprinted in (Davidson 2001b), 137-153. *

We wish to thank Marcello Frixione for reading a draft of this introduction and giving us precious advice, both on theoretical and stylistic matters.

32 Davidson D. (1984a), “Introduction”, in (Davidson 1984b), xv-xxiii. Davidson D. (1984b), Essays on Truth and Interpretation, Oxford University Press, Oxford. Davidson D. (1985), “Reply to Quine on Events”, in: E. Lepore and B. MacLaughlin (eds.), Actions and Events: Perspectives on the Philosophy of Donald Davidson, Basil Blackwell, Oxford, 172-176. Davidson D. (1986), “A Nice Derangement of Epitaphs”, in: E. Lepore (ed.), Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, Basil Blackwell, Oxford, 433-446. Davidson D. (1989), “What is Present to the Mind”, reprinted in (Davidson 2001b), 53-68. Davidson D. (1990), “Epistemology Externalized”, reprinted in (Davidson 2001b), 193-204. Davidson D. (1994), “The Social Aspect of Language”, in: B. McGuinness and G. Oliveri (eds.), The Philosophy of Michael Dummett, Kluwer, Dordtecht, 1-16. Davidson D. (1999), “Replies”, in: E.H. Hahn (ed.), The Philosophy of Donald Davidson, Open Court, Chicago, Ill. Davidson (2001a), “What Thought Requires”, in: The Foundations of Cognitive Sciences, J. Branquinho, Oxford University Press, Oxford, 121-132. Davidson D. (2001b), Subjective, Intersubjective, Objective, Oxford University Press, Oxford. Føllesdal D. (1999), “Triangulation”, in: E.H. Hahn (ed.), The Philosophy of Donald Davidson, Open Court, Chicago, Ill., 719-728. Lepore E. and K. Ludwig (2005), Donald Davidson. Meaning, Truth, Language, and Reality, Oxford University Press, Oxford. Pagin P. (2001), “Semantic Triangulation”, in: P. Kotatko, P. Pagin and G. Segal, Interpreting Davidson, CSLI Publications, Stanford, Ca., 199212. Quine W.V.O. (1985), “Events and Reification”, in: E. Lepore and B. MacLaughlin (eds.), Actions and Events: Perspectives on the Philosophy of Donald Davidson, Basil Blackwell, Oxford, 162-171.

FLY SWATTING: DAVIDSONIAN TRUTH THEORIES AND CONTEXT R. M. SAINSBURY (University of Texas at Austin, USA)

1. The ideal of homophony One key idea in Davidson’s approach to language is that the output of a semantic theory should focus on interpretation. To interpret an utterance is to redescribe it in a way that will give an ideally correct and complete answer to the question “What did so-and-so say?”, asked by an ordinary language user in a typical context. Normally this redescription will be one that the utterer herself should and can accept, and so will not involve concepts (for example the concept of a set or infinitary sequence) which need not be exercised in competent use. This immediately distances Davidson’s project from one which takes as central the provision of revealing descriptions of what individual words mean. No doubt for some purposes it is useful to say of the word “snow” that it refers to atmospheric water vapour frozen into ice crystals and falling in light white flakes or lying on the ground as a white layer. But one who says “Snow is white” does not say that atmospheric water vapour frozen into ice crystals and falling in light white flakes or lying on the ground as a white layer is white; he says that snow is white. Davidson’s project valorises “homophonic” interpretations, ones in which the very same words are used in the interpretation as were used in the utterance to be interpreted. Is this is a missed opportunity? We should recall, first, that all languages can be expected to have primitive expressions, ones whose meaning cannot be specified by using others (colour words are obvious examples); and, second, that Davidsonian interpretations have to be derived from wordrelated axioms “by logic alone”, a demand which provides a quite specific perspective on what it is to reveal the compositional mechanisms of a language. Homophony is sufficient but not necessary for interpretiveness. Its sufficiency is illustrated by the fact that an utterer of “Snow is white” cannot demur at being reported as having said that snow is white. Its lack

34 of necessity is illustrated by the fact that one can correctly report an utterance using a language the utterer does not understand, thereby describing her utterance in a way she would not recognize as correct. A monolingual Italian who says “La neve è bianca” thereby says that snow is white, even if she cannot recognize as correct this non-homophonic description of what she said. Indexicality itself shows that homophony is not necessary for interpretiveness. If you utter “I am hungry” you do not say that I am hungry but that you are. Interpretation seems to bring together two kinds of knowledge. There is standing knowledge about the “meaning” of “I”, and occasion-specific knowledge of who uttered the sentence. Both are required for interpretation, though the latter kind of knowledge is not the property of a special language module. It belongs to quite general cognitive abilities (as is perhaps more conspicuous in the case of “that”, where all kinds of information can be used in discovering the referent). This means that, assuming Davidsonian theory delivers a full semantic account, semantic information is not sufficient for interpretation. The upshot is that Davidson’s project is doomed, if it is construed as using truth theory to organize information knowledge of which would suffice for interpretation. But we can broaden the Davidsonian perspective: think of semantic knowledge as sufficient for interpretation only in combination with non-semantic knowledge. We still might be able to use Davidson’s approach both to constrain what counts as interpretation, and to do full justice to the specifically semantic aspects of language use. That is the optimistic idea I will pursue in this paper.

2. Classical indexicality Davidson said that indexicality is a “very large fly in the ointment” (Davidson 1967: 33), but he made some preliminary suggestions about how it might be handled truth-theoretically. Here are two: “I am tired” is true as (potentially) spoken by p at t if and only if p is tired at t.

35 “That book was stolen” is true as (potentially) spoken by p at t if and only if the book demonstrated by p at t is stolen prior to t (Davidson 1967: 34). These are presumably schemata designed to indicate the form that appropriate T-theorems should take. They do not as such indicate the axioms from which such theorems might be derived. We cannot accept the schemata. An instance of the second is: “That book was stolen” is true as (potentially) spoken by Davidson at noon on 02/02/02 if and only if the book demonstrated by Davidson at noon on 02/02/02 is stolen prior to noon on 02/02/02 (Sainsbury 2005: 54). This is not interpretive. The report draws on concepts, like the time and date, which an utterer of the target sentence would have no need to exercise in a fully competent utterance. We start to address these problems, while also ensuring a clear demarcation of the province of semantics, by thinking of our goal as the provision of conditional rather than absolute truth conditions, as envisioned by Higginbotham, for example If u is an utterance of “today is July 4” by s, and s refers with the utterance of “today” therein to δ, then u is true ↔ δ is July 4 (Higginbotham 1994: 94). We still have uninterpretive results. Start with this instance of Higginbotham’s schema: If u is an utterance of “today is July 4” by Sally, and she refers with the utterance of “today” therein to July 3, then u is true ↔ July 3 is July 4. If we are to use this for interpretation, we will presumably use the antecedent to derive:

36 In uttering “today is July 4”, Sally spoke truly iff July 3 is July 4 and then report Sally as having said that July 3 is July 4. But one who says this is manifestly irrational, whereas one who uttered what Sally uttered may not be. We therefore need to refine Higginbotham’s suggestion, in terms of what I have called the scene/content structure. Intuitively, we want to report Sally along these lines: On July 3, Sally said that it was July 4. The “it” depends anaphorically upon the interpreter’s use of “July 3” outside the content reported. The report is entirely neutral about how Sally referred to July 3, and so is consistent with her having uttered something expressing a belief she could rationally hold. We could implement this by replacing the second occurrence of δ in Higginbotham’s schema by “it” (perhaps with subscript δ). Then, applying obvious background information, we might end up with a redescription of Sally’s utterance as follows: Sally uttered “today is July 4”, and by her use of “today” in that utterance referred to July 3. So what she said is true iff it [July 3] is July 4. This is near enough to the intuitively desired form of report. This approach relinquishes the idea that I can always make myself an absolute samesayer with someone whose speech I correctly report. I may have to do some scene-setting (e.g. “Sally, speaking on July 3, said that…”) before I can introduce the content-specifying part (“it was July 4”), and this is not a stand-alone specification of a content, but rather one that essentially looks backwards, through anaphoric dependence, to the scene-setting. A Davidsonian supplies user-instructions along with his truth theory. In the unreformed version, they go something like this: if you encounter an utterance u of a sentence s by a speaker a, check the truth theory for a canonical theorem for s; if it says that s is true iff p, conclude that a said that p. Reforming these instructions for the proposed revision of

37 Davidsonian truth theory, they become rather more complex, on the following lines: Derive a conditional truth condition for s. It will be along the lines: “If in an utterance u of s the speaker referred to objects x, y, and z then u is true iff …x …y … z…. Determine each of the referents, and select corresponding expressions x*, y* and z* (constants, not variables) to refer to them. Take an instance of the conditional truth condition in which the starred constants replace the corresponding variables in the antecedent, and suitably related anaphoric pronouns replace them in the right hand side of the biconditional consequent. The result will be along these lines: “If in an utterance u of s the speaker referred to objects x*, y*, and z* then u is true iff …itx* …ity* … itz*…. Delete the “If”, the left hand side of the biconditional consequent, and the biconditional sign (“iff”); replace the “then” by “he/she thereby said that”. The result will have the form: In her utterance u of s referring to x*, y* and z* she thereby said that …itx* …ity* … itz*…. If all goes well, this should be an interpretive report. The conclusion is that the spirit of Davidsonian truth theory can accommodate at least these kinds of cases of indexicality, and in the process makes an intuitively plausible division between genuinely semantic knowledge (represented by the conditional truth conditions) and the collateral knowledge needed to attain interpretations.

3. Radical Contextualism against truth conditional semantics Charles Travis has made the bold claim that no sentence completely expresses a thought, so no sentence’s content can be captured truth theoretically: call this Radical Contextualism. He says that it: blocks truth conditional semantics. For suppose I say, ‘The sentence “Sid grunts” is true iff Sid grunts’. Either I use that last “grunts” on some particular understanding of being a grunter—one understanding

38 among many—or I do not. If I do, then I assign the sentence a property it does not have. For it does not speak of being a grunter on any special understanding of this. But if I do not, then I fail to state any condition under which anything might be true. Being a grunter on no particular understanding of being one is just not a way for Sid to be (Travis 2006: 47-8). Sentences containing “grunt” do not completely express a thought, for the words do not settle how “grunt” is to be understood—what sort of behaviour is to count as making one a grunter (is one little grunt enough or does one have to be a habitual or serial grunter?). This is resolved in different ways in different contexts. (Compare Travis’s examples in earlier work: a brown leaf painted green may count as green in some but not other contexts.) The first horn of the dilemma for truth theory is that we treat the right side of the biconditional which specifies truth conditions as completed by context, so that it completely expresses the thought that S; then the biconditional will falsely say that every utterance of “Sid grunts” is true iff S. Alternatively, if the right side is not completed by context, we do not have a genuine biconditional: “I fail to state any condition under which anything might be true”. In his early paper (Davidson 1967), Davidson addressed indexicality by quantifying over persons, times and places. This enabled differences in persons, times and places to lead to different T-theorems. Might one not extend this idea, and deal with context-sensitivity at a single blow by quantifying over contexts? As a first attempt, we might aim at T-sentences on these lines: s is true as uttered in C iff: in C, p. The sentence in the slot marked by “p” translates the sentence referred to by what is in the slot marked by “s”. We might initially be encouraged by examples like this: “All beer is good” is true as uttered in Australia iff: in Australia, all beer is good.

39 This may sound true, at least given various simplifying assumptions. But any semblance of a serious contribution to our problem is illusory. The biconditional does not even begin to do proper justice to sensitivity to context. It is derived from a schema which also delivers: “All beer is good” is true as uttered in Sydney iff: in Sydney, all beer is good. Since an utterance in Sydney is also an utterance in Australia, unrestricted application of the schema will yield distinct and potentially conflicting truth conditions for the same unambiguous utterance. There is a distinct problem, one which confronts an aspect of Borg’s recent defence of truth theoretic semantics. Discussing seemingly monadic versions of expressions which also have an explicitly relational form (“ready” [ready for], “married” [married to], “raining” [raining at], “continue” [continue doing such-and-such]) Borg suggests truth conditions on the following lines: If u is an utterance of ‘Jane can’t continue’ in a context c then u is true iff Jane can’t continue something in c (Borg 2004: 230). Bill and Sally are talking about Jane; take this fact as an instantiation on the implicitly universally quantified variable c. Bill is trying to communicate to Sally that Jane cannot continue with the marketing research project she started a month earlier. Borg’s generalization yields: If u is an utterance of ‘Jane can’t continue’ in a context in which Bill and Sally are talking about Jane then u is true iff Jane can’t continue something in a context in which Bill and Sally are talking about Jane. On all likely scenarios, Jane isn’t doing anything in the Bill-Sally context (they are talking about her in London, and she is asleep in California), and if she isn’t doing anything in it there’s nothing she can continue doing in it. So the right side will be false on all likely scenarios, whereas the left side might be true. The problem here is that what we need on the left side is the conversational context, whereas what we need on the right side is some

40 activity which that context determines. It would be surprising were we to get both these different things using two occurrences of a single variable; we certainly could not count on any such coincidence. If we reconsider Travis’s dilemma for truth conditional semantics, it may seem we have material which would lead to just such a surprise. Concerning a putative T-theorem “Sid grunts” is true iff Sid grunts, the dilemma was that either the right side is taken in a way divorced from any specific understanding of what it is to grunt, in which case it does not provide a genuine condition at all, or else a specific understanding is in play, in which case it says falsely that this is the only way for “Sid grunts” to be understood. “Understandings”, as Travis uses the term, relate both to sentences (like “Sid grunts”) and to ways things can be (like being a grunter), so perhaps it is a notion which can effect just the kind of transition needed by Borg’s project. (This would be ironic, for Travis was out to attack truth conditional semantics, whereas Borg defends them.) An understanding of a sentence is some kind of additional constraint on what it takes for the sentence to be true. An understanding of a kind of event or state of affairs is an additional constraint on what is involved in its obtaining. (We do not need to insist that this notion of understanding is entirely unequivocal. It is enough that for every understanding of a sentence there is a unique understanding* of a corresponding state of affairs.) By quantifying over understandings, we might coordinate further specificity for the sentence “Sid grunts” with further specificity concerning what it is to grunt, along these lines: s, on the understanding U, is true iff: on the understanding U, p. Let “s” be replaced by a name for “Sid grunts”, and “p” by that sentence itself. Truth is a property of a sentence-understanding pair. There is no absolute what it is to grunt, but only understandings of what it is to grunt. All truth theory need do is coordinate the understandings. Suppose on one understanding of the sentence “Sid grunts” it says that Sid makes grunting

41 noises in the course of the majority of his conversational exchanges in the year 2006, and this (or the correlated) understanding of what it is for Sid to grunt is that it is for him to make grunting noises in the course of the majority of his conversational exchanges in the year 2006. Then an instance of the schema just displayed is: “Sid grunts”, on the understanding to make grunting noises in the course of the majority of his conversational exchanges in 2006, is true iff: on the understanding to make grunting noises in the course of the majority of his conversational exchanges in 2006, Sid grunts. If this is intelligible, we seem to have steered between Travis’s dilemma. The instance speaks of grunting on an understanding, and so does state a condition under which something can be true. The generalization does not mention any understanding in particular, and so is not open to the charge that it associates a sentence with an excessively specific truth condition, one favouring just one of many specific understandings. This is all mere hocus pocus. (No doubt Travis would enthusiastically agree.) On a Davidsonian picture, an interpreter is supposed to use the truth theory to arrive at interpretations. But once understandings are quantified over, an interpreter will need to know which understandings are appropriate in order to arrive at an interpretation. This knowledge in itself, however, is the semantic knowledge the theory was supposed to represent, but instead of being represented, it is presupposed. To fill out this point, we can turn to the presumed axiomatic basis of truth theory. Truth conditional semantics are supposed to be compositional. The quantification over understandings cannot be suddenly imposed upon standard unquantified T-theorems: if these are true, the quantification is unnecessary and inappropriate, and if they are false, the theory is false. We would need to think of axioms applying to expression-understanding pairs, not in an understanding-by-understanding way (for there are too many possible understandings to itemize), but in some general way, for example: for all “grunt”-appropriate understandings U, for all x, x satisfies iff: on U, x grunts.

42 The restriction on understandings to ones appropriate to “grunts” is essential. Otherwise we will have instances like: for all x, x satisfies iff: on the understanding that it is enough to be painted green, x grunts. The anomalous understanding is one that would be needed in an account of “green”, so it will be a member of the domain of quantification over understandings. But this instance is either nonsense, or it delivers the wrong result. (For the second alternative: an understanding of what it is to be a grunter on which it is enough to be painted green would ensure that suitably painted benches are grunters.) Hence understandings must be restricted to ones appropriate to “grunts”. Using the theory now presupposes that the user knows in advance which these understandings are, for the theory does not say. But to know which understandings of “grunts” are appropriate entails knowing what “grunts” means. The knowledge the theory was supposed to state has not been stated but has been presupposed. I conclude that a Davidsonian cannot respond to the kind of global argument we have considered by quantifying over contexts or understandings.

4. Unspecific meanings How, then, should Travis’s point be met? For examples like “grunt” I think the main thing is to distinguish the common phenomenon of unspecific meaning from semantic context-dependence. I’ll illustrate with an example that will be uncontroversial for many Contextualists (though I fear may not be so for Radical Contextualists of Travis’s kind). There are many ways to run, east or west, to work or to the gym, in the morning or in the evening. An utterance merely of “John runs” does not provide any of these details, though if the utterance is true, it will be made true by an event which resolves every such issue. Context may make some more specific way of running salient, but in doing so, the semantics are not touched. The test is that one can coherently deny that John runs in a salient way, without this

43 being either a retraction or a contradiction. Hence the salient way of running is not part of the semantics. For example, it would be natural to interpret the utterance of “John runs”, as it occurs in the following context Jill walks to work. John runs. as committing the utterer to the claim that John runs to work. The question is whether this commitment (supposing it to be genuine) emerges from the semantics of “John runs”. A negative answer is suggested by the following possible variant: Jill walks to work. John runs. Indeed, he runs 20 miles a week. But never to work, on account of the traffic. The coherence, and absence of retraction, suggests that in this utterance the semantics of “John runs” does not assign it the content “John runs to work”. This suggests that the same is true of the shorter utterance, for the longer one has the shorter one as a proper part. By the time the interpreter had reached the second full stop in the longer utterance, he should presumably have reached just the state he would have reached when interpreting the shorter one, and so, on the rival view, would have believed that John had been said to run to work. Such an interpreter would have to regard the remainder of the longer utterance as either containing a contradiction or a retraction of the earlier part. Intuitively, however, that is not the case. The same point can be reached by a slightly different route. Consider Bob walks to work. Jill doesn’t run. But she runs a quarter marathon every Sunday. On a Contextualist view, it should be easy to hear this as consistent, for the second sentence will be equivalent to “Jill doesn’t run to work”. In fact it is hard to hear the whole as consistent, suggesting that the second sentence tells us that Jill doesn’t run anywhere (or in any way).

44 Unspecific meaning is the category to which the Davidsonian should assign Travis’s grunter. True, there are many ways of grunting, as there are many ways of doing anything. If an attribution of grunting is true, it is made so by some specific form of grunting. None of this entails that the semantic content of “grunts” varies from context to context (nor that the pragmatic content varies). A test is this: if we can add something equivalent to “in some way or other” without making a significant difference, the verb is semantically neutral concerning the way it is to be satisfied. The default reading of “Sid grunts” is that he grunts in some way or other: the truth conditions are unspecific relative to various modes of grunting. The default reading of “Sid doesn’t grunt” is that he doesn’t grunt in any way. The same goes for colour terms: “red” applies to the things that are red in any one of possibly indefinitely many ways (on the inside, on the outside, naturally, through being painted, etc.). If one of these ways is highly salient, we may criticize a speaker for applying “red” to something not red in the salient way; we may voice this criticism by saying that what the speaker said is not true. But we normally do not care about the distinction between semantic content and what a speaker meant, and so we would not discriminate between these different targets of our criticism. Suppose external redness is salient, and someone says, of something which is red inside but not outside, that it is red. It would be natural to respond like this: You’re wrong: it’s not red in the relevant way. It may be red inside, but it’s not red outside. The whole exchange does not require any more specific semantics for “red” than that it is satisfied by something which is red in some way or another. This permits a sensible story about “red inside”, which is hard to tell if the salience of external redness made the contained occurrence of “red” apply only to things externally red. These cases contrast sharply, I believe, with other Contextualist examples. “Jill is ready” is not equivalent to the near-trivial “Jill is ready for something or other”, and “This girder is strong enough” is not equivalent to the trivial “This girder is strong enough for something or

45 other”. We should not let the fact that there are specific ways to grunt undermine our confidence in the full correctness (barring considerations related to tense) of the claim that “Sid grunts” is true iff Sid grunts. By contrast, “ready” and “enough” may well demand a treatment which reveals their content as context-sensitive. We cannot offer a global response to the global argument offered by Radical Contextualists. We have to look at their examples case by case. One avenue of response is to say, as with “grunts”, that they confuse many ways in which a sentence can be made true with many distinct contents, when really we have a single rather unspecific content. This will not do for all cases. In what follows, I consider further weapons which a Davidsonian should stock in his armoury.

5. Other context-driven problems for truth theories: rain and possession There are very many locutions and kinds of locution that have been held to cause trouble for truth theoretic semantics: small flies in Davidson’s ointment (or parts of the one large fly). A thesis of this paper is that there is no general argument to show that they can all be dealt with truththeoretically. Piecemeal examination is required. In this final section, I will consider just two examples: “It is raining”, and possessives. Suppose that “It’s raining” is uttered in Austin on 07/07/07. Let’s suppose, with the contextualist, that the semantic content of the utterance is that it’s raining in Austin on 07/07/07. (This could be partially justified by the apparent inconsistency of adding “But not anywhere”.) A Davidsonian apparently cannot do justice to this as there is no lexical element in the utterance corresponding to a location. This case can be addressed by exploiting conditional truth conditions. We can treat the sentence uttered as essentially unstructured (tense once more is set aside), so that it receives an axiom along the following lines:

46 if in an utterance u of “it’s raining” the speaker referred to place or range of places p, then u is true iff it’s raining somewhere there [within p].1 (“There” indicates the use of scene/content structure.) Non-semantic knowledge will be required to determine the range of places. This gives the right result even for cases in which the speaker does not refer to her current location. If my wife in Austin utters Mark’s in London. It’s raining she would normally be rightly understood as having said that it’s raining where Mark is (in London). It also gives the right result for more recherché cases, like the example offered by François Récanati (Récanati 2002: 317), in which many rain detectors are each monitoring one of many locations, each set to sound a bell if rain is detected. Hearing the bell, but not knowing which detector has sounded, someone exclaims “It’s raining”. He has referred to the range of places being monitored and has said that it’s raining somewhere there.2 Turning to possessives, the data are that a noun phrase formed with a possessive (e.g. “John’s car”) may invoke a wide range of distinct relations. One Davidsonian response would be the conditional truth condition approach: for all possession relations R, all objects z, all referring expressions X, all predicates Y: if in an utterance u of “X’s Y” the speaker refers to relation R, then u thereby refers to z iff z = the satisfier of Y which stands in relation R to the referent of X. 1

Davidson himself, without comment, inserts location into the truth condition: “‘Es regnet’ is true-in-German when spoken by x at time t if and only if it is raining near x at t” (Davidson 1973: 135). The example in the text above shows that this way of incorporating location will not always give the right truth condition. 2 Rain has been a surprisingly frequently discussed topic recently. For an overview of some positions, see (Récanati 2007).

47 A problem with this approach is that it may deliver something uninterpretive. For example we could use some collateral knowledge, along with the envisaged semantics, to infer that one who uttered John’s car is rusty could properly be reported as follows: Referring to the car with VIN number 1079856291 he said that it was rusty. Even if true, this is not interpretive. In contrast to reports of bare demonstratives, some conceptual material relating to cars and John should occur in a fully adequate report. An alternative approach sees this as a case of deep unspecificity. Some evidence for this might come from the fact that whereas one who utters a sentence containing “John’s car” should have some kind of idea what car is at issue, he may have little thought for the relation which holds between it and John. This suggests something like: “X’s Y” refers to the referent of “X”’s satisfier of “Y”. In short, the proposal is that this case be treated like “grunts” and “runs”: the possessive construction is simply unspecific in its meaning. One will typically hope that context will reveal greater specificity, but this is not semantically required. These suggestions indicate some of the weapons a Davidsonian will need to sustain the viability of truth conditional semantics. No doubt there will be others (a firm distinction between the pragmatic and the semantic will be one). And perhaps there are some kinds of contextual phenomena which defeat a Davidsonian approach. As far as I am aware, however, these have yet to be described.

48 References Borg E. (2004), Minimal Semantics, Clarendon Press, Oxford. Davidson D. (1967), “Truth and meaning”, reprinted in (Davidson 1984), 17-36. Davidson D. (1973), “Radical interpretation”, reprinted in (Davidson 1984), 125-39. Davidson D. (1984), Essays on Truth and Interpretation, Oxford University Press, Oxford. Higginbotham J. (1994), “Priorities in the philosophy of thought”, Supplementary Proceedings of the Aristotelian Society 68, 85-106. Récanati F. (2002), “Unarticulated constituents”, Linguistics and Philosophy 25, 299-345. Récanati F. (2007), “It is raining (somewhere)”, Linguistics and Philosophy 30, 123-46. Sainsbury R.M. (2005), Reference without Referents, Clarendon Press, Oxford. Travis C. (2006), “Insensitive semantics”, Mind and Language 21 (1), 3949.

FREGE AND DAVIDSON ON PREDICATION Eva PICARDI (University of Bologna, Italy)

1. Predication: what is the issue and what is at issue The ancient problem of predication, i.e. the most basic connection in which words can enter to form a semantic unit, may be seen as bearing on at least four clusters of issues: (i) what requirements on word meaning and grammatical form have to be satisfied in order for a collection of words to count as a sentence of a given language? (ii) what constraints on word meaning and logical form have to be satisfied for a sentence to express a proposition upon which a thinker may pass judgement, i.e. accept it as true or reject it as false? Plausibly, our acceptance of a given sentence as true and our rejection of it as false require that we understand at least the meaning of its component words and the way they are put together. This brings us to the third and fourth issue that a theory of predication has to address: (iii) how do the syntax and logical form that we ascribe to sentences relate to their truth-conditions? (iv) how does the syntactic structure that we read into (simple) sentences relate both to their inferential potential, i.e. to their occurrence as premises and conclusions of deductive arguments and to the vast variety of compound and quantified sentences that may be seen as “constructed” out of them? For, obviously, an account of syntactic structure that did not explain how sentences connect with one another would be of no help for casting light on the notions of semantic consequence and valid argument. It is essential that we bear in mind the architectonic demands that any account of predication should satisfy, in order that we may be able to evaluate its success or failure. The main claim of this paper is that, in spite of all its difficulties, including those pointed out by Davidson, Frege’s account of predication is the only candidate at our disposal that satisfies such demands. Broadly speaking, we may distinguish deflationary accounts of predication of the Quine-Davidson variety, with their attendant disquotational schema:

50 (a) “F” is true of x iff x is (an) F and inflationary accounts of predication, with their attendant “robust” nondisquotational schema: (b) “F” is true of x iff x instantiates the property of being F. (MacBride 2006) offers a less committal formulation of non deflationary accounts of predication, for he is concerned chiefly with the semantic issue of predicate reference, and not with the metaphysics of particulars, universals, properties, tropes, facts, states of affairs. Perhaps, if we can provide a convincing account of (b), not all is lost for correspondence theories of truth, heirs to the logical atomism of Russell and Wittgenstein (cf. e.g. Newman 2002). As an alternative to (b), (MacBride 2006) suggests the following schema: (b*) “F” is true of x iff x instantiates the referent of F. According to MacBride, the issue of predicate reference is only loosely connected to the ontological commitments undertaken by the bound variables of quantification, and pertains chiefly to questions of identity and identification concerning the referents of predicates (if any). Quine held such questions intractable, and this was one of the reasons why he rejected quantification over attributes, relations and propositions, and second-order logic in general. MacBride is inclined to think that quantification does not commit us to the entities our bound variable range over unless such a commitment had already been undertaken in language or thought. However, the lively contemporary debates in meta-ontology on the one hand, and the renewed interest in second-order logic and abstraction principles, on the other, show that there is no consensus as to how ontological issues should be framed, let alone settled. Quine’s criterion is still, in my opinion, a valuable one. I have mentioned (b*) mainly because it is against this way of construing Frege’s account of the role of concept-words in predication that Davidson’s criticisms in his Lectures on Predication are directed.

51 However, I am inclined to think that (b*) does not capture the functional model relevant to Frege’s account of predication. Besides, talk of “instantiation” is reminiscent of traditional theories of universals, and is wedded to the subject-predicate model, which Frege considered inadequate to explain both the logical role played by concept-words in deductive argument and the way concepts are given to us in thought and language. Frege held that in order to account for predication in general, and for singular predication in particular, we ought to refrain from assimilating the logical role of predicates to the role played by proper names. It is a wellknown (and questionable) feature of Quine’s (and Davidson’s) account of predication that the referential role is discharged not by singular terms but by individual variables occurring in a position accessible to first-order quantifiers. Schema (b*), on the other hand, may be relevant to Frege’s epistemology of language, and can be seen as offering an answer to his characterization of what it is for an object to fall under a concept, and derivatively, what it is for an object to belong to the extension (Umfang) associated to the concept in question. As is well known, after Russell’s discovery of the contradiction, Frege no longer availed himself of the notion of extension conceived as a logical object associated to a given concept, while holding fast to his view that concepts are a special kind of function. These two issues (the semantic issue of predication and the epistemological issue of how concepts are given to us in thought and language), though closely linked, should not be conflated. One may subscribe to the functional model,1 without endorsing the basic tenets of Frege’s epistemology of language.

2. Compositionality and Truth-Theories Davidson seminal paper of 1967, Truth and Meaning, opens with the question: How do the meanings of sentences depend upon the meanings of 1

By “functional model” I mean the model to be found in Frege’s mature writings, e.g. in (Frege 1891) and (1893), as distinct from the programmatic distinction between function and argument that he put forward in his Begriffsschrift of 1879.

52 individual words? This is possibly THE fundamental question of semantics. If wonder is what sets philosophy in motion, this is the appropriate place for it to set in. Frege’s Third Logical Investigation opens with the observation: It is astonishing what language can do. With a few syllables it can express a incalculable multitude of thoughts, so that even if a thought has been grasped by an inhabitant of the Earth for the very first time, a form of words can be found in which it will be understood by someone else to whom it is entirely new. This would be impossible, if we could not distinguish parts in the thought corresponding to the parts of a sentence, so that the structure (Aufbau) of the sentence can serve as a picture of the structure of the thought. To be sure, we talk figuratively when we transfer the relation of whole and part to thoughts (Frege 1984: 390). A key-word in the above quoted sentence is structure. For, as Frege goes on to remark, “the question now arises how a thought comes to be constructed, and how the parts are so combined together that the whole amounts to more than the parts taken separately”. This is a version of what (Russell 1903) called the problem of “the unity of the proposition”, which is at the centre of Davidson’s inquiry in Truth and Predication. Frege’s answer to the question of compositionality rests on his distinction between the different semantic role played by saturated vs. unsaturated parts of a thought, where saturation and combination are not meant as temporal processes.2 It is well to bear in mind that the part-whole model is not 2

The metaphor of “saturation” made its appearance very early in Frege’s writings, earlier, that is, than his mature view that a concept can be conceived as a special kind of function. As early as 1882, in his letter to Anton Marty, Frege writes: “A concept is unsaturated, in that it requires something that falls under it; hence it cannot subsist by itself. That an individual falls under it is a judgeable content, within which the concept appears as the predicate, and is always predicative. In this case, in which the concept is an individual, the relation of subject to object, is not a third added to the two of them, but belongs to the content of the predicate, which is what makes it unsaturated” (Frege 1980: 101, my emphasis). Empty concepts, unlike vague concepts, are admissible in logic, for they too are capable of an ascription of number.

53 applicable at the level of Bedeutung or semantic value: it does not make sense to say that we discern parts within a truth-value (a logical object). The functional model is not a part-whole model, as Frege himself was at pains to point out, after having helped himself to such an interpretation.3 This is by no means a minor quibble, for Davidson’s interpretation of Frege’s functional character of concepts rests on the assumption that a function (the Bedeutung of a predicate) is a separate ingredient of reality, alongside its arguments and values. To be sure, concepts and functions must be in the domain of second-order quantifiers, but it is not through this feature that they discharge their role in predication. At any rate, functions and concepts are better conceived as an objective pattern that we discern in the world (or in a structure), not as a separate ingredient of it. Frege thought that to us human beings most concepts are accessible only through the medium of language. The ability to correctly apply a concept-word to an object can be seen as the ability to fit things into a pattern. Concepts, as much as objects, must be given to us under a mode of presentation. He held, moreover, that only concept-words with sharp boundaries are suited to pick out concepts. The semantic role played by a concept-word in predication could be conceived as an objective method for fitting arbitrary objects into a pattern: we ourselves may lack the means for finding out whether an arbitrary object fits into the pattern, hence such method falls short of being a constructive one. Frege’s (or Cantor’s, for that matter) conception of functions takes it for granted that there is a determinate value that a well-defined function yields for any arbitrary object as argument.4 Davidson’s picture, on the other hand, fits very well Russell’s account of the unity of the proposition to be found in his masterpiece of 1903, The Principles of Mathematics (Russell 1903). I will return to this in section 4. Some philosophers (e.g. Geach 1975) have found the metaphor unintelligible that a thought should have parts, and suggested that we apply the functional model also to the thought. This idea should be resisted, for it misconstrues the relation between sense and reference and takes us back to a conception of content as a complex of representations (a 3 4

Cf. in his essay “Über Sinn und Bedeutung”. For a discussion of this issue cf. (Dummett 1991).

54 “Vorstellungsverbindung”) which Frege (Frege 1879: §2) had inherited from the Kantian tradition and was deeply dissatisfied with. In fact, he discarded it as soon as he had the functional model fully worked out, i.e. after 1890. As Dummett has pointed out: The condition for the truth of the thought has to be viewed as the satisfaction by a particular object, given in a certain way, of a certain condition on it. […] In more Fregean language, the sense of the proper names determines an object as its referent; and the sense of a predicate determines a mapping from objects to truth values, that is to say, a concept. The sentence is true or false according as the object does or does not fall under the concept, that is according as it is mapped by it on the value true or on the value false. The mapping of objects on to truth-values is not the sense of the predicate, but its referent: the sense is, rather, some particular way, which we can grasp, of determining such a mapping. But the sense of the predicate is to be thought of, not as being given directly in terms of a mapping from the sense of proper names on to anything, but rather in terms of mapping of objects onto truth-value (Dummett 1981: 252-3). I have mentioned the Geach/Dummett controversy over the part/whole model because when reading Davidson I sometimes have the impression that he too, like Geach, construes Frege’s principles of compositionality and functionality as if we could determine what thought is expressed by a sentence in advance of determining how its constituent parts—i.e. proper names, quantifiers, identity and predicate expressions—contribute to its truth-conditions. Moreover, Geach’s model suggests that a predicate contributes via its sense to the truth-conditions of the sentences in which it occurs, contrary to Frege explicit endorsement of the extensional point of view against the “intensional” logicians.5 Frege’s statement that the thought expressed by a sentence could be rendered as the thought that its truth-conditions are satisfied (Grundgesetze der Arithmetik, vol. I, §34) could not be more explicit, and has an obvious bearing on the present discussion. This strategy of explanation is not available to those 5

I have dealt with such matters elsewhere, e.g. (Picardi 1983) and (1987).

55 philosophers (such as e.g. (Künne 2005) and (Horwich 1998)) who maintain that the concept of truth and the concept of a proposition can, and hence must, be explained independently of each other. Neither Frege, nor Davidson, nor Dummett, for that matter, are among them. An additional reason for being dissatisfied with Frege’s doctrines is presented in the second part of Truth and Predication, and concerns the problem of predication. Predication had already been mentioned in Truth and Meaning, for there Davidson wrote: One proposal is to begin by assigning some entity as meaning to each word (or significant syntactic part) of the sentence; thus we might assign Theaetetus to “Theaetetus”, and the property of flying to “flies” in the sentence “Theaetetus flies”. The problem then arises how do the meaning of a sentence is generated from these meanings. Viewing concatenation as a significant piece of syntax, we may assign to it the relation of participating in or instantiating: however it is obvious that we have here the start of an infinite regress. Frege sought to avoid the regress but his doctrine seems to label a difficulty rather than solve it (Davidson 2001: 17, my emphasis). In the course of that paper the building-block conception of meaning— which he attributes to Frege, among others—is criticised and discarded: to be told that the meaning of a sentence depends on the meanings of its component words and the way they are put together does not carry us anywhere, for it appeals to an unspecified notion of meaning. Frege’s metaphor of saturation vs. insaturation merely labels the difficulty, but does not solve it.6 Admittedly, the model of saturation applies also to non sentential units, such as functional expressions whose value is not a truthvalue. Hence the distinction between saturated and unsaturated parts of a 6

And yet, as (Sainsbury 1996) points out in a paper devoted to Russell’s multiple relation theory of judgement, and sympathetic ante litteram to Davidson’s approach to the problem of predication, many philosophers nowadays have no qualms in helping themselves to the ingredients of Russellian singular propositions. They hardly pause to reflect on what turns such items into the particular proposition which happens to be the object of belief or judgment for a thinker.

56 compound, without an appeal to truth and truth-conditions, does not, as it stands, “solve” the problem of predication. However, it does a good deal more than merely “labelling” the problem. For, on Frege’s own grounds, we cannot account for the predicative role of a concept-word without putting forward a picture of the way in which it contributes systematically to the wealth of thoughts expressible in a language. Since Frege’s account of the contribution of the sense of a predicate expression to a judgeable content is inseparable from his conception of how it contributes, in virtue of its semantic value (Bedeutung), to the truth-conditions of the sentence in which it occurs, the link is forged between predication and truth, which Davidson himself considers the key to the problem.

3. “True of” Davidson admits that Frege had made enormous progress over his predecessors: So far, Tarski’s method has not been distinguished from Frege’s except that it associates no entities which express generality with predicates or any entities at all with sentences. The focus is on the role of variables or the spaces they occupy is analogous to Frege’s, and was inspired by him. Tarski’s essential innovation it to make ingenious use of the idea that predicates are true of the entities which are named by the constants that occupy their spaces or are quantified over by the variables which appear in the same spaces and are bound by the quantifier (Davidson 2005: 159). Davidson then goes on to mention infinite sequences: it is such sequences that satisfy sentences, whether open or closed. Sequences are abstract objects, just as Frege’s mappings. However, the satisfaction relation belongs to the meta-language, and the ontological commitments undertaken in the meta-language should not be mixed up with the ontological commitments undertaken in the object-language. Quite so, but, as we shall see, Frege’s notion of Bedeutung too, as applied to predicates

57 of a formalized language, can bear a quasi meta-linguistic construal. Davidson does not deny that predicates have “extensions”, for it would be very strange indeed that within an extensional framework co-extensional predicates should not be intersubstitutable salva veritate. Co-extensional predicates are true of the same objects. Frege, for one, denied that unsaturated expressions could flank the identity sign; functions can be said to be equivalent, but not identical. Davidson, too, assumes that predicates have extensions, but he does not tell us what sort of abstract objects extensions are; only he insists that we should not think of extensions as sets.7 To appreciate the semantic role of predicates is to appreciate how they contribute to the truth-conditions of the sentences in which they occur. Tarski’s account provides such a general method for specifying the role of each predicate in the language, plus “a non recursive axiom which says under what conditions it is true of any number of entities taken in the order in which its blanks occur” (Davidson 2005: 161). The only way of avoiding the regress involved in inflationary explanations of predication, while, at the same time, offering an account of the semantic role of predicates, is to employ the “true of” idiom. Since a language couched in first-order logic does not need set-theory to model its semantics, at no stage in the object language a commitment to an ontology of sets is undertaken. Doubtlessly, however, Davidson was well aware that there is much in natural language that goes beyond the expressive bounds of firstorder logic.8 It may not be amiss to recall the reasons why the Urheber of the “true of” coinage thought that we should excise all reference to attributes in dealing with predication. In Methods of Logic, section 39 Quine writes:

7

Could they be (as large as) classes, as Parsons surmises on p. 146. footnote 5 of (Davidson 2005). 8

(Lepore and Ludwig 2007) seem to have waved aside Davidson’s preference for first-order logic. Moreover, Davidson follows Quine in considering substitutional quantification unintelligible. Whether he follows Quine also in taking second-order quantifiers as first order-quantifiers over sets is unclear to me, owing to the lack of an explanation of what the extension of a predicate is supposed to be.

58 Despite this correspondence the singular abstract term differs from the general concrete term in an important way: it purports to name one and only one object, abstract object though it be, while the general term does not purport to name at all. The general term may indeed “be true of” each and many things, viz. each red thing, or each man, but this kind of reference is not called naming; “naming” at least as I shall use the word, is limited to the case where the named object purports to be unique. This is one of the early occurrences of the expression “true of” in Quine’s writings addressed to a general public; the expression had already occurred in his early logical writings. Charles Parsons justly complains that Davidson gives too much philosophical credit to Tarski and too little credit to Quine and to himself (Davidson 2005: 152, footnote 14). It is to Quine that we owe the appreciation of the philosophical significance of Tarski’s technical notion of satisfaction. At any rate, most of us were instructed by Quine as regards the advantages of this way of formulating matters over the traditional logic of terms. Peter Geach’s work, Reference and Generality, has contributed like few others to expose the pitfalls into which the doctrine of supposition can lead us. Quine closes section 39 with an observation characteristic of his whole approach to issues of semantics and ontology: The distinction between general terms and abstract singular terms is a remnant of medieval logic which some modern logicians do not share my concern to preserve. Actually the significance of the distinction is clearer since the rise of quantification theory than it had traditionally been; singular terms are accessible to positions appropriate to quantifiable variables, while general terms are not. In the foregoing paragraphs it has accordingly been urged that general terms have the virtue, as against abstract singular terms, of letting us avoid or at least postpone the recognition of abstract objects as values of our variables of quantification. Some logicians, however, attach little value to such avoidance or postponement. This attitude may be explained in some cases by a Platonic predilection for abstract objects; not so in other cases, however, notably Carnap’s. His attitude is rather that

59 quantification over abstract objects is a linguistic convention devoid of ontological commitment; see his “Empiricism, semantics, and ontology” (Quine 19723: 220-221). Those who want to dissociate quantification from ontological commitment, and, so to speak, “travel light” on variable-binding operators, should side with Carnap. Davidson is on Quine’s side as regards the philosophical advantages of the “true of” idiom. However, his emphasis differs from Quine’s, because, unlike the latter, he is interested in the construction of theories of meaning and he is weary of the slogan that truth is disquotation that some have read into (Quine 1970). Also his attitude concerning ontological and metaphysical issues is different from Quine’s (and Carnap’s). “There is—he says—the metaphysical question how particulars are related to properties, and the semantic question how subjects and predicates are related” (Davidson 2005: 83). The metaphysical question, Davidson surmises, may be relatively easy to answer, whereas the semantic question is the difficult one. Hard-boiled students of Quine and Carnap may very well wonder whether, once the semantic question has been settled, there is any (easy or difficult) metaphysical question left to be answered. And, above all, who is supposed to answer that question, and by what means? Davidson’s main thesis, however, is not just that the metaphysical question may be easier to answer than the semantic one, but that whatever answer we favour—no matter whether we are nominalists, platonists, trope theorists, arch-realists, scientific materialists, and what not—it has no bearing on the issue of predication. But are the metaphysical and epistemological issues which Davidson does not address really irrelevant to the problem of predication? Plainly, those who expect from a theory of predication a clue as to the difference between essential predication and contingent predication will be dissatisfied with Davidson’s proposal. Their dissatisfaction, however, does not concern me. Equally dissatisfied will be those who expect from a theory of predication a clue concerning how to handle vague predicates, predication in fictional and metaphorical contexts, predicates with a tendency to generate paradoxes, etc. My concern is with the epistemology of the satisfaction relation.

60 Davidson’s deflationary account of predication is counterbalanced by an “inflationary” account of truth. It is via the theory of radical interpretation that the connection is forged between truth and belief. But can a theory of belief afford to be silent about the epistemology of concepts? For, after all, in predication we display our mastery of conceptual connections, possibly (though not for Davidson) of conceptual schemes.9 Is it really credible that questions concerning concept formation and concept identification should be irrelevant to the semantic issue of predication? What is the epistemology of the relation of satisfaction? How does it connect with belief and acceptance? Do all predicates conform to the same pattern when it comes to satisfaction? Toward the close of his Lectures on predication Davidson remarks: It will be noted that I have explained satisfaction in terms of truth. If I were out to define truth, this would be circular. But defining truth in not my aim, for it cannot be done. I was not defining but using the concept of truth, which, however beset by paradox, is the clearest and most basic concept we have. What my strategy amounts to, then, is to show how a grasp of the concept of truth can explain predication (Davidson 2005: 160-1). But, as Davidson himself had pointed out (Davidson 2005: 75), the concept of truth must connect essentially with belief, and belief and acceptance are epistemic notions par excellence. It is this connection that is left unexplained by Tarski’s account. The whole construction would be, as it were, hanging in the air unless we were able to show that the concept of truth is really involved in the practice of speaking a language. This is the task of the theory of radical interpretation, whose goal is to give empirical substance to the truth-theory for a language, by connecting semantic notions with speakers’ attitudes of holding true directed at sentences. A 9

For a theory of concept possession that aims to capture certain features of Frege’s semantic (and metaphysical) realism cf. (Peacocke 1992). Under the heading of “concept formation” fall issues relating e.g. to conceptual and objectual abstraction, and to the various kinds of definition by means of which new concepts are added to a given language or theory.

61 semantic theory which purports to give an account of what it takes to understand the meaning of the most basic form of linguistic unit cannot afford to be silent on questions concerning concept formation and concept identification. For, clearly, Davidson sees more in Tarski’s notion of satisfaction than a mere technical device, and this is why more should be said in support of the claim that it is through this relation that language hooks upon the world. If all we can say about the satisfaction relation is exhausted by a statement of the satisfaction clauses, the promise that the theory of radical interpretation will cast light on the connection between truth and belief (and hence meaning and use) is bound to remain unfulfilled.

4. Regress arguments In Truth and Predication Davidson touches on Plato’s truly striking (and soon forgotten) mention of singular predication in the last sections of the Sophist, rehearses the probable and improbable deeds of Theaetetus, and goes through the regress argument implicit in considering instantiation and participation as the semantic counterparts of concatenation. The regress argument is the master-argument that Davidson employs to undermine a number of attempts made in the history of philosophy to “solve” the problem of predication by resorting to an assignment of a referent to predicate expressions. The problem of the unity of the proposition was especially pressing for the founding fathers of modern logic, and for the author of Principles of Mathematics in particular. Indeed, Davidson’s regress argument seems tailored on Russell’s own account of it, as put forward in Chapter 4 of (Russell 1903), and it is well worth reviewing in broad outline. In § 54 of (Russell 1903) Russell remarks: The verb, when used as a verb, embodies the unity of the proposition, and is thus distinguishable from the verb considered as a term, though I do not know how to give a clear account of the precise nature of the distinction” (Russell 1903: § 54, 50).

62 In Russell’s opinion the difference between a concept occurring as a predicate in the proposition “Socrates is human” and as a term in the proposition “Humanity belongs to Socrates” has nothing to do with “the intrinsic nature of the terms” (Russell 1903: § 49, 46). The concept is the same in both instances: the difference is merely grammatical, and has to do with “external relations”. And yet, this conception of “humanity” and “human” as indicating terms of equal standing causes troubles for the unity of the proposition. We have just reviewed Quine’s reasons for not making this move. For, given Socrates and Humanity, one may be tempted, in order to restore the unity of the proposition, to invoke the copula of instantiation, or an abstract relation of participation. Davidson takes it for granted that all the linguistic “connectors” or copulae that one may be tempted to invoke for connecting terms, introduce new components into the sentence, which call for new constituents in the realm of reference, thus starting off a search for new copulae or higher order connectors. An alternative open to Russell would have been to enter the difference between object and concept among the “indefinables of logic”, as Frege had suggested.10 According to Frege (as read by Russell in 1903) concepts, owing to their predicative nature, can only appear as relating relations (or as Frege, and later (Russell 1924), will put it, always with an indication of their argument-places and range of significance). However, far from maintaining that concepts themselves cannot figure as subjects of predication, Frege insisted that the relation in which a concept of first level stands to a concept of second level into which it falls is analogous to the relation in which an object stands to a first-level concept under which it falls. Frege’s hierarchy of functions, much as Russell’s later theory of types, prevents the occurrence of sentences such as “( ) is mortal is mortal”. In Appendix A of The Principles of Mathematics, devoted to an examination of Frege’s logical doctrines, Russell comes to the conclusion that “on the whole, the doctrine of concepts which cannot be made subjects seems untenable” (Russell 1903: § 483, 510). Probably Russell’s failure at 10

For a discussion of this and related issues cf. (Picardi 2003).

63 the time to appreciate the beauty of Frege’s theory of first and second level concepts was due to Frege’s own bogus “paradox” of the concept horse. As a matter of fact, however, already in a letter to Russell, dated 19 June 1902, Frege had pointed out in passing that the concept of a function is a second level concept, whereas in spoken language it often masquerades as a first level concept. Similar consideration might be brought to bear on the so-called paradox of “the concept horse is not a concept”. Frege tried various ways out of his self-inflicted puzzle. For instance we may try the rendering: “That which ‘is a horse” refers to is a concept”11; its second-level formulation as “(∃F) (x) (F(x) iff Horse(x))”, sounds tame enough. In the first of his three essays on the Foundations of Geometry, published in 1903, Frege comments in a footnote on Russell’s way of using the word “term” in The Principles of Mathematics. He there writes: Russell does not want to concede that a concept is essentially different from an object; concepts too are always supposed to be terms. I have treated of this difficulty in my essay Über Begriff und Gegenstand. It is clear that we cannot present a concept as independent, like an object; rather it can occur only in connection. One may say that it can be distinguished within, but it cannot be separated from the context in which it occurs. All apparent contradictions that one may encounter here derive from the fact that we are tempted to treat a concept like an object, contrary to its unsaturated nature. This is sometimes forced on us by the nature of our language. But it is merely linguistic (Frege 1903: 282). Those who have taken the so-called paradox of the concept horse as a reductio ad absurdum of Frege’s conception of the predicative character of concepts are, I think, badly off the mark.12 11

For a discussion of Frege’s way out the bogus paradox of The Concept Horse cf. (Dummett 1973) and (1981), (Picardi 1994), (Mendelsohn 2005). (MacBride 2006) offers a useful discussion of the relevant literature. 12 During the last twenty years or so Jaakko Hintikka, Thomas Ricketts, Joan Weiner, Warren Goldfarb, Cora Diamond and others, have claimed that those who praise Frege

64 Russell, accordingly, discards Frege’s solution, but the dual nature of the verb poses problems to his own account of the unity of the proposition: By transforming the verb, as it occurs in a proposition, into a verbal noun, the whole proposition can be turned into a single logical subject, no longer asserted and no longer containing in itself truth or falsehood. […] “Caesar died” and “The death of Caesar” will illustrate this point. If we ask: What is asserted in the proposition “Caesar died”? the answer must be “the death of Caesar is asserted”. In that case, it would seem, it is the death of Caesar which is true or false; and yet neither truth nor falsity belongs to a mere logical subject. The answer here seems to be that the death of Caesar has an external relation to truth or falsehood (as the case may be), whereas “Caesar died” in some way or others contains its own truth or falsehood as an element (Russell 1903: § 52, 48). Transforming the “relating” verb into a verbal noun is one of the two ways in which the unity of the proposition is lost. For the function of the verb is not just that of “relating” the terms but also that of conveying an assertion. It appears to Russell that it is the act of assertion (in the logical sense) that brings it about that that a proposition is internally related to truth or falsity. “The death of Caesar” and “Caesar died” seem to denote the same complex, and yet only the latter has an internal relation to truth and falsehood:

for his insights on the semantics of natural languages are mistaken. Allegedly, the absence of a distinction between object-language and meta-language together with the lack of the conception of variable domains render semantics de facto ineffable for Frege, and de iure ineffable for early Wittgenstein. I, for one, fail to see why Frege’s contention that only informal elucidations of the “indefinables” of logic are available, should be incompatible with the construction of a formalized language endowed with a semantics. For a different, and more balanced, appraisal of Frege’s contribution to the philosophy of language cf. (Heck and May 2006). Linsky’s valuable paper of 1992 on the unity of the proposition in Frege and Russell takes the paradox of the Concept Horse to be fatal for Frege’s conception.

65 There appears to be an ultimate notion of assertion, given by the verb, which is lost as soon as we substitute a verbal noun and is lost when the proposition in question is made the subject of some other proposition […] Thus the contradiction which was to have been avoided, of an entity which cannot be made a logical subject, appears to have become inevitable. This difficulty, which seems to be inherent in the very nature of truth and falsehood, is one which I do not know how to deal satisfactorily. The most obvious course would be to say that the difference between an asserted and an unasserted proposition is not logical, but psychological (Russell 1903: § 52, 49). This is not the regress on which Davidson focuses in Truth and Predication, but is possibly more instructive as a way of showing that by turning a verb into a verbal noun we sever the connection with the concept of truth, which was later to become also for Russell a defining feature of a proposition (an “internal” relation, as he and Wittgenstein put it). Russell (tentatively) suggests that what must be added to the complex is not a copula in its role of a connecting glue, but the copula as a bearer of assertoric force. Frege, on the other hand, had split the two roles that traditional logic assigned to the copula thus: in its “connecting” role “is” forms part of the predicate, whereas in its role of force indicator is represented by the vertical stroke of the assertion sign, whose scope is the whole sentence. Hence predication and assertion part company. The “is” of identity and of concept-subordination are not “copulae”, but expressions indicating the identity function and the conditional connective, which in Frege’s formalized language is a function of first level of two arguments. The unity of the proposition is lost also by treating the relation as a term, as metaphysical pluralism requires: Consider […] the proposition “A differs from B.” The constituents of this proposition, if we analyze it, appear to be only A, difference, B. Yet these constituents, thus placed side by side, do not reconstitute the proposition. The difference which occurs in the proposition actually relates A and B, whereas the difference after analysis is a notion which has no connection with A and B. […] A proposition, in fact, is

66 essentially a unity, and when analysis has destroyed the unity, no enumeration of constituents will restore the proposition (Russell 1903: § 54, 49-50). As soon as we analyse the proposition into its constituents, the relating relation, expressed by the verb and responsible for binding the constituents into a unit, turns out to be a term on a par with the remaining constituents. In 1903 Russell, who was well conversant with Bradley’s regress arguments, failed to see a satisfactory way out of the problem. We cannot discuss all the variants of Russell’s accounts of the unity of the proposition during the years 1910-1918. We may surmise that Wittgenstein’s conception of the relation between propositions and states of affairs, as put forward in the Tractatus logico-philosophicus did, to a considerable extent, satisfy Russell. His former pupil, however, did not approve of the theory of types—Russell’s full account of predication—but we cannot here go into such matters.13 However a hint to the theory of types may not be amiss, for, as I said, this is Russell’s full account of predication, an account, that is, that satisfies the architectonic demands I mentioned in Section 1 of my paper. In the essay on Logical Atomism of 1924 Russell admits that in 1903 he had badly underestimated the importance of language in general, and had overlooked the many traps that grammar has in store for philosophers and logicians. He there writes: A very unfortunate effect of the peculiarities of language is in connection with adjectives and relations. All words are of the same logical type; a word is a class of series, of noises or shapes according as it is read or heard. But the meanings of words are of various different types, an attribute (expressed by an adjective) is of a different type from the objects to which it can be (truly or falsely) attributed […], a relation is of a different logical type from the terms […] There is not one relation of meaning between words and what they stand for, but as many relations of meaning, each of a different logical type, as

13

See e.g. (Griffin 1985), (Candlish 1996), (Sainsbury 1996).

67 there are logical types among the object for which there are words. […] “Brutus killed Caesar” is significant, but “killed killed Brutus” is nonsense, so that we cannot replace “Brutus” by “killed”, although both words have meaning. This is plain common sense, but unfortunately almost all philosophy consists in an attempt to forget it. The following words for example by their very nature sin against it: attribute, relation, complex, fact, truth, falsehood, not, liar, omniscience. To give a meaning to these words we have to make a detour by way of words or symbols and the different ways in which they mean […] Attribute-words and relation-words are of the same logical type, therefore we can say significantly “attribute-words and relation-words have different uses”. But we cannot say significantly “attributes are not relations”. By our definition of type, since relations are relations the form of words “attributes are relations” must not be false but meaningless, and the form of words “attributes are not relations” similarly must be not true, but meaningless. Nevertheless the statement “Attribute-words are not relation-words” is significant and true (Russell 1972: 153). Russell’s defence, in 1924, of the theory of types, and his new way of addressing Bradley’s regress arguments, bear a striking resemblance to Frege’s complaints about the logical “imperfections” of ordinary language.14 Russell’s essay foreshadows Carnap’s distinction, as put forward in the Logical Syntax of Language, between formal and material modes of expression. Carnap’s doctrine of “Allwörter” echoes, in turn, Wittgenstein’s distinction between formal concepts and genuine concept, as put forward in the Tractatus logico-philosophicus, while advancing a criticism of Wittgenstein’s distinction between saying and showing. 14

“But with regards to unities the question is more difficult. The topic is one with which language, by its very nature, is peculiarly unfitted to deal. I must beg the reader, therefore, to be indulgent if what I say is not exactly what I mean, and try to see what I mean in spite of the unavoidable linguistic obstacles to clear expression” (Russell 1972: 157). The Concept Horse strikes back, or?

68 In the Tractatus logico-philosophicus the saying/showing distinction together with the distinction between formal and genuine concepts enabled Wittgenstein to place the function/argument model in a new light. It is an Urbild, a sort of logical archetype, displayed in all propositional functions, not a constituent of any of them. A sentence does not contain among its constituent the logical form it displays: and this is why, among other things, there can be no representative in the proposition of the logic of facts. The logical constants do not stand for anything, and certainly do not stand for functions, as Frege had maintained. This is a criticism of Frege different from Davidson’s, which cannot be discussed in the present paper. Yet, Wittgenstein’s idea that a sentence displays its logical form is in harmony with Frege’s conception of functional application as the logical Urform of simple sentences.

5. Frege’s epistemology of language What enabled Frege to get round a problem that was so vexing for young Russell? The answer is suggested by Frege himself: Unlike the author of The Principles of Mathematics, he did not begin with the constituents (terms) given in advance, but extracted them out of complete sentences by means of decomposition and analysis. Frege always took pride in this achievement, indeed, he thought that this was a (the?) peculiarity of his approach to logic: What is distinctive about my conception of logic is that it begins by giving pride of place to the content of word “true” and then I immediately go on to introduce a thought as that to which the question “is it true?” is in principle applicable. So I do not begin with concepts and put them together to form a thought or judgement. I come to the parts of a thought by analysing the thought. This marks off my concept-script from a similar invention of Leibniz and his successors, despite what the name suggests; perhaps it was not a happy choice on my part (Frege 1919: 253).

69 This is no means a “rational reconstruction” ex post of the spirit that had informed his logical work. In an unpublished essay, written around 1882, known as the “Anti-Boole”, young Frege, after giving an example of what he means by the several “decompositions” to which the same equation, e.g. 24 = 16, is amenable, writes: Instead of putting a judgement together out of an individual as a subject and an already previously formed concept as predicate, we do the opposite and arrive at a concept by splitting up the content of a possible judgement. Of course if the expression of the content of a possible judgement is to be analysable in this way, it must already be itself articulated. We may infer from this that at least the ideas of these properties must have their own simple designations. But it does not follow that these properties and relations occur apart from objects: on the contrary they arise simultaneously with the first judgement in which they are ascribed to things (Frege 1980: 17). The contrast between young Frege and young Russell could not be sharper. Frege’s logic is from the very beginning conceived as a symbolic language suited to represent the structure of sentences which express judgeable contents. It carries with it from the very beginning an essential appeal to the notion of truth which should have received Davidson’s enthusiastic approval. In section §9 of Begriffsschrift Frege’s goal is to show that the argument/function distinction is to be preferred to the traditional analysis in terms of subject and predicate. In dropping one or more occurrences of the same proper names from a sentence (such as “Cato killed Cato”), we highlight the argument-places of a function where substitutions can be effected. In so doing we extract different concepts from the original sentence (e.g. “c killed x”, “x killed c”, “x killed x”). To the isolation of the characteristic pattern discernible in an array of sentences Frege attributed the utmost importance. The priority of the sentence over its constituents and components plays a crucial role also in Frege’s epistemology of language. Once Frege had fully worked out the programmatic distinction between function and argument outlined in the Begriffsschrift, he was able to give a full account of the semantic contribution of concept-words to the truth-conditions of sentence in which they occur in a predicative role.

70 He held that concepts are given to beings with our physical and mental constitution only via the medium of language. To be sure, they are not created by language, but without a symbolic medium (without signs) we would not have access to them. Young Frege started with the Kantian distinction between Anschauungen and Begriffe, only to find it unsatisfactory on several grounds. Aristotle, as much as Boole, operated with a conception of concepts as given in advance of the sentences by means of which we formulate our judgement, or in restricting artificially the range of such sentences to those exemplifying the subject-predicate model. While we may possibly have access to (spatio-temporal) objects (the “unbestimmtes Etwas” or the “determinable X”, or the “something = x” invoked as a substratum in traditional accounts of predication) independently of singular terms, our access to concepts, functions, and relations is made possible only through the employment of a symbolic medium. Only language users can form sentences endowed with a grammatical and logical structure that enables them to express a judgeable content, i.e. a content capable of being accepted as true or rejected as false. This claim was perhaps at the time a striking and controversial one. Nowadays, after the so-called “cognitive turn”, is almost universally rejected, except by “linguistic” philosophers, among which the writer of this paper counts herself. Frege found it hard to see how one can do mathematics without a symbolic medium. However, he did not confine his claim to the province of mathematics. There is no finite (or infinite) stock of concepts given in advance of our engaging in the practice of forming a judgement and putting it into words in some language. Frege replaced the Kantian question: How concepts combine with intuitions in the synthesis of judgement? With the question: What is the role played by concept-words in the context of a declarative sentence capable of being accepted as true or rejected as false? The answer young Frege suggested to this question in a short essay, On the Scientific justification of a Conceptual Notation, published in 1882, is this: Without symbols (Zeichen) we could hardly attain conceptual thinking. Thus in applying the same symbol to different but similar things we actually no longer designate the individual thing, but rather what they have in common: the concept. And we attain the concept only in the

71 process of designating it; for the concept itself is not something that can be apprehended in intuition (unanschaulich), but it stands in need of a sensible representative (er dedarf eines anschaulichen Vertreters) in order that it may appear to us. Thus a sensible element (das Sinnliche) discloses to us the world of what is not sensible (Frege 1972: 84, translation modified and completed).15 In his First Logical Investigation, Der Gedanke, of 1918 Frege returns to this cluster of problems, and fits them into his new conception of the third realm. Young Frege, however, was no such arch-realist, though he plainly was a conceptualist of sorts. In section 26 of Grundlagen der Arithmetik he characterises “objectivity” as independence from sensation, intuition and ideas, but not as independence from reason. The objectivity of arithmetic rests on the objectivity of the concepts involved in statements of number, and concepts fall within the province of logic. Logic does not traffic in “pure forms”: it is from the very beginning contentual logic, i.e. a logic apt to express the judgeable contents conveyed by our sentences, irrespectively of their subject matter. In this respect, Frege’s conception of logic is very different from Kant’s. At the heart of Frege’s logicism there is the conviction that concepts, functions and relations (of first and second level) have the epistemological priority over classes, be they conceived as sets, aggregates or collections. Concepts are apprehended by us (are “given to reason”) in the process of issuing a judgment on the content expressed by a sentence that satisfies certain syntactic and semantic requirements. A class, as Frege never tires to emphasize, has its Bestand in the concept whose extension it is: the latter should not be conceived as the collection of objects that fall under it concept in question, but as an individual (logical) object. However, with the help of concepts alone, we are unable to justify the claim that numbers are individual logical objects—a claim that Frege regarded as crucial for his foundational programme. Frege’s logicist epistemology mandates an objectivist conception of concepts, functions and relations. His epistemology requires that concepts 15

In the Bartlett/Bynum’s translation the last sentence in the quoted passage is omitted, and, as a result, the reader cannot appreciate the connection between (Frege 1882) and (1918). Also the allusion to the Kantian Anschauung is lost.

72 be construed as that which concept-words refer to in the context of a sentence. The notion of reference as applied to predicates is therefore forced on us. However, predication requires that the semantic role of predicates should not be construed according to name-bearer prototype. Perhaps one could try to develop an objectivist conception of functions and relations, short of a full-fledged Platonist construal of them. My suggestion that the semantic role of predicates be seen as an objective method for fitting objects into a pattern goes in that direction.

6. The Bedeutung of predicates in formalized languages In Truth and Predication Davidson writes: any attempt to give full explications of the semantics of predicates by associating them with objects of any kind is doomed. […] Nor will it help to distinguish, as Frege did, between objects, which is what singular terms refer to, and concepts, which is what one-place predicates refer to. To say that predicates are functional expressions, and that therefore are incomplete or unsaturated, and that what they refer to is similarly full of holes or spaces waiting to be filled in, does not help: entities are entities, whatever we call them (Davidson 2005: 136). Now, if all that Frege had to offer were a bunch of metaphors, few would disagree with Davidson. But before we dismiss Frege’s insight into the functional character of concepts, we should try to understand the rationale behind the metaphor by recalling the peculiar features of Frege’s formalized language, as put forward in the first volume of Grundgesetze der Arithmetik. Dummett’s emendation in the book of 1981, The Interpretation of Frege’s Philosophy, of his earlier (Dummett 1973) criticism of Frege’s functional characterization of concepts, is now seen by Davidson as a move backwards. One of Dummett’s reasons for being ill at ease with Frege’s account of concepts as functions is Frege’s assimilation of sentences to complex terms, standing for one of the two logical objects, the True and the False. But both the idea that True and False are logical

73 objects, and the conception of sentences are complex names of such objects can be given up or reinterpreted, without abandoning Frege’s important insight into the semantic role of predicates. The counterintuitive construal of sentences in Frege’s formalized language could be (and has been) remedied by appeal to categorial grammar or type-theoretic approaches, which make finer distinction within the category of terms. Besides, Frege himself emphatically denied that in asserting a sentence are we naming anything: this is what the assertion sign is there for. It is a fact, however, that Frege’s syntax is hospitable to many semantic interpretations—including combinatorial, type-theoretic, constructivist ones. Under all semantic interpretations we need to know what the semantic value of a predicate is supposed to be. Frege took the semantic value of a concept-word (an expression with one or more argument places) in predicative position as consisting in a mapping from objects to truthvalues, whereas Tarski, like most mathematicians, took the semantic value of the predicates in the object-language to be a set. Davidson, too, maintains that predicates have an extension. His main claim is that no matter how such “extensions” are construed, they play no role in explaining predication. There is a further charge that Davidson levels against Frege, i.e. the duplication of semantic entities corresponding to predicates. For he writes: If we take predicates as referring to entities we introduce a shadowy level of explanatory machinery between the expressions and the work they do. […] But if predicates have a referent, this is in addition to their sense and extension. This is the wheel that becomes redundant: to describe the semantic value of a predicate is not to introduce another level of explanation (Davidson 2005: 139). Here Davidson mentions the notion of extension, but “extensions” were conceived by Frege as the referents of names of value ranges, and they never played a role in his account of what it is for an object to fall under a concept, or what it is for a first level concept to fall within a second level concept. Because of extensionality, all concepts-words that take the same value for the same arguments count as standing to each other in a relation analogous to that of identity.

74 In his book on the Interpretation of Frege’s Philosophy Dummett argues that the correct construal of the notion of Bedeutung is that of semantic value—a programmatic notion, that can be variously interpreted, depending on the theory of meaning one subscribes to. I would go even further. Frege’s notion of Bedeutung has a dual role. It functions as theoretical concept required for laying down the semantic contribution that different types of expressions make to the truth-conditions of sentences belonging to a formalized language. In Grundgesetze der Arithmetik Frege employed the notion of Bedeutung for spelling out, within the object language, as it were, an equivalent of the recursive clauses in the definition of satisfaction that Tarski formulated in the meta-language. Frege’s specification is viciously circular, but this is a different matter.16 Thus in stating in the formal mode, relative to a given formalized language, the truth-conditions of “Theaetetus flies” we can say that the sentence in question is true iff the function designated by “x flies” maps Theaetetus on to the True, and is false otherwise. Let us consider more closely the notion of satisfaction. We may say that a predicate is true of an object or a sequence of objects, but also that there is a function that assigns objects to variables as satisfying a predicate; it satisfies it when the result of interpreting the free variables in it as referring to the object assigned to them by the function is true. This formulation in terms of functions is the one actually used by Tarski in his essay of 1936. Thus, to put it a bit impressionistically, we may say that Frege’s mistake, according to Davidson’s reconstruction, is due to his ignorance of the distinction between meta-language and object-language. It is in the meta-language that we need to appeal to the existence of a function of the sort just described: no function needs to be mentioned in stating, within the object-language, the truth-conditions of simple and compound sentences. Once we appreciate that the theoretical role of the notion of Bedeutung, as applied to a functional expression within a formalized language, does not consist in assigning a referent to predicate, but rather in spelling out, in the object language how e.g. the predicate “sits” contributes to truth-conditions of sentences which can be constructed in the formalized language, the contrast between Frege and Tarski appears 16

Cf. (Kutschera 1989: ch. 7) and (Dummett 1991).

75 less stark. The matter is, I think, really entangled, for though it is true that the distinction between object-language and meta-language appears only very late in Frege’s writings, he made an attempt to state the semantics of his formalized language within the very same language. It is by no means clear that all such attempts are doomed to failure, and that Frege’s strategy is hopeless. However, if we see matters under this angle, would Frege’s appeal to functions as the semantic value of concept-words still be unwarranted also from Davidson’s own perspective? Frege’s notion of Bedeutung, however, does not play just the role of a theoretical tool, useful for laying down a systematic assignment of truth conditions for the sentences of a formalized language. The notion of Bedeutung is meant to account for what it is to understand a sentence belonging to a natural language, built out of words whose sense and mode of composition is familiar to the speakers of the language. The understanding of the thought expressed by a sentence of a given language requires that we have a conception of what it is for a given name to designate an object and of what it is for a certain predicate to single out a method for fitting objects in a pattern of discerned similarities. The pattern is not of our own invention: it is the joint work of mind and world. Davidson’s criticisms notwithstanding, Frege’s account of predication still strikes me as the best one at our disposal. One may wonder whether the metaphor of saturation is a felicitous one for elucidating predication. Here opinions may differ. Nonetheless, it may not be amiss in this context to recall Frege’s conception of the constraints by which the originators of new metaphors should abide: The purpose of elucidations is a pragmatic one; and once it is achieved, we must be satisfied with them. And here we must be able to count on a little goodwill and cooperative understanding, even guessing; for frequently we cannot do without a figurative mode of expression. But for all that, we can demand from the originator of an elucidation that he himself know for certain what he means; that he remain in agreement with himself; and that he be ready to complete and emend his elucidation whenever, given even the best of intentions,

76 the possibility of misunderstanding arises (Frege 1906: 301, translation slightly modified).17

References Beaney M. (1997), The Frege Reader, Blackwell, Oxford. Candlish S. (1996), “Russell’s theories of judgement”, in: R. Monk and D. Palmer (eds.), Russell and the Origins of Analytic Philosophy, Thoemmes Press, Bristol, 1996, vol., 103-135. Carnap R. (1937), The Logical Syntax of Language, Routledge and Kegan Paul, London. Davidson D. (1974, 20012), Inquiries into Truth and Interpretation, Clarendon Press, Oxford. Davidson D. (2005), Truth and Predication, Harvard University Press, Cambridge Mass. Dummett M. (1973, 19812), Frege: Philosophy of Language, Duckworth, London. Dummett M. (1981), The Interpretation of Frege’s Philosophy, Duckworth, London. Dummett M. (1991), Frege’s Philosophy of Mathematics, Duckworth, London. Frege G. (1879), Begriffsschrift, eine des reinen Denkens abgeleitete Formelsprache, Engl. transl. in: T.W. Bynum (ed.), Gottlob Frege, Conceptual Notation and Related Articles, Clarendon Press, Oxford 1972. Frege G. (1882), “Über die wissenschaftliche Berechtigung einer Begriffsschrift”, Engl. transl. in: T.W. Bynum (ed.) Gottlob Frege, Conceptual Notation and Related Articles, Clarendon Press, Oxford 1972. Frege G. (1884), Die Grundlagen der Arithmetik, Engl. transl. by J.L. Austin, The Foundations of Arithmetic, Blackwell, Oxford 1950. Frege G. (1891), Function und Begriff, reprinted in (Frege 1990). 17

For criticisms and suggestions I wish to thank Fabrizio Cariani, Antonio Ferro, Daniele Mezzadri, Carlo Penco and Elisabetta Sacchi.

77 Frege G. (1892), “Über Sinn und Bedeutung”, Zeitschrift für Philosophie und philosophische Kritik 100, 25-50, reprinted in (Frege 1990), Engl. transl. in (Beaney 1997). Frege G. (1893), Grundgesetze der Arithmetik, vol. 1, Pohle, Jena. Frege G. (1903), “Über die Grundlagen der Geometrie”, I, Jahresbericht der Deutschen Mathematiker-Vereinigung 12, 368-75, reprinted in (Frege 1967), Engl. transl. in (Frege 1984). Frege G. (1906), “Über die Grundlagen der Geometrie”, II, Jahresbericht der Deutschen Mathematiker-Vereinigung 15, 293-309, reprinted in (Frege 1967), Engl. transl. in (Frege 1984). Frege G. (1918), “Der Gedanke. Eine logische Untersuchung”, in: Beiträge zur Philosophie des Deutschen Idealismus I, 58-77, Engl. Transl. in (Beaney 1997). Frege G. (1923-26), “Logische Untersuchungen. Dritter Teil. Gedankengefüge”, in: Beiträge zur Philosophie des Deutschen Idealismus III, 36-51. Engl. transl. in (Beaney 1997). Frege G. (1967, 19902), Kleine Schriften, ed. by I. Angelelli, Olms, Hildesheim, Engl. ed. by B. McGuinness, Collected Papers on Mathematics, Logic, and Philosophy, Blackwell, Oxford 1984. Frege G. (1969), Nachgelassene Schriften, ed. by H. Hermes (et alii), Meiner, Hamburg, Engl. transl. by P. Long and R. White, Posthumous Writings, Blackwell, Oxford 1979. Frege G. (1976), Wissenschaftlicher Briefwechsel, ed. by G. Gabriel, H. Hermes, F. Kambartel, C. Thiel, A. Veraart, Meiner, Hamburg, Engl. abridged by B. McGuinness and transl. by H. Kaal, Philosophical and Mathematical Correspondence, Blackwell, Oxford 1980. Geach P.T. (1962), Reference and Generality, Cornell University Press, Ithaca. Geach P.T. (1975), “Review of M. Dummett, Frege: Philosophy of Language”, Mind 85, 287-302. Griffin N. (1985), “Russell’s multiple relation theory of judgement”, Philosophical Studies 47, 213-47. Heck R., R. May (2006), “Frege’s Contribution to the Philosophy of Language”, in E. Lepore and B. Smith (eds.), The Oxford Handbook of Philosophy of Language, Oxford University Press, Oxford, 3-39. Horwich P. (1998, 19902), Truth, Oxford University Press, Oxford.

78 Künne W. (2005), Conceptions of Truth, Oxford University Press, Oxford. Kutschera, F.v. (1989), Gottlob Frege, Walter de Gruyter, Berlin-New York. Lepore E., K. Ludwig (2005), Donald Davidson. Meaning, Truth, Language, and Reality, Oxford University Press, Oxford. Lepore E., K. Ludwig (2007), Davidson’s Truth-Theoretical Semantics, Oxford University Press, Oxford. Linsky L. (1992), “The unity of the proposition”, The Journal of the History of Philosophy 30, 243-273. MacBride F. (2006), “Predicate Reference”, in E. Lepore and B. Smith (eds.), The Oxford Handbook of Philosophy of Language, Oxford University Press, Oxford, 422-475. Mendelsohn R. (2005), The Philosophy of Gottlob Frege, Cambridge University Press, Cambridge. Newman A. (2002), The Correspondence Theory of Truth. An Essay in the Metaphysics of Predication, Cambridge University Press, Cambridge. Parsons C. (1983), Mathematics in Philosophy, Cornell University Press, Cornell. Peacocke C. (1992), A Study of Concepts, MIT Press, Cambridge Mass. Picardi E. (1983), “On Frege’s Notion of Inhalt”, in Atti del Convegno internazionale di Storia della logica, San Gimignano, 4-8 dicembre 1982, CLUEB, Bologna, 307-312. Picardi E. (1987), “Premessa”, in: G. Frege, Scritti postumi, Napoli, Bibliopolis, 9-66. Picardi E. (1994), “Kerry und Frege über Begriff und Gegenstand”, History and Philosophy of Logic XV, 9-32. Picardi E. (2003), “Frege, Peano e Russell sulle idee primitive della logica”, in: N. Vassallo (ed.), La filosofia di Gottlob Frege, FrancoAngeli, Milano, 181-210. Quine W.v.O. (19723, 1953), Methods of Logic, Routledge and Kegan Paul, London. Quine W.v.O. (1970), Philosophy of Logic, Prentice-Hall, Englewood Cliffs, N.J. Russell B. (1903), The Principles of Mathematics, Cambridge University Press, Cambridge.

79 Russell B. (1924), “Logical Atomism”, reprinted in D. Pears (ed.), Russell’s Logical Atomism, Fontana/Collins, London, 143-165. Sainsbury M. (1996), “How can some thing say something”, in: R. Monk and D. Palmer (eds.), Russell and the Origins of Analytic Philosophy Thoemmes Press, Bristol 1996, 137-153. Tarski A. (1936, 1956), “The Concept of Truth in Formalized Languages”, in: J. H. Woodger (ed.), Logic, Semantics, Metamathematics, Oxford University Press, Oxford, 152-278. Wittgenstein L. (1921), Tractatus Logico-Philosophicus, Routledge and Kegan Paul, London.

EVENTS AND CONSERVATIVITY: CLUES TOWARDS LANGUAGE EVOLUTION* Massimo PIATTELLI PALMARINI (University of Arizona, USA)

1. Introduction: The relevant (for us) components of the neo-Davidsonian program Davidson taught us that words have basic “application” conditions—e.g., the French word ‘neige’ applies to snow; this allows the Tarskian project to be extended to phrasal constituents, a most welcome outcome. These application conditions determine, in accordance with a formal, Tarskian, procedure, the “correctness” conditions of sentences—e.g., the sentence ‘La neige est blanche’ is correct in French iff snow is white; this applies across languages and a speaker understands a language, in the sense of being able to discern all the semantic properties of expressions in the language, if the speaker knows the basic application conditions for the relevant words and knows how to determine the correctness conditions of sentences in the language (compositionality). To this we, of course, add Davidson’s momentous introduction of a predicate-place for events (originally (Davidson 1967)), i.e. his classical data and arguments to the effect that words and expressions of natural languages typically refer to, and quantify over, events. His most famous example is: (1) John buttered the toast slowly, deliberately, in the bathroom, with a knife, at midnight. There is no elegant way, by means of the classical, pre-Davidsonian, logical apparatus, to see the truth conditions and all the implications of (1) come out right. If one is limited to quantifying over objects only, and does not want to change the adicity of predicates for each new predicate that is introduced, then five pieces of toast easily pop up, contrary to our *

Work in progress with Juan Uriagereka and Paul Pietroski—University of Maryland.

82 straightforward understanding of (1). Bringing them back to one piece only, as we should, introduces inelegant objectual identifications. The Davidsonian apparatus, on the contrary, straightens things out very nicely. Introducing a predicate place for events, and making, thus, the transitive predicate butter a di-transitive, the Logical Form (hereinafter LF) of this sentence (omitting some inessential details), is: (1a) ∃e: Butter (John, toast, e) ∧ [Agent (John, e) ∧ Theme (toast, e)] ∧ in the kitchen(e) ∧ slowly(e) ∧ deliberately(e) ∧ with a knife(e) ∧ at midnight(e) There is an event, it is an event of buttering, John is the agent, and the toast is the theme, and that event took place in the kitchen1, and it took place slowly, and so on. The adicity of butter is fixed once and for all, and now all the obvious inferences come out right (as being obvious) by conjunction-reduction. This is exactly what one wants to see happening. Such elegant Davidsonian strategy has changed for good the way of doing the semantics of natural language. We owe it to James Higginbotham (Higginbotham 1985) to have crucially enriched this picture by pointing out that it is the syntactic operator Tense (T) that binds Davidsonian events. Interestingly, there are hard cases, when the introduction of the argument-place for events makes the derivation of the LF particularly elegant, straightforward and perspicuous. These cases have been tackled successfully by Barry Schein (Schein 1986, 1993) and Paul Pietroski (Pietroski, 2003) (2005 and in print), among others. These are paradigmatic examples of the Schein-Pietroski sentences (I omit some details for the sake of brevity): (2) Five professors wrote six papers in March (quickly, under pressure, and inelegantly).

1

In the contemporary semantics of natural language, it’s useful to write something like in(m,e), at(n, e), and so on. I will gloss over these details.

83 LF: ∃e ∃X ∃Y [five professors(X) ∧ six papers(Y) ∧ writing(X,Y,e)] ∧ quickly (e) ∧ under pressure (e) ∧ inelegantly(e). In Pietroski’s more detailed analysis, we have:

My own favourite example is: (3) The buildings are darker and darker as you drive North Obviously, no individual building becomes darker, and our driving North is not the cause of the buildings becoming darker. What (3) says is that the distinct events of buildings being darker than other buildings, and of buildings being positioned further North of other buildings, and of one’s driving to the North are strictly related. A full analysis would require a better characterization of plurals, as we will see below. It’s evident, we think, that the LF of these kinds of sentences could not be derived without events. Events are not just a “luxury”, an elegant formal device, but a sine qua non. In this general Davidsonian frame, we wish now to proceed to the crucial puzzle of the conservativity of natural language determiners (quantifiers being a special, particularly interesting, sub-case).

84 2. The puzzle of conservativity 2.1. Brief history The relevant literature is rich and has a long and distinguished pedigree. Ever since the medieval times William of Sherwood and others noted the equivalence of All men are mortal to All men are mortal men (see (Higginbotham 1993) (for a simple general introduction, see (Larson 1995) and (Larson and Segal 1995)). Higginbotham and May noticed this universal constraint on determiners in 1978, calling it “intersectivity.”2 About the same time Barwise and Cooper called it “conservativity”, and published it in Linguistics and Philosophy (Barwise and Cooper 1981). Dag Westerstahl improved on things, with a general theorem, in the Handbook of Philosophical Logic, in 1989 (Westerstahl 1989). Higginbotham wrote up, independently, a version stressing the role of not only conservativity but a more powerful property in giving characterizations of truth, in “Grammatical Form and Logical Form”, where it appeared as an appendix (Higginbotham 1993). In 1986, an influential paper by Keenan and Stavi (Keenan and Stavi 1986) had enlarged the horizon to all determiners in a great variety of languages (very likely, as far as we can tell today, all natural languages). Further fine interweaving with syntactic theory (in a Minimalist Framework) was proposed by Norbert Hornstein and Juan Uriagereka (see their contribution in (Epstein and Hornstein 1999)). As we will see at the end of this paper, Elena Herburger importantly showed how the apparently deviant case of “only” is, after all, not deviant. Some of the evolutionary consequences of conservativity have been sketched already by Juan Uriagereka and myself in 2005 (PiattelliPalmarini and Uriagereka 2005). So much for a brief history of the subject, now to the facts.

2

The paper was later published in volume 1, number 1, of The Linguistic Review (Higginbotham and May 1981).

85 2.2. The facts The arguments of natural language determiners have to be ordered. Determiners relate sets in a specified order. All Tuscans are Italians (TRUE) ≠ All Italians are Tuscans (FALSE) All (Tusc, Ital) ≠ All (Ital, Tusc) Determiners are two-place predicates, in an interesting parallelism with transitive verbs, as shown by Richard Larson (Larson and Segal 1995, Larson 2005 and in print) (for a recent overview, see (Larson 2004), for a nuanced discussion, see (Ludlow 1995)). One place is the “restriction” (Tuscans) (syntactically, the internal argument). The other is the “scope” (Italians) (syntactically, the external argument). Conservativity is a formal property of relations among sets. A twoplace relation R is said to be conservative iff R (A,B) ⇔ R (A, A∩B) If R is true of the sets A and B (in this order), then it is also true of the sets A and the intersection of A and B (again in this order). Some, but not all, relations between functions (from individuals to truth values) are conservative. Therefore, it is very interesting that all natural language determiners are conservative. As William of Sherwood had noted, All Tuscans are Italians = All Tuscans are Tuscan Italians.

2.3. Consequences It is not a truth of reason that it be so, nor is this universal property of languages something one has any hope of explaining by means of general properties of efficiency of communication, memory optimization or any other generic (let’s insist, generic) property of the human mind. Pace the standard (alleged) evolutionary reconstructions of the origins of language by the neo-Darwinians, something else is needed to account for the

86 conservativity of determiners. This is the central consideration of the present paper. It is important also to stress an obvious fact: Children the world over do not “learn” that the determiners of their native language are conservative. No one has to “teach” them such basic fact, nor do adults even realize that this property is inherent in their language. Speakers tacitly know this as they know many other subtle properties of syntax, and the interface between syntax and semantics. To put it very bluntly and somewhat inaccurately: the conservativity of determiners is innate. In a more guarded characterization, we will say that the syntax and the semantics of natural language determiners is un-learned, because it’s part of the speaker’s tacit knowledge of language.

3. Simple examples 3.1. Standard conservative determiners They are best presented, we think, by means of an obvious semantic equivalence: (4) All children love ice cream iff All children are ice-cream-lovers that are children. (5) Most Basques are Spaniards iff Most Basques are Basque Spaniards. (6) Most engineering buildings were designed by Frank Gehry iff Most engineering buildings are Frank-Gehry-designed engineering buildings. In Barwise and Cooper’s terminology (loc. cit.), determiners “live on” their internal arguments. All “lives on “children”, most “lives on” “Basques”, and so on.

87 3.2. Impossible non-conservative Dets3 What would a non-conservative determiner look like? It takes some effort of imagination to invent examples, but here we go: “The” versus impossible *gre As every native speaker of English well knows (or even a non-native one, like myself), the sentence The bottle fell is true iff (7) a bottle, and there was only one, is a bottle that fell = The bottle is the bottle that fell. Therefore, we have this mandatory equivalence: The bottle fell = The bottle is a bottle that fell. Now, let’s introduce a strange determiner, marked with an asterisk, because no natural language could harbour it: *gre Let’s introduce it by means of semantic equivalence, get an intuition of its (impossible) meaning, then treat it as one treats ordinary determiners and see where this leads us: (8a) *Gre bottle fell (imagine it being) TRUE iff (8b) A bottle is the only thing that fell Therefore: (8c) *Gre bottle is a bottle that fell (should be = to) A bottle is the only bottle that fell. 3

Adapted and simplified from Paul Pietroski’s lectures and from examples used by Andrew Nevins (Harvard, Department of Linguistics) in his lectures.

88 But let’s consider the following, quite trivial, state of affairs: bottle-1 FELL

bottle-2 DID NOT FALL

cup-1 FELL

It’s perfectly obvious that such state of affairs makes (8c) true, but (8b) false. We are, thus, verifying that *gre is not conservative. Another impossible determiner (Nevins’s shrewd invention): *galoochy Let’s define it via the following equivalence: (9) *Galoochy engineering buildings were designed by Frank Gehry iff Galoochy Frank Gehry-designed buildings were engineering buildings. Notice that, unlike a real determiner (unlike, say, most) it is not the case that the following equivalence holds: (9a) Galoochy engineering buildings were designed by Frank Gehry iff Galoochy engineering buildings were Frank-Gehry-designed engineering buildings. If it were so, then galoochy (no asterisk now, please notice) would be a possible natural language determiner. It just sounds strange for an English determiner, but its syntax and semantics would be OK. On the contrary, our infamous *galoochy (with an asterisk) is another non-conservative (and therefore impossible) determiner. These two examples may suffice to bring home, rather intuitively, we hope, the data on conservativity we care for. No speaker uses such determiners, no child could learn their meaning upon simple exposure to typical sentences that contain them.4

4

We are presently planning real experiments to show that it is so.

89 4. The underlying syntax Higginbotham has stressed that nodes in syntactic trees are the quintessential meaning-assigners. A node in a syntactic tree is what it is because of the structure of the tree and its position in it. The internal argument of a determiner (similarly, the internal argument of a transitive verb, as Larson has perspicuously shown) corresponds to a precise node and to a precise sub-tree configuration. The external argument corresponds to a different, but equally precise, node-in-the-tree configuration. Lexical meanings and the configuration of the nodes that dominate them together explain compositionality.

The Davidsonian event position of the thematic grid of the verb is discharged at the point where VP meets Inflection (TP). The following schematic representations (due to Paul Pietroski) can give an idea of what is involved:

90 We have a process known in syntax as “raising”. Intuitively, something in the sentence must receive a syntactic (and therefore also semantic) value, and this value-assignment can only be done by something else in the structure, something that is “above” it hierarchically. Raising is what accomplishes this mandatory valuation. Lexical requirements that, for syntactic reasons, cannot be satisfied locally, must be satisfied by means of raising. Arguments for events raise to receive their value from the verb inflection, more specifically from the syntactic operator Tense (present, past-tense, future etc.). The functional constituent TP (Tense Phrase) is what is needed to satisfy the lexical requirement. There is, however, a further refinement, one introduced by the late George Boolos and adopted and enriched by Paul Pietroski. We will see it quite succinctly.

5. Plurals (not sets) (Boolos, 1984, 1998a, 1998b) The typical Boolos sentence is: (10) The rocks rained down on the mountain huts. (10) cannot be true of any rock in particular, because no single rock can “rain down”. But it is not true, either, of any set or collection of rocks, because no set or collection of rocks as such can “rain down”. Natural language predicates such as “rain down” are not satisfied by one thing, nor by any one set or collection of things. Rain down can only be true of some succession of falls by rocks, one after the other, not of a set of rocks all falling together as a set, or collection. Boolos suggested an alternative construal of second-order quantification: distinctively plural quantification over xs, such that x “is one of” the Xs (not that x∈X). Some thingsx are such that each thingx is one of themX iff itx is not an element of itselfx. In essence Boolos’s proposal is that we should invoke distinctively plural quantification over singular entities, not singular quantification over distinctively plural entities. In his own words, distinctively plural quantifiers are so characterized:

91 [N]either the use of plurals nor the employment of second-order logic commits us to the existence of extra items beyond those to which we already committed [...] We need not construe second-order quantifiers as ranging over anything other than the objects over which our firstorder quantifiers range [...] a second-order quantifier needn’t be taken to be a kind of first-order quantifier in disguise, having items of a special kind, collections, in its range. It is not as though there were two sorts of things in the world, individuals and collections of them, which our first- and second-order variables, respectively, denote. There are, rather, two (at least) different ways of referring to the same things, among which there may well be many, many collections (Boolos 1998: 72). Taking the syntax seriously is not optional, we have to do it. And, once we allow that a variable can have MANY values relative to a SINGLE assignment it’s not at all extravagant that PLURAL variables are variables of this sort, and that plural variables can be arguments of PLURAL PREDICATES like ‘rained down’. And it’s not at all extravagant that quantification is deeply related to plurality. So we can at least ask if determiners are plural predicates that combine with expressions that can have many values relative to a single assignment of values to variables. And it turns out that this is perfectly coherent: the values of determiners are pairs, just as suggested by the syntax. Determiners are satisfied plurally by such things; the internal (nominal) argument of a determiner imposes a condition on the relevant individuals; the external (sentential) argument of a determiner imposes a condition on the relevant truth values. The remaining task is just to devise a formal way of spelling out this extension of Tarski, via Boolos, to the structures generated if quantifiers raise. We can now return to our determiners and to their conservativity.

92 6. Determiners and their arguments 6.1. Arguments and their values As we have seen, a determiner takes as its external argument a sentential one (are Spaniards, like ice-cream, etc.), whose value is TRUE or FALSE relative to any assignment of values to variables. Proper assignments to these values, then, make the whole expression TRUE or FALSE. But all, every, most etc. indicate relations between sets. How can this be? (without cheating). How do sentential arguments and sets, and plurals, jibe well with one another? Pietroski’s reasoning goes a bit like this: It’s perfectly canonical that determiners map pairs of predicates to sentences, and that determiners are thus of the type . But that can't be the final word, even setting aside the need to explain conservativity. Given raising, the external argument of a determiner sure looks like a sentence with a variable in it. And while a relative clause also looks like a sentence with a variable in it, a relative clause cannot be the external argument of a determiner. This becomes especially evident when you don’t pronounce the relativizer, as in ‘Every man I met’. In fact, ‘Every man (who) I met’ fails to have a reading on which the sentence is true iff every man is such that I met him. If, once again, we take the syntax seriously, and say that the external argument of ‘every’ in ‘I met every man’ really is a sentence with a variable (an open-sentence), then determiners are of the type . Initially, this does look extravagant, since the value of an open-sentence is still a truth value (relative to an assignment of a value to the variable), just as the value of ‘he’ is still an individual (relative to an assignment of a value to a variable). But it’s not extravagant, after all. Expressions don’t change their semantic type just because a constituent name or indexical gets replaced with a variable. Which is why the standard view is so odd in this respect: ‘I saw you’ and ‘I saw him’ are said to be of type , but ‘I saw TRACE’5 ends up getting treated as of type 5

TRACE is, in Generative Grammar, a standard unpronounced linguistics element. What is “left behind” after something has been moved to its proper place in the

93 (predicate). This is inconsistent, unless you say (with Heim and Kratzer 1998) that the external argument of a determiner is really like a relative clause with a covert relativizer. Under this treatment, the suggestion is coherent, but very implausible empirically, since (as just noted) relative clauses cannot be external arguments of determiners. In Pietroski’s terms, a sentence with one variable is in may ways like the corresponding predicate of type . A determiner like, say, every raises to a position in which its (lexical) requirements are met. How do we express a function of type ? Pietroski’s proposal is that determiners are predicates of FregePairs. FregePairs are ordered pairs of the form , where v is a truth-value (TRUE or FALSE) and x is the internal or the external argument, i.e. one of the things over which (singular or plural) variables range. In other words, if the open sentence S appears as the external argument of an indexed determiner, then relative to any Assignment A, some FregePairs (the Fs) are semantic values of that external argument iff for eachf of themF: itsf External Participant is TRUE iff the open sentence is TRUE relative to the (minimal) variant of A that assigns itsf Internal Participant to the indexed variable.6 This may well be, admittedly, a bit too technical for the present readers. In a nutshell: Quantification over value-entity pairs via argument positions (internal/external) takes the syntactic parallel between determiners and verbs (à la Larson), and the syntax-semantics of plurals (à la Boolos), very seriously (as we think one should). In the light of what we saw a moment ago, the advantage of this is to unify the semantics of syntactic raising across the board. Most importantly, this is done under a constraint of strict minimality and strict locality. For reasons of space, I will not pursue this further here (see recent work by Paul Pietroski, Norbert Hornstein and Juan Uriagereka). sentence. A simple example would be Which book did you read? The proper analysis of this sentence, making TRACE explicit, is Which book did you read TRACE. In fact, “book” is pronounced after “which”, but it is obviously understood as the object of “read”. Its semantic value is assigned “elsewhere” than the place of pronunciation. 6 The “big” F and the “small” f indices denote “big” and “small’ events respectively. This has to do with singulars and plurals in the framework to which I have alluded above.

94 We dare insist on the fact that the constraints of strict locality and strict compositionality are absolutely central to the mental machinery that carries out linguistic derivations and interpretations, but they are not part of the generic machinery of perception, thought, memory, or action. We are dealing here with another significant instance of the specificity of the language faculty, of what traditionally is labelled as the autonomy of syntax. We can make the parallelism between verbs and determiners even more explicit: The (syntactically) familiar Theta-roles in verbs are: An internal argument (the theme) and an external argument (the agent). For Determiners we have an internal argument (the restrictor, a Noun) and the external argument (the scope, or the predicational s). This parallel leads us to an immediate and interesting connection:

6.2. The UTAH hypothesis Mark Baker proposed several years ago (Baker 1997) a very interesting general hypothesis called UTAH (where UTAH stands for Uniformity of Theta Assignment Hypothesis). What it states is that, in every language, the same syntactic configurations always assign the same Theta-roles (thematic roles). The internal argument is the theme (in particular for canonical transitive verbs, the “object”), while the external argument is the “agent” (syntactically, the subject). This is something that every speaker tacitly knows and something that no child ever has to “learn”. Rather, the innate availability of such principle is what allows children to acquire the local lexicon and fix the local syntactic parameters.7 The crucial fact is that (invented) verbs that would violate UTAH would be un-learnable by the

7

For recent neat evidence of the crucial role that syntactic clues and the UTAH principle play in the case of the child’s acquisition of the meaning of verbs like “believe” “think” and “know” (when there is absolutely nothing one can “show” perceptually to the child), see (Papafragou, Cassidy and Gleitman 2007) and (Gleitman, Cassidy, Nappa, Papafragou and Trueswell 2005)

95 child. What they would mean if they could exist is easy to state by means of circumlocutions, but no single verb-meaning can be like that. A simple example: (11) My sister *ploves me her children *plove(x,y,e) = an event of y causing x to love y This sentence, therefore, if such verb could exist in real languages, would mean that my sister’s children are such that they cause themselves to love me. The thematic roles and the syntactic nodes would be switched, but that is impossible. No child could learn such verbs.8 The thematic hierarchy cannot be reversed. What UTAH adds to this is that the syntax and the semantics map to one another uniformly for all verbs in a language. Across languages, this uniformity always holds, though, at least in principle, the mapping could go the other way, but still uniformly. In (11) the syntax of English tells us that my sister is the subject (the external argument, the agent), that me is the benefactive and her children the theme. But the invented semantics of *plove goes the other way. No uniformity in such case. In the same fashion, we have reasons to think that, for exactly the same kind of reason, no child could learn a non-conservative determiner. The imaginary cases presented earlier consist precisely in the swapping of the internal and the external arguments. This is why those determiners are “impossible”. It repays, we think, to insist that this is in no way a limitation of “thought”. The child can easily succeed in performing mental tasks that are orders of magnitude more complex than these. Nothing here is “generically” hard. It’s syntactically and semantically impossible, but the corresponding “thoughts”, when expressed via circumlocutions, are perfectly banal. Let’s see another example, also due to Paul Pietroski. 8

Try to compute the truth conditions for sentences like the following: No parent *bloves the other parent their children without it being manifest. John and Mary *blove themselves only some of each other’s children. It takes paper and pencil to do this, unlike ordinary sentences containing ordinary verbs.

96 6.3. Equinumerosity Equinumerosity is a familiar concept. “As many as”, “no more and no less than”, “exactly as many as”, and similar expressions convey equinumerosity in ordinary language, without any problem. But let’s consider the impossible determiner *equi (12) *Equi bottles fell TRUE iff the bottles samenumber the fallen If *equi were conservative, then (12a) Equi bottles fell TRUE iff Equi bottles are bottles that fell TRUE iff the bottles samenumber the bottles that fell (12b) *Equi bottles fell TRUE iff the bottles samenumber the fallen This determiner is impossible because it switches the internal and the external arguments. The star of *equi is well deserved, because this imaginary determiner “lives on” the scope (the fallen objects) and not on the restrictor (bottles). Of course, such thoughts about such states of affairs are perfectly thinkable, in the abstract, but not a possible (a natural, learnable, cognitively manageable) determiner of a natural language. The reason is syntactico-semantic, not generically psychological. In a nutshell (contrary to the good determiners we saw above) in *equi the external argument of the DP (the fallen) cannot raise to Spec of TP and then become the internal argument of the DP. This would be a syntactic monstrosity, not an aberrant “thought”. It’s, therefore, possible to invent non-conservative determiners and “think” about them as an intellectual exercise (with paper and pencil). It’s also possible to find circumlocutions that express linguistically what a nonconservative quantifier would express, if it existed. But no natural human language has any such determiners. They would be un-learnable by the child and un-manageable by ordinary speakers in everyday discourse.

97 7. Lessons from conservativity There is no explanation of any of this by means of standard predicateargument relations, elementary quantification, set-theoretic relations, or some generic “laws of thought”. The psychology of reasoning has nothing at all to tell us about “impossible” (non-conservative) determiners. Even less can we hope to explain this by means of facts and theories about pragmatic constraints, efficacy of communication, charitableness in translation, cultural conventions and the like. The explanation is exquisitely syntactico-semantic. It involves, as we have seen quite succinctly, deep similarities between verbs and determiners (à la Richard Larson), a natural generalization of Mark Baker’s UTAH, the asymmetry between Det restriction and Det scope, constituency (a DP must be a constituent), and (possibly) parametric differences between languages.9 Persuasive suggestions have been made to the effect that, in minimalist terms, it involves raising to check interpretable features and checking + deleting un-interpretable features. It also seems to involve chain-formation and re-projections (Uriagereka, in press). Always applying strict semantic compositionality. The main lesson from conservativity is, in our opinion, that a quite central universal of language (the conservativity of determiners) has a syntactic and semantico-syntactic explanation, but no generic “functional” explanation. It is alien to standard neo-Darwinian adaptationism.10 Of course, the linguistic computational capacities of our species must have been the result of some evolutionary story. Not, however the story that 9

The U of UTAH states that the mapping of thematic roles onto syntactic configurations must be uniform, within a given language, but such mapping could in principle have quantal variation across different languages, provided that uniformity is everywhere respected within each language. 10 Reconstructions of the evolution of language in the optic of neo-Darwinian adaptationism, based on claims that general functional and cultural factors have been crucial, are too numerous to be listed here. A short list of representative publications is (Arbib 2005) (Lieberman 2006) (Tomasello 2003) (Deacon 1997) (Jackendoff 2002). A recent gallery of anti-generativist pieces is to be found in the special issue of The Linguistic Review, Vol. 22 (2-3) (2007) “A review of the Role of Linguistics in Cognitive Science” (Nancy A. Ritter Ed.)

98 neo-Darwinians like to tell us (even well-informed neo-generativists neoDarwinians like Steven Pinker and Ray Jackendoff)11 (for earlier congenial data and arguments, see (Lightfoot 2000), (Uriagereka 1998, 2002; Epstein and Hornstein 1999), and (Piattelli-Palmarini and Uriagereka 2005, and in print)).

8. An interesting non-counter-example: only 8.1. The facts Conservativity grants that the truth conditions for all, most, some, many involve a check restricted to the intersective sets. It does not require inspection of any other set. The whole wide world of possible truth-makers is narrowed down to the intersection, and nothing else. Maybe, that’s why conservativity is cognitively so central (Higginbotham 2005). Prima facie, “only” appears to be a glaring exception. To check whether (13) Only sharks eat bluefish is true or false, one has to look also at other species, over and above the sharks. In fact, this much wider inspection is not just suggested, but imposed. One counter to this counter is that only is not a determiner strictu sensu12. “Only N” is not a DP. At variance with other determiners and quantifiers, only is highly moveable. In fact, we have as almost perfect synonyms: (14a) Children only like ice cream. 11

See the recent brisk exchange on language evolution between Hauser, Chomsky and Fitch on one side, and Pinker and Jackendoff on the other (initial position paper (Hauser, Chomsky and Fitch 2002), critique (Pinker and Jackendoff 2005), reply (Fitch, Hauser and Chomsky 2005) and counter-reply (Jackendoff and Pinker, 2005)). 12 Gennaro Chierchia (personal communication) is adamant in stressing this fact as the sole counter that is needed.

99 (14b) Children like only ice cream. (14c) Children like ice cream only (and nothing else). Try doing this with most, every, all, etc. and you will see that they cannot be likewise moved around. Again prima facie, “only” looks like a non-conservative “all”. It seems (adopting Barwise and Cooper’s terminology) to “live on” the scope, not the restrictor. (15a) All angels have wings. = All angels are winged angels. (15b) Only angels have wings. = All individuals that have wings are angels. (15a) involves a standard quantification, where all ‘lives on’ the restrictor (the angels). In customary ‘conservative’ fashion, the quantification is first computed over the restriction, and next the scope. In contrast, to get the proper logical form of (b), it would seem that “only” must first be computed over its scope and then its restriction. Elena Herburger (Herburger 2000), however, has insightfully noted that there is an element of focus involved. This can be made more evident by means of adjectivation. Rendering focus graphically by means of capitals, we have (16) Only YOUNG angels have wings (old ones don’t). The paraphrase is 16b, not 16a. (16a) All individuals that have wings are young angels. (16b) All individuals that are angels and have wings are young. If the ‘anti-conservative’ theory of only were right, there would be no reason why a logical form as in (16b) should emerge. Focus has to be implicated here—the issue is how. Summarizing drastically (see Herburger’s monograph for a full treatment, (Herburger 2000)) there is a proper syntactic treatment of focus. Following this through, “only” can be seen as a standard determiner.

100 8.2. The syntax of focus The key is that the restriction and scope of “only” are not base-generated. Rather, these are acquired in the course of the syntactic computation. Lexical information arranges itself so that, in the course of the derivation, “only” moves to a focus site, whose specifier hosts the displaced focal element that “only” “associates with” (overtly in some languages). The remainder undergoes what is technically called “remnant movement”. After these syntactic processes take place, the focused material is the scope of “only” and the rest of the clause ends up joining the restriction of this element. The result is a logical form akin to (16b), where the quantifier has a standard conservative shape. A comparison: (17) All sharks ate fish. Standard paraphrase: (17a) ‘all sharks are sharks that ate fish’. If all were anti-conservative, (17) should be able to mean (17b). (17b) ‘all fish are fish that sharks ate’. Patently, this is not the case. The logical form for (17), if it could be paraphrased as (17b), would be obtained by first assigning relevant thetaroles to sharks (as predators) and fish (as prey), then having the quantifier all, crucially by itself, raise to some scope-taking position; and finally the predicate fish raise to be in-construction-with the raised all. But this would be a syntactic monstrosity!13 The effects of focus, and the syntax of focus are, of course, much more general.

13

Let’s notice, once gain, the sharp difference between a syntactic monstrosity and “unthinkable” thoughts, or hard-to-think thoughts.

101 Again representing graphically focus by means of capitals, let’s have a look at (18) Only WHITE sharks ate fish. (18a) All fish eaters were white sharks. (18b) All shark fish eaters were white (sharks). Again, the correct paraphrase of (18) is (18b), not (18a). How do we get (18b)? In complex syntactic ways (by raising—but not by itself) only ‘associates’ to focused material, which then in some very non-trivial sense becomes eliminated from what is presupposed in the sentence. So, in the end, Herburger shows that (after a complex series of LF processes) even only has a conservative analysis. And we have just seen what it would have meant for it not to have been that sort of quantificational element.

Conclusion: Wider implications Let’s rewrite (16) (16) YOUNG angels have wings (old ones don’t). Herburger suggests, quite generally, and very interestingly, that the familiar Neo-Davidsonian operator introducing event quantifications in standard propositions is, contrary to customary assumptions, a binary existential quantifier. It involves both a restriction and a scope—not just a scope. The analysis just sketched for only directly extends to more general instances. Herburger’s punch-line is (in our wording) that there is no such thing as a sentence without a focus. If so, every single human judgment expressed through a proposition would have to participate in the syntactic organization that conservativity presupposes: The standard quantifier (here a binary existential event quantifier) ‘lives on’ its restriction and obtains its scope through a ‘quantifier raising’ mechanism. Our (Piattelli-Palmarini, Pietroski, Uriagereka) punch-line is, then, the following: What Herburger has added, technically, is the possibility that

102 restrictions too enter into ‘raising’ processes, in focal instances, in the same league as standard qantifier raising (QR) for scopal dependencies. Philosophically, if all propositions happen to involve the syntax of binary quantification, she has also added the need to reckon with a quite elaborate and subtle syntax every time a judgment is made. This adds further evidence to all we have seen above. Language is essentially a medium for “thought”, under severe and specific constraints (strict locality, strict cyclicity, strict compositionality, uniformity and inevitability of raising, conservativity of determiners). It is used also (of course) for communication, but it has not been shaped by it! Here, as elsewhere in language, adaptationist neo-Darwinian “fables” (see footnote 9), based on the efficacy of communication, or cultural complexity, or motor precursors, go completely awry. The story to be told is one involving very abstract, very specialized, internally highly structured syntactico-semantic structures, and computations over them that are severely constrained by strong locality, and strong compositionality. A very different evolutionary story. Wholly mechanistic, but not adaptationist.

References Arbib M.A. (2005), “From monkey-like action recognition to human language: An evolutionary framework for neurolinguistics”, Behavioral and Brain Sciences 28, 105-167. Baker M. (1997), “Thematic roles and syntactic structure”, in: L. Haegeman (ed.), Elements of Grammar, Kluwer, Dordrecht. Barwise J. and R. Cooper (1981), “Generalized quantifiers and natural language”, Linguistics and Philosophy 4, 159-219. Boolos G. (1984), “To be is to be the Value of a Variable (or the Values of some Variables)”, Journal of Philosophy 81, 430-449, reprinted in (Boolos 1998a). Boolos G. (1998a), Logic, Logic, and Logic, Harvard University Press, Cambridge, MA. Boolos G. (1998b), “On second order logic” (Chapter 3 of Logic, Logic, and Logic), in: (Boolos 1998a), 443.

103 Davidson D. (1967), “The logical form of action sentences”, in: N. Rescher (ed.), The Logic of Decision and Action, Pittsburgh University Press, Pittsburgh, PA. Deacon T.W. (1997), The Symbolic Species: The Co-evolution of Language and the Brain, W.W. Norton & Company, Inc., London. Epstein S.D. and N. Hornstein (1999) (eds.), Working minimalism, MIT Press, Cambridge, Mass. Fitch W.T., M. D. Hauser and N. Chomsky (2005), “The evolution of the language faculty: Clarifications and implications”, Cognition 97 (2), 179-210. Gleitman L., K. Cassidy, R. Nappa, A. Papafragou and J.C. Trueswell (2005), “Hard words”, Language Learning and Development 1 (1), 23-64. Hauser M.D., N. Chomsky and W.T. Fitch (2002), “The faculty of language: What it is, who has it, and how did it evolve?” Science 298 (22 November), 1569-1579. Heim I. and A. Kratzer (1998), Semantics in Generative Grammar, Blackwell, Oxford, UK. Herburger E. (2000), What Counts: Focus and Quantification, MIT Press, Cambridge, MA. Higginbotham J. (1993), “Grammatical form and logical form”, in: J. Tomberlin (ed.), Philosophical Perspectives, vol. 7: Philosophy of Language and Logic, Ridgeview Publishing Company, Altascadero, Cal., 173-196. Higginbotham J. and R. May (1981), “Questions, quantifiers and crossing”, The Linguistic Review I, 41-80. Higginbotham J. (1985), “On semantics”, Linguistic Inquiry 16 (4), 547593. Jackendoff R. (2002), Foundations of Language: Brain, Meaning, Grammar, Evolution, Oxford University Press, New York-Oxford, UK. Jackendoff R and S. Pinker (2005), “The nature of the language faculty and its implications for evolution of language (Reply to Fitch, Hauser, and Chomsky)”, Cognition 97 (2), 211-225. Keenan E.L. and J. Stavi. (1986), “A semantic characterization of natural language determiners”, Linguistics and Philosophy 9, 253-326.

104 Larson R. (1995), “Semantics”, in: D.N. Osherson, L.R. Gleitman and M. Liberman (eds.), An Invitation to Cognitive Science (2nd. ed.). Volume 1: “Language", Bradford Books/The MIT Press, Cambridge, Mass. 361-380. Larson R. (2004), “The Projection of DP”, Talk given at the Guest lecture series at the University of Stuttgart, January. Larson R. and G. Segal (1995), Knowledge of Meaning: An Introduction to Semantic Theory, The MIT Press, Cambridge, MA. Lieberman P. (2006), Toward and Evolutionary Biology of Language, Harvard University Press, Cambridge, MA. Lightfoot D. (2000), “The spandrels of the linguistic genotype”, in: C. Knight, M. Studdert-Kennedy and J.C. Hurford (eds.), The Evolutionary Emergence of Language: Social Function and the Origins of Linguistic Form, Cambridge University Press, Cambridge, U.K., 231-247. Ludlow P. (1995), “The logical form of determiners”, Journal of Philosophical Logic 24 (1), 47-69. Marantz A. (1997), “No Escape From Syntax: Don’t Try Morphological Analysis in the Privacy of Your Own Lexicon”, in: A. Dimitriadis, L. Siegel, C. Surek-Clark and A. Williams (eds.), Proceedings of the 21st Annual Penn Linguistics Colloquium, University of Pennsylvania Working Papers in Linguistics 4.2., Philadelphia, 201-225. Papafragou A., K. Cassidy and L. Gleitman (2007), “When we think about thinking: The acquisition of belief verbs”, Cognition 105 (1), 125-165. Piattelli-Palmarini M. and J. Uriagereka (2005), “The Immune Syntax: The Evolution of the Language Virus”, in: L. Jenkins (ed.), (Chapter 4 of) Variations and Universals in Biolinguistics, Mouton-De Gruyters, Amsterdam, 341-377. Pietroski P.M. (2003), “The character of natural language semantics”, Epistemology of Language, Oxford University Press, Oxford, UK. Pinker S. and R. Jackendoff (2005), “The faculty of language: what’s special about it?” Cognition 95, 201-236. Schein B. (1986), Event logic and the interpretation of plurals. Unpublished Thesis Ph.D., Massachusetts Institute of Technology, Dept. of Linguistics and Philosophy 1986. Schein B. (1993), Plurals and events, MIT Press, Cambridge, MA.

105 Tomasello M. (2003), Constructing a Language: A Usage-Based Theory of Language Acquisition, Harvard University Press, Cambridge, MA. Uriagereka J. (1998), Rhyme and Reason: An Introduction to Minimalist Syntax, The MIT Press, Cambridge, MA. Uriagereka J. (2002), Derivations, Routledge, London U.K. Westerstahl D. (1989), “Quantifiers in formal and natural languages”, Handbook of Philosophical Logic 4, 1-131.

DAVIDSON AND DUMMETT ON THE SOCIAL CHARACTER OF LANGUAGE Jennifer HORNSBY (Birkbeck, University of London, UK)

Introduction There is a feature of Donald Davidson’s overall picture of language that I have never been able to come to terms with. I believe that we don’t appreciate the social character of language unless we allow that when language is used, a shared language is used whose norms are determinative of what utterances mean. This belief of mine runs contrary to claims that Davidson makes in several papers. One of these (Davidson 1994) is addressed to Michael Dummett; and Dummett replied to it (Dummett 1994). So it might be thought that insofar as I disagree with Davidson on this matter, Dummett will have armed me with the arguments I need against Davidson. But that is not so. I think that Dummett fails to engage with Davidson’s argument. Davidson and Dummett are separated by their different answers to the question: What constitutes the essentially social character of language?1 My argument is going to be that we need materials from both Dummett and Davidson in order to answer this question rightly. On the matter of shared languages, we need to take Dummett’s side against Davidson. But we cannot dispense with those claims of Davidson which, as it seems to me, Dummett fails to take on board. When Dummett responds to Davidson, he finds an enormous amount to agree with Davidson about. But he nevertheless leaves us with the impression that some fundamental bone of contention remains. My first objective here is to unearth the bone. I want to see what leads Davidson to reject the view about shared languages which informs Dummett’s work. 1

My concern here is with Davidson’s views as they bear upon this question. Many authors have responded to the claims of Davidson that Dummett puts in contention, but without being interested specifically in what separates Davidson from Dummett. An example is (Reimer 2004) with which I’m in almost total agreement.

108 This is the task of §1. Carrying it out will put me in a position to take Dummett’s side on the matter of shared languages. But in taking Dummett’s side, I make use of an idea of a communicative intention got from Davidson.

§ 1.1. The particular debate I am concerned with consists of two papers each by Dummett and Davidson. The paper of Davidson’s that I’ve mentioned was a reply to a reply by (Dummett 1986) to an earlier paper by (Davidson 1986). Given that I think that the two philosophers are sometimes at crosspurposes, it will serve my own present purposes to cut a long story short. And I can keep the story short by focusing on just three claims in Davidson. (1) There’s no such thing as a language, not if a language is anything like what many philosophers and linguists have supposed. (2) The idiolect is conceptually primary, the public language secondary. (P) Prior knowledge of a shared language is neither necessary nor sufficient for communication. (1) is the claim that is prominent in (Davidson 1986), (2) in (Davidson 1994).2 As for (P), I think that we shall see that it is pivotal for Davidson— that (P) serves as his basis for both of (1) and (2). I shall try to put pressure on (P) in due course. What I need to do first is to explain how Davidson arrives at his claim (1). What conception of a language was Davidson concerned with when he said that there was no such thing as one? Well, the conception at issue is one according to which a language is a vehicle of communication, and meaning in a language meets three conditions: it is systematic; it is learned 2

Claims (1) and (P) are quoted from (Davidson 1986), and (Davidson 1994). So far as claim (2) is concerned: see the first sentences of Davidson’s and Dummett’s (1994).

109 and conventional; and it is shared. It is systematic inasmuch as meanings of sentences are composed from meanings of words. It is learned and conventional inasmuch as a speaker and audience on occasions of communication bring to bear a conventionally determined competence which they have gained in advance.3 It is shared inasmuch as speaker and audience bring the same competence as one another to bear on occasions— they bring to bear a shared understanding of the speaker’s words. Davidson thinks that a philosopher has no need of a conception of something to which we can attribute all three of these features. He reaches this conclusion by considering that someone may understand the meaning of a speaker’s words even though they have no advance knowledge of what those words as uttered by the speaker then will mean. We have examples of this when a familiar word is used to mean something other than it standardly means. Here are examples of two sorts. (A) Archie Baker says “Let’s have some music to break up the monogamy”. (M) George W. Bush says: “I am surprised at the amount of distrust that exists in Washington. I'm sorry it's the case. I'll work hard to try to elevate it”.4 In examples of both these sorts, a distinction can be made between the understanding of words that an audience brings to linguistic exchanges, and the understanding that attaches to words used by a speaker on the 3

There might be a question here what is involved in the competence being conventional. Suffice to say that for my purposes, it is enough that we are dealing with non-natural meaning in Grice’s sense. 4 George W. Bush, interview on U.S. National Public Radio, Jan. 29, 2007. The example is of course of a malapropism. Davidson illustrates these by way of Mrs. Malaprop: his title “A nice derangement of epitaphs” being a quotation from her. My “M” can stand for malapropism. I note that examples of these two sorts don’t exhaust those by which Davidson attempted to make his case. We have an example of a different sort from these two, for instance, when a completely new word is used. But what I say in §3.2 below is meant to cover examples of all sorts.

110 occasion of a particular exchange. According to Davidson, the fact that we have this distinction shows that we should not think of a vehicle of a communicative exchange both as shared and as learned and conventional. For suppose that we think of a language as given by reference to the understanding of words that a person brings to linguistic exchanges. Then, in these examples, although we can say that there is understanding which is learned-and-conventional, it isn’t shared. But suppose that we think of a language as given by reference to the understanding that a speaker achieves on the occasion of a particular exchange. Then, in these examples, although there is shared understanding, it isn’t learned-andconventional.5 If we are inclined to believe in shared languages, then we may react to this argument by saying that we should treat examples of the two sorts differently. Archie Baker does get his message across by relying on his audience’s prior knowledge of what each of his words literally mean, but he relies also on having them recognize that he wanted to convey something beyond what his words literally mean. George Bush, by contrast, gets his message across despite the fact that his audience cannot rely only upon his audience’s prior knowledge of what his words literally mean. If this is our reaction, then the idea of a language which is both learned-and-conventional and shared will be in the picture whichever kind of case we look at. Obviously this won’t satisfy Davidson. Its account of Bush loses sight of the distinctively communicative intention which Davidson treats as fundamental. A speaker’s distinctively communicative intention is to be taken to mean what one wants to be taken to mean. And in Davidson’s view, in order to know what Bush’s utterance meant, you have to grasp what Bush wanted to mean in using the words he did, your grasping of which ensures that his intention was fulfilled. Grasping this, you find that 5

I’ve simplified considerably here. Davidson puts all of this in terms of a discrepancy between prior theories and passing theories. The simplification seems appropriate since Davidson conceives of a theory of meaning simply as a satisfactory description of the competence of a speaker or audience. At any rate, Davidson doesn’t think that ordinary speakers know theories of meaning, and I think that talk about knowledge of theories in the present context is apt to introduce a red herring.

111 the idiolect of Bush at the time he spoke was one in which the word ‘elevate’ means ALLEVIATE.6 The debate might reach stalemate here. If we are on Dummett’s side, then we say that Bush used a shared language, and failed to say exactly what he meant. If we are on Davidson’s side, then we attribute meaning to Bush’s words in accordance with what he intended to be taken to mean at the particular time; otherwise we fail to see Bush as having the intention characteristic of a speaker. But although we may reach a sort of stalemate here, we can at least isolate what is in contention. We see that Davidson and Dummett are separated by Davidson’s belief in the fundamental place in an account of language of speakers’ communicative intentions on particular occasions. We are then in a position to begin to appreciate the exact role that (P) plays for Davidson. If you relied on prior knowledge of English alone, then you would fail to reach a complete understanding of the utterance in the case of examples like (A). For an account of what Archie Baker communicates requires more than the assumption that he spoke a certain language which he shared with his audience. So such examples show that prior knowledge of a language may be insufficient for communication. Examples like (M), of George Bush, by contrast, show, on Davidson’s accounting of them, that prior knowledge is unnecessary for communication. For by Davidson’s lights, it is only when a very particular idiolect is attributed to Bush that any communicative intention is seen to have been at work. Now since Dummett need not disagree with Davidson that attribution of a language shared by Baker and his audience falls short of accounting for what Baker gets across, the only claim that is bound to be in contention is the claim that shared knowledge of a language isn’t necessary for communication— one half of (P), as it were.

6

There may be other interpretations of Bush. But I shall assume that this is the correct one. At any rate, we need to settle on some definite understanding if we are to take Davidson’s point. For his point is that we can know what someone wanted to be taken to mean even though they use a word as meaning something other than what it means in a shared language.

112 At this point, then, I set aside the claim that shared knowledge of language is not sufficient for communication. But I shall come back to it in §3.

§ 1.2. I turn now to Dummett’s reaction to Davidson. Dummett argues against (1) in his first reply to Davidson. But in the second, he is apt to be dismissive of (1): he is doubtful whether anyone ever held the conception of a language which Davidson attributes to ‘many philosophers and linguists’. So Dummett comes to thinks that Davidson can be right about (1); but that insofar as he is right, he does not have any opponents. When it comes to (2), Dummett’s reaction is different. He speaks as if he had a real issue with Davidson. Still, he says that the question of whether (P) is true is largely independent of whether (2) is true. So, once again, so far as Dummett is concerned, (P) is neither here nor there. But Davidson surely thinks that (P) bears on both (1) and (2). We have just seen that the sorts of examples that he uses to establish (1) are examples that might be used to demonstrate (P). And we have also seen how one part of (P)—its claim about what communication requires—might belong in an argument that philosophers are not entitled to begin from their usual conception of a language. Davidson’s thought is surely something like this: Our fundamental understanding of linguistic communication cannot be built upon a conception of a prior shared language, since linguistic communication has no need of prior knowledge of wordsmeanings. Our conception of a shared language must, then, derive from a conception of something which comes first—from a conception of an idiolect spoken by someone at a time. We gain some confirmation that this is Davidson’s thinking in a passage from an earlier paper: Knowledge of the conventions of a language is [...] a practical crutch to interpretation, a crutch we cannot in practice afford to do without— but a crutch which under optimum conditions for communication, we

113 can in the end throw away, and could in theory have done without from the start (Davidson 1984: 278).7 Here Davidson tells us that linguistic communication has no need of shared linguistic knowledge. And his reason for thinking this is that, in theory, there can be linguistic communication between two people without there being any linguistic knowledge which the two people share. Well, if we are meant to suppose that two people might communicate without any reliance whatever on shared linguistic knowledge, this can seem incredible. If nothing that you already knew were relevant to the understanding of any of the noises I produced, then however much I wanted to be taken to mean something by producing a sequence of noises, the chances that you would take me to mean what I wanted to be taken to mean would be astronomically small. And one cannot intend to do what one knows there is only an astronomically small chance of one’s doing. We might speculate about whether there could be two gods who made noises at one another; each of them somehow picking up on the current disposition of the other to use noises as having a certain current meaning, even though neither of them knew anything about the other’s past uses. Perhaps such gods are possible. But it is not obvious that we would take their interactions to be examples of the kind of communication which we want to account for. The claim that an account of the communicative interactions of human beings can dispense with the idea of shared linguistic knowledge can seem outrageous. Well, what Davidson actually believes is somewhat less outrageous. At one point he places another requirement on communication. He says that a speaker must have adequate reason to believe that the hearer will succeed in interpreting him as he intends.8 This affects things considerably. In the first place, it might help to explain how the gods we imagined could 7

In (Davidson 1994: 115), he puts it like this: “There seems to me to be no reason, in theory at least, why speakers who understand each other ever need to speak, or to have spoken, as anyone else speaks, much less as each other speaks”. 8 This is from (Davidson 1984). In (Davidson 1994) Davidson talks of a speaker “reasonably believing” that he will be interpreted in a certain way, and tells us that “reasonable belief” is a very “flexible concept”.

114 communicate by making their noises. Perhaps gods can satisfy Davidson’s requirement because they are beings who know that they have adequate reason to think whatever they find themselves thinking. But secondly, and more importantly, the requirement makes all the difference when it comes to human beings’ communicating. I said that I would never intentionally produce a sequence of noises for you if I had no expectation that you would take me to mean what I should like to be taken to mean by producing them. But I definitely might produce a sequence of noises if I took myself to have a reason for thinking that you would take me a certain way when I produced them. And it is not my states of mind alone that the new requirement affects. For if I have good reason to think that you’ll understand me as I want, then there must surely be something about your dispositions to understand me that I can take for granted. Evidently I could take for granted what I need to if we shared knowledge of the significance of the noises I produce.

§ 2. I do not know how one might prove that knowledge of a language is more than the merely ‘practical crutch’ to interpretation that Davidson thinks it is. But I want to try to make it plausible that shared knowledge of meaning is necessary for linguistic communication as we know it. In cases of linguistic communication, a certain characteristic state of affairs is reached: the audience’s understanding of the speaker’s utterance on the occasion of the exchange is matched with how the speaker wants to be taken. If we can find such states of affairs outside examples of uses of language, then we can perhaps see what distinctively communicative intentions involve without for the moment entering into the controversial territory of language use. So I take a different tack now. I want to look at an example of non-linguistic communication (although I shall call the parties to the communication “speaker” and “audience”). One can come to know that someone has measles by seeing spots of a certain sort on their body. Spots of this sort mean measles. The meaning here is what Grice called natural meaning. But it can be put to use to communicative ends. Suppose that Anna wants to be taken to mean that a certain child has measles. She rolls up the child’s sleeve, and directs Carlo’s attention to the spots on the child’s arm. Carlo takes it that Anna

115 means that the child has measles. This is something that Anna has communicated to Carlo. Of course Carlo could have learnt that the child had measles without Anna’s help—he might just have observed the spots. Again Anna could cause Carlo to know that the child had measles without communicating this. She might, for instance, roll up the sleeve in such a way that the spots in fact become visible to Carlo, but without realizing that Carlo is present. But in the example as we have it, Anna rolls up the sleeve wanting to be taken to mean that the child has measles, and Carlo’s knowledge is got through a process by which Anna and Carlo arrive, as one might say, at mutual understanding: Notice how easily the mutual understanding is achieved in this case. Given that Anna and Carlo each know that spots of that sort mean measles and that the other knows this, it seems that nothing more is required for the fact that the child has measles to be communicated to Carlo than that he should realize that Anna intends to communicate it. We can vary the example now. Suppose that there is nettle rash on the child’s arm, but no measles spots there. Can Anna now communicate to Carlo that the child has measles? No. But suppose that Anna wrongly thinks that she is showing measles spots to Carlo. Can she then communicate that the child has measles? Still, No. For Anna’s thinking that she has displayed measles spots to Carlo, is obviously no help to her in getting it across to him that the child has measles unless Carlo somehow gets to think that Anna wrongly thinks this. Well, perhaps Carlo might reason as follows: “Anna is trying to get something across, but she surely isn’t trying to get it across that the child has nettle rash. Some of the other children have measles. Very likely she is confused about measles spots, and she is trying to tell me that the child has measles.” We now have Carlo taking Anna to mean what Anna wanted to be taken to mean. Have we then arrived at a case of communication? Well, we very certainly don’t have a paradigm case. In the original case, we saw how very easily mutual understanding was achieved. We have no mutual understanding of that easy sort now. This case, where Anna has things wrong and Carlo works things out, shows the point of the requirement that Davidson introduces—that a speaker must have adequate reason to believe that the audience will succeed in being taken as intended. Anna lacks adequate reason in this

116 case, since it is only because she wrongly thinks that Carlo will see measles spots that she supposes that she’ll be taken to mean that the child has measles. So Davidson’s requirement explains our judgment here. Davidson’s requirement shows also why we should think that the audience must have adequate reason to think that he takes the speaker as she intends. We can turn the story around to show this. Imagine, then, a case in which it is Carlo who doesn’t know what measles spots look like; it just pops into his head that Anna means to convey that the child has measles. Again, this doesn’t seem to be a case in which Anna communicates that the child has measles. But again here Anna lacks adequate reason for thinking that Carlo will take her as she means to be taken. Carlo’s ignorance of what Anna hopes to trade on (the meaning of certain sports) prevents Anna from being in the position she needs to be to have the relevant communicative intention. When there is communication, an appropriate match is effortlessly achieved between what the speaker wants to convey and the audience’s understanding. The match is achieved effortlessly when the speaker has adequate reason to think that she’ll be understood as she intends. And it can seem that the speaker’s having such reason, when she does, depends simply upon her and Carlo bringing shared knowledge to the occasion. In the particular case, Anna and Carlo both know, and know that the other knows, that certain spots mean measles.9 In linguistic communication, we don’t rely upon our sharing with an audience knowledge of natural meaning to provide us with adequate reason to believe that our distinctively communicative intentions will be successful. So what do we rely upon? Well, surely upon knowledge of what Grice called non-natural meaning—upon artificial meaning. This is what speaker and audience bring to linguistic exchanges. So prior shared 9

Those who hold a certain sort of conventionalist account of language (from which I meant to distance myself at n. 3 above) may think that we have common knowledge of meaning (which is to say that S and A not only know that the other knows, but also know that the other knows that they know, and that … ad infinitum). I argue against this in Hornsby (forthcoming). For present purposes, it is enough to note that the argument for a requirement of S and A’s sharing knowledge delivers only that each of them knows that the other knows, and so lends no support to conventionalist accounts which work with the notion of common knowledge.

117 knowledge of what artificial devices mean would seem to be necessary for communication as we know it. A human natural language consists of a set of such artificial devices, systematically related—of words.

§ 3.1. My strategy has been to look at communication working through natural meaning. This enables us to see the part that prior shared knowledge plays in achievements of the kind of effortless understanding that characterizes human communication. Thus can we understand why one may think, as Dummett does, that in a central case of language use, the parties share knowledge of a language. Still, the materials that we introduce, in order to appreciate the part that prior shared knowledge plays, are got not from Dummett but from Davidson. Two different claims of Davidson belong in this line of thought. One concerns the intentions characteristic of a speaker. The other is a claim about what is presupposed to someone’s having such intentions. Evidently the latter claim is crucial. In the strategy I used, we have to think realistically about how communicative intentions can be realized. We think realistically inasmuch as we resist the idea that human communication could work through magic, or that human beings who communicate with one another might rely on god-like, telepathic powers. These aren’t real possibilities. (Notice that in a similar way, if we take perception to require the operation of senses, we do not allow for the possibility that perception might work through magic, or that human beings might have god-like, extra-sensory perceptual powers. Again these are not real possibilities.) No doubt a realistic account of linguistic communication will have more to say about shared linguistic knowledge. Presumably it will recognize, on the part of infant human beings, a predisposition to learn the language of some community and a readiness to learn the language of the particular community among whom they find themselves. Recognizing these things helps to show why natural languages should be systematic (in Davidson’s sense), and it provides some insight into how shared linguistic knowledge is arrived at. A philosophical account of language can then be an account of an actual phenomenon present in the world wherever human beings are. And once it is allowed that the use of language is an actual phenomenon, I think that it becomes hard to take seriously Davidson’s idea

118 that shared knowledge is merely “a practical crutch” to linguistic communication.

§ 3.2. But more must be said. For Davidson himself is very certainly concerned to be realistic about the use of language. The examples which his own argument takes off from are real enough (examples of the sorts we encountered in §1). And we don’t have a complete response to Davidson until such examples are accommodated. So we need now to look again at Davidson’s claim (P)—the claim that prior knowledge of a shared language is neither necessary nor sufficient for communication. When I distinguished a claim about the sufficiency of prior knowledge from a claim about its necessity, what was at issue was the role of such knowledge in particular communicative exchanges—exchanges in which particular words are uttered by a particular speaker at a particular time. But in enquiring into how a speaker can have adequate reason to think that they will be understood, our sights have been set upon more than particular communicative exchanges. We think now about what puts people in a position to communicate. We consider the epistemic situation of language users. I have suggested that shared knowledge of meaning must then come into the picture. But I have not suggested that it is all that is required for communication. And obviously, the kind of effortless understanding which can be achieved through shared knowledge does not depend upon such knowledge alone: it depends upon people’s capacity to have, and to recognize, communicative intentions. (Consider that Anna’s communicating to Carlo depends at least upon Carlo’s being apt to think that Anna means to get something across.) Now the sorts of examples which Davidson uses to show that knowledge of a shared language is not sufficient for communication are examples in which, on anyone’s accounting, more comes into play than simply an ability to recognize communicative intentions. We saw that in Dummett’s accounting of these examples, the speaker will be said to rely upon having the audience recognize that they want to convey something beyond what their words literally mean. Seeing how it is that speakers can rely upon this, we shall find that there is no problem about treating examples which might be taken to show that knowledge of a shared

119 language isn’t necessary in the case of some particular communicative exchange. So how are speakers able to convey thoughts beyond those that are expressed by the words in a language that their audience shares? Well, we can look to Davidson for the answer. Davidson says that understanding is achieved through the exercise of imagination and wit, through appeal to general knowledge of the world, and through awareness of human interests and attitudes. This is surely right. Our capacities for recognizing communicative intentions would be very much more limited than actually they are if we lacked imagination, general knowledge, and so on. Imagination and general knowledge can be called upon when someone misuses a piece of language. Consider again the example of Bush (at (M) above). In order to understand his utterance, we need only a little knowledge of what he might have been minded to say. Having such knowledge, it is easy enough to correct for his mistake, so that Bush achieves his intended understanding in spite of his failure to use an English word which means what he meant.10 We then take him to mean ALLEVIATE, even though the word he used was “elevate”. (Here is an analogy. If you know that I want a sip of grappa, then you might, without my realizing it, switch the glass of water that I mistakenly suppose contains grappa for one that actually contains grappa. Thanks to your good offices, I get a sip of grappa, in spite of my mistake.) It is true that at the particular point when Bush uttered a certain word, knowledge of a shared language is of no help to his audience. The fact that “elevate” means elevate (knowledge of which is presumably shared by some speakers of English) is not something one needs to know in order to know Bush’s intent. But that could hardly show that one’s being in a position in which one can take Bush to mean what he meant to be taken to 10

In speaking of correcting for Bush’s mistake, I may seem to assume a prescriptivism about language which Davidson denies. But I think that the only prescriptivism to which such talk of mistakes commits one is the sort that we are bound to accept as soon as there is a question whether someone knows, or is ignorant of, the meaning of some word. When Davidson charges Dummett with prescriptivism, he takes the prescriptivist to hold that those in the social swim know best what words mean, and that linguistic change is on the whole undesirable. It may be that Dummett for his part holds some version of these claims. But they need not be at issue here.

120 mean does not depend upon one’s knowing a language, as well as upon a great deal else. The fact that verbal communication relies upon a great deal more than sharing a language is one that Davidson repeatedly emphasizes. As he puts it when replying to Dummett: We need to appreciate the extent to which understanding, even of the literal meaning of a speaker’s utterances, depends on shared general information and familiarity with non-linguistic institutions (a “way of life”) (Davidson 1994: 119). The notion of “literal meaning” here could be the notion of meaning in a language of which knowledge was shared. If so, what Davidson asks us to appreciate here is how rarely communication conforms to the philosophers’ usual paradigm, according to which what is communicated can be deduced from knowledge of literal meaning—of meaning in the language shared by speaker and audience. We can take Davidson’s point. We can allow that it is only relatively rarely that we can identify the literal meaning of a speaker’s utterance with what the speaker got across. When we recognize that people’s conveying something other than what their words literally mean is utterly pervasive, we see that we should need very much more by way of resources than a theory from which it could be derived what any of the sentences of a particular language mean if we were to give an account of all that is actually communicated when language is used. (Or perhaps we see that no account could be given: perhaps we cannot codify that which it takes imagination and wit to appreciate.) Still, this does not show that a certain capacity we have to understand, one which we have when we are able to know the literal meaning of utterances, is not among the things we rely upon as communicators.

Conclusion When Davidson treats communicative intentions as fundamental, he finds a social element in speaker meaning. When Dummett treats cases in which

121 the words of some language are used in conformity with shared norms as central cases, he finds a social element in linguistic meaning. It is obvious that there is no problem about finding social elements at both places. But in the picture I have reached, these are not separable social elements. For the states of mutual understanding which are the upshot of successful linguistic communicative intentions are characteristic of uses of language. And a language—a system of artificial signs, whose meanings are learnt— is a device by which people intentionally make themselves understood. Dummett has no truck with Davidson’s claims about speaker meaning, about communicative intentions. He thus fails to address Davidson’s argument, and ignores one aspect of the social element in language. If I am right, he also misses out on a line of argument on his side, against Davidson. One of Davidson’s charges against Dummett is that, by locating a social element in shared languages, he simply helps himself to an unexplained notion of meaning. Well, the notion to which Dummett helps himself is, I think, just the notion that we need—or more exactly the notion which we need to connect with speakers’ knowledge—if we are to understand realistically how linguistic communication is possible. So far as the charge is concerned, I should have thought that Davidson, as much as anyone, has helped to cast light of the sort that can be cast on the notion of meaning at issue.

References Davidson D. (1984), “Communication and Convention”, in: D. Davidson, Inquiries into Truth and Interpretation, Oxford University Press, Oxford, 265-280. Davidson D. (1986), “A Nice Derangement of Epitaphs”, in: E. Lepore (ed.), Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, Basil Blackwell, Oxford, 433-446. Davidson D. (1994), “The Social Aspect of Language”, in: B. McGuinness and G. Oliveri (eds.), The Philosophy of Michael Dummett, Kluwer, Dordtecht, 1-16.

122 Dummett M. (1986), “‘A Nice Derangement of Epitaphs’: Some Comments on Davidson and Hacking”, in: E. Lepore (ed.), Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, Basil Blackwell, Oxford, 459-76. Dummett M. (1994). “Reply to Davidson”, in: B. McGuinness and G. Oliveri (eds.), The Philosophy of Michael Dummett, Kluwer, Dordtecht, 257–67. Hornsby J. (forthcoming), “Knowledge of Meaning and Epistemic Interdependence”, in: R. Schantz (ed.), Meaning, De Gruyter, New York. Reimer M. (2004), “What Malapropisms Mean: A Reply to Donald Davidson”, Erkenntnis 60, 317–334.

DAVIDSON ON EPISTEMIC NORMS Pascal ENGEL (University of Geneva, Switzerland)

1. Introduction: norms epistemic and others Prima facie at least, there are, in addition to practical norms, about how we ought to act, epistemic norms, in the sense of correctness conditions for our beliefs such as the following: a belief is correct if and only if it true, one ought to believe only what is based on appropriate evidence or justification, one ought to believe what follows from one’s beliefs, a belief is rational if it is coherent with certain logical principles, and so on and so forth. There are disputes about how many such norms there are and about the proper formulation of these principles (do they have to contain “oughts” and explicit prescriptions or do they involve only permissions?) and about their connections to other notions, such as the notion of value (is believing correctly a matter of believing what one ought to believe, or a matter of what is good to believe?) or the notion of reason (some philosophers claim that they are irreducible to other principles, others claim that they are reducible to the more fundamental notion of reason). These are, again prima facie, distinct from practical norms, norms about the way we should act or about what we ought to do, just as practical reasons for doing things are not the same as reasons for believing things. Philosophers disagree about the relationship between epistemic norms and practical norms, epistemic and practical reasons. Pragmatists, for instance, want to say that the latter can override the former, perhaps that one can reduce the former to the latter. Some philosophers just deny that there are any such norms—they deny, for instance, that there are general conditions under which a belief is correct, or that there is something, truth, to which our beliefs should “aim at”. Other philosophers accept their existence but want to reduce these norms to other properties, such as values, or interests, and claim to have naturalistic analysis of them. In one sense, to specify what the epistemic norms are is to specify in what sense beliefs are justified, at least in the sense in which stating that a belief respects a certain norm appropriate to it is a way of justifying the

124 belief.1 And in so far as knowledge is justified true belief, a theory of epistemic norms belongs to a theory of knowledge or to epistemology. But we can also understand the task of analysing the nature of epistemic norms as a more general kind of inquiry, which would tell what kind of norms a correct theory of knowledge should satisfy. In this sense, for instance, the choice between a theory according to which knowledge should be defined as a form of internalist justification and a theory according to which knowledge should be defined in an externalist way as reliability is a choice between two distinct conceptions of epistemic norms. Or the question whether an epistemology should depend upon a foundationalist or a coherentist conception of justification is matter of confronting two kinds of epistemic norms. We can call this more general kind of inquiry not epistemology proper, but meta-epistemology. Meta-epistemology would be to epistemology what meta-ethics is to ethics. In this respect we can raise for epistemic norms some questions which are comparable to those that meta-ethicists raise about moral norms, and in particular the following. (a) Their nature. What are the norms? How to specify them? As statements about reasons? As rules? As commitments? As permissions? As prescriptions and imperatives? Imperatives of what sort? Hypothetical ? Categorical ? (see in particular (Skorupski 1990), (Broome 2000), (Millar 2005)). (b) Their regulation. How are they applied to particular cases? How are they followed, or, to use a generic term to designate their conditions of application, how do they regulate our beliefs and our epistemic practices? (see in particular (Velleman 2000), (Shah 2003)). (c) Their hierarchy. What justifies them? Which ones are more basic, which ones are derived? For instance is truth more basic than justification or coherence for beliefs? (d) Their ontology. Are they genuine properties of our beliefs? Are they essential to the nature of our beliefs? Are they real (the object of cognitive attitudes) or only the product of our psychological attitudes 1

Although to say that a belief is justified, or that there are reasons for holding it is not the say thing as saying that it obeys certain norms. See (Engel forthcoming).

125 (in the non cognitivist sense)? How to place them within the natural and causal order? (Blackburn 1998), (Jackson 1998). In this article, I would like to explain how Davidson deals with such questions about epistemic norms.2 I shall first describe Davidson’s views about mental norms and his form of normativism, the view that norms are in some sense essential to the mind3, and explain why these norms are, in a sense, no norms at all, but idealised principles of interpretation. I shall raise some difficulties for Davidson’s version of normativism, and try to suggest a better version.

2. Davidson on mental norms There is a lot, in Davidson’s writings, about the nature of norms of rationality and about normativity in general, both practical and theoretical, but he never addresses the questions (a)-(d) above directly. In order to understand his approach, we can take our starting point from his essay “Three Varieties of Knowledge” (Davidson 1991). There he distinguishes three sorts of knowledge: knowledge of the world, knowledge of our own minds, knowledge of other minds. Davidson’s discussion of norms and normativity occurs mainly in the first and in the third kind of context: knowledge of the world or epistemology, for instance when he discusses empirical content and the coherence theory of knowledge and scepticism (Davidson 1983), and knowledge of other minds in the context of his theory of interpretation. The distinctiveness of his approach, however, consists in the fact that he does not discuss these issues separately, but within his theory of interpretation. The third and the first sort of knowledge 2

In the past I have already tried to deal with such questions, often in papers to which Davidson himself gave a reply (Engel 1999, 2001), (Davidson 1999, 2001a). It is his strong disagreement with my interpretation of his views about the normativity of the mental and with my our conception of it which prompts me to try again here. Sadly, he is not here any more to reply, but I hope to have been, at least in the interpretive part of this article, more faithful to his views than he thought I was before. 3 This label is used, as far as I know, by (Wedgewood 2007), (Zangwill 2005) talks of “normative essentialism”

126 are intrinsically connected, since the conditions of our interpretation of others’ beliefs are the general conditions of interpretation of belief as a candidate to knowledge. According to his famous slogan, interpretation is “epistemology seen in the mirror of meaning” (Davidson 1975). It is in the context of interpreting others that the question of the nature of epistemic norms arises. For Davidson, however, there is no point in distinguishing various kinds of norms. There are general norms of rationality which make interpretation possible, which are also “norms of the mental” because having a mind is being susceptible to be interpreted. Davidson thus defends a form of normativism, understood as the view that the mind is essentially normative and mental contents cannot be specified without mentioning normative conditions: If we are intelligibly to attribute attitudes and beliefs, or usefully to describe motions as behaviour, then we are committed to finding, in the pattern of behaviour, belief, and desire, a large degree or rationality and consistency (Davidson 1980b: 236-7). The argument for anomalous monism rests in part upon normativism. One of the main reasons why there are no strict psychological laws and no psychophysical laws is that the norms of rationality that we have to use in interpretation have no “echo” in physical theory (Davidson 1970). Normativism is also the basis for two other central doctrines of Davidson’s philosophy, externalism and holism: What I think is certain is that holism, externalism, and the normative feature of the mental stand or fall together. […] There can be not serious science of the mental. I believe the normative, holistic, and externalist elements in psychological concepts cannot be eliminated without radically changing the subject (Davidson 1995a: 122-123). Now, what are exactly the norms of rationality or principles of interpretation which are held to be constitutive of our understanding of the mind? There is, in the first place, a principle of Coherence, which prompts the interpreter to discover a degree of logical consistency in the thought and in the actions of the speaker. Here the principles are the principles of

127 logic for belief, and the basic principles of decision theory for actions (Davidson 1980b). In the second place, there is the Principle of Correspondence, which prompts the interpreter to take the speaker to be responding to the same features of the world that he (the interpreter) would be responding to under similar circumstances. Both principles can be (and have been) called principles of charity: one principle endows the speaker with a modicum of logic, the other endows him with a degree of what the interpreter takes to be true belief about the world. In this sense successful interpretation necessarily invests the person interpreted with basic rationality. It follows from the nature of correct interpretation that an interpersonal standard of consistency and correspondence to the facts applies to both the speaker and the speaker’s interpreter, to their utterances and to their beliefs (Davidson 1991). As we shall see, these two kinds of principles operate at a very general level, since they do tell us how we can use these principles in particular cases. Sometimes, however, Davidson is a little bit more specific. It is the case when he talks about the role of the Principle of total evidence, which is one of the best candidates for being an epistemic norm for beliefs. The principle of evidence is the relatively uncontroversial principle that belief should be based on appropriate evidence, and that the formation of belief should be regulated by the best total evidence available. Davidson describes it in the context of his analysis of the irrationality of weakness of the will and of self-deception: Weakness of the will is analogous to a certain cognitive error, which I shall call weakness of the warrant. Weakness of the warrant can occur only when a person has evidence both for and against a hypothesis; the person judges that relative to all the evidence available to him, the hypothesis is more probable than not; yet he does not accept the hypothesis (or the strength of his belief in the hypothesis is less than the strength of his belief in the negation of the hypothesis). The normative principle against which such a person has sinned is what Hempel and Carnap have called the requirement of total evidence for inductive reasoning: when we are deciding among a set of mutually exclusive hypotheses, this requirement enjoins us to give credence to the hypothesis most highly supported by all available relevant

128 evidence. Weakness of the warrant obviously has the same logical structure (or, better, illogical structure) as weakness of the will; the former involves an irrational belief in the face of conflicting evidence, the latter an irrational intention (and perhaps also action) in the face of conflicting values (Davidson 1986: 201-202). But although in this case Davidson gives us an account of how this norm is violated and how its violation produces irrational behaviour, most of the time his characterisation of rationality is very general in the sense that he does not give us any theory about the application of rational principles. And there is a good reason why it is so. A number of interprets of Davidson (in particular (Wiggins 1987), (McDowell 1986), (Child 1992)) have derived form his work the thesis of the uncodifiability of rationality, which Child, for instance, describes thus: There is no fixed set of rule or principles from which, together with a statement of the circumstances of any particular case, we could deductively derive a complete, detailed specification of what one ought to do or think in that case. Since the norms of rationality are the principles governing the interpretation of a subject’s words and attitudes, we could express the same point about the uncodifiability of rationality as a point about interpretation: there is no definite set of rules or principles for arriving at the best interpretation of an agent. When we interpret others, we strive to make sense of them: in doing so we draw on our own conception of rationality to form judgments in each particular case; and we can draw on those resources without limit. So, in describing and applying our conception of rationality, there is no stage at which we can say that the canons of rationality have been exhaustively enumerated, that there is nothing more to add (Child 1992). If rationality is uncodifiable, not only there is no privileged set of principles which have to be invoked in any case of individual interpretation, but also there cannot be any theory of how these principles have to be applied to specific cases. It would be absurd, for instance, to try to draw a list of the kinds of norms that we have use in a specific case, or

129 to try to converge on certain norms. It is enough to say that there are certain general principles of rationality: The issue is not whether we all agree on exactly what the norms of rationality are; the point is rather that we all have such norms, and that we cannot recognize as thought phenomena that are too far out of the line. Better say: what is too far out of line is not thought. It is only when we can see a creature (or ‘object’) as largely rational by our own lights that we can intelligibly ascribe thought to it at all, or explain its behaviour by reference to its ends and convictions. Anyone […] when he ascribes thoughts to others, necessarily employs his own norms in making the ascriptions. There is no way he can check whether his norms are shared by someone else without first assuming that in large part they are; to the extent that he successfully interprets someone else, he will have discovered his own norms (nearly enough) in that person (Davidson 1990: 97-98). The consequences of such an account of the norms of rationality are twofold. In the first place, as many critics (e.g. Peacocke 1992) have noted, there is a risk of a certain amount of subjectivity in interpretation, in the sense that every interpretation of another’s thought and action is irreducibly individual and contextual. In the second place, the uncodifiably of rationality means that the rational norms which are supposed to rule interpretation cannot give rise to any genuine prescription. Thus, the norm of coherence, which enjoins us to be consistent, cannot give rise to any specific prescription except very general advices such as: “Be consistent”, or: “Try to find consistency in another’s belief”. They do not tell how and why we have to be consistent in a given case. This is why, in particular, Davidson is so insistent in taking the principle of charity as a principle on maximisation of understanding rather than as a principle of maximisation of truth or of agreement. A direct consequence of this is that the norms of rationality, according to Davidson, are not regulative: they are not prescriptive and they do not give us rules about how to apply them.

130 In this sense, as Timothy Schroeder (Schroeder 2003) has well pointed out, the norms of rationality are not normative. Schroeder distinguishes usefully between two notions of norms4: (a) as categorisation or classification schemes, in the sense of general idealised principles of description (b) as force makers, that is as prescriptions or governance principles giving us aims to follow. For a norm to be genuinely “normative”, it has to be capable not only of resting upon general principles in the sense of (a), but also of having a certain force to move us to something or to think something. Another condition of a genuine norm, emphasised in particular by Railton (Railton 1997) is that it should also give us the appropriate freedom not to follow it. Norms, at least for actions, are such that, even when they are imperative in the categorical sense, a subject can have the possibility of not following it. Even the moral ought or the categorical imperative in the Kantian sense cannot be a norm unless the subject can have the capacity of disobeying it. In other words, a norm which we have no other choice but to follow is not a norm at all. Such a “norm” may state a necessity, but if it is a law, it merely describes what there is, and does not prescribe where there ought to be. The laws of logic, for instance, in Davidson’s sense, are not such that we could have the possibility of not to follow them. The norms of rationality in the sense of general framework principles that any rational agent has to follow are not normative in this sense. They are descriptions of an idealised competence, but they do not tell us how to exercise this competence. And a theory of mind, such as Davidson’s, based on normative principles of this sort, is not normative at all. For a theory of mind to be normative in the genuine, it is necessary that it make use of (a) a categorisation scheme for which there is some force-maker, and (b) the fact that the scheme has a force-maker. As Schroeder says, Davidson’s theory of mind, although it invokes normative principles in sense (a) is not 4

For a similar distinction see also my article (Engel 1992), where I distinguish three senses of normativity: as idealised principles, as prescriptions, and as evaluative principles or values.

131 normative in the sense (b). The norms of rationality, to use Schroeder’s phrase, do not have any “normative oomph”. To see in what sense Davidson’s conception of normativity is only committed to the first sense (a) of norm as a categorisation scheme and not the second (b) as a force maker, it is useful to consider the analogy that he draws between the process of interpretation and the act of measuring length or other quantities: We may think of an interpreter who aims to understand a speaker as matching up sentences of his own with the utterances and states of mind of the speaker. The totality of evidence available to the interpreter determines no unique theory of truth for a given speaker, not just because actually available evidence is finite while the theory has an infinity of testable consequences, but because all possible evidence cannot limit acceptable theories to one. Given the richness of the structure represented by the set of one’s own sentences, and the nature of the connections between the members of this set and the world, we should not be surprised if there are many ways of assigning our own sentences to the sentences and thoughts of someone else that capture everything of significance […] The situation is analogous to the measurement of weight or temperature by assigning numbers to objects. Even supposing there are no errors of measurement, and that all possible observations have been made, an assignment of numbers to objects that correctly registers their weights is not unique: given one such assignment, another can be produced by multiplying all the numbers by any positive constant. In the case of ordinary temperature (not absolute temperature), any correct assignment of numbers can be converted to another by a linear transformation. Because there are many different but equally acceptable ways of interpreting an agent we may say, if we please, that interpretation or translation is indeterminate, or that there is no fact of the matter as to what someone means by his or her words. In the same vein, we could speak of the indeterminacy of weight or temperature. But we normally accentuate the positive by being clear about what is invariant from one assignment of numbers to another, for it is what is invariant that is empirically significant. The invariant is the fact of the

132 matter. We can afford to look at translation and the content of mental states in the same light (Davidson 1991: 214). The analogy between norms and principles of measurement helps us to see why Davidson’s notion of norm is the purely descriptive one (a) and not the prescriptive one (b). In this sense, as Schroeder points out, Davidson’s theory of mind is not normative at all: His interest in rationality is thus an interest in it only insofar as it picks out a certain set of propositional attitude clusters (those which it would be fairly rational to hold) and distinguishes them from a different set of propositional attitude clusters (those which it would be wildly irrational to hold). The fact that the patterns exhibited by the propositional attitudes of a rational organism are normatively commanded—that there exists a force-maker for the patterns—is of no significance in Davidson’s theory (Schroeder 2003).5 Let is now consider, in the light of this, Davidson’s specific claims about epistemic norms, norms about belief and knowledge. In face of the familiar claim that truth is a norm of belief in the sense that the point or aim of belief is truth ((Dummett 1959), (Williams 1969), (Engel 1999, 2001)), Davidson is quite explicit that truth is not, for him, a norm at all: our beliefs are supposed to be true or false, but there is no norm of truth which our beliefs should follow: When we say we want our beliefs to be true, we could as well say we want to be certain that they are, that the evidence for them is overwhelming, that all subsequent (observed) events will bear them out, that everyone will come to agree with us. It makes no sense to ask for more. But I do not think it adds anything to say that truth is a goal, of science or anything else. We do not aim at truth but at honest justification. Truth is not, in my opinion, a norm (Davidson 1999: 461).

5

See also (Glüer 2000), (Glüer and Wikforss 2007).

133 It could not even be said that our belief should follow an evidential norm, in the sense of the principle of total evidence mentioned above. That our beliefs “should” be based upon appropriate evidence is a framework requirement on belief, but it does not carry, for Davidson, any particular ought about belief, just as the fact that our beliefs “should” be true does not carry any particular ought for them, in the normative force sense. Just as there are no genuine epistemic norms about belief for Davidson, there are no genuine norms in the ontological sense. As I mentioned in the introduction of this article, the ontology of norms is concerned with questions such as: are the norms real? Are they “out there”? Where to place them within nature? Are they genuine properties of our beliefs? In order to understand Davidson’s position on this point, we have to turn to his writings on the nature of moral values (see in particular (Davidson 1983, 1995b)). The debate about moral values in meta-ethics is framed by the distinction between various kinds of realism about values (or of cognitivism) and various kinds of anti-realism (or non cognitivism). In his 1995 essay on the objectivity of values Davidson rejects this opposition. His stance is neither the anti-realist one that they are mere expressions of our psychological attitudes or feelings, nor the realist one that they are genuine features of the world. For Davidson values are “nowhere” to be found. His view, however, does not amount to Mackie’s (Mackie 1977) “irrealism” in the meta-ethical debate, according to which our ethical statements about values are all false. Our statements about values are susceptible to be true, and in this sense Davidson is a realist, but they are not true in virtue of any state of affairs in the world. They are true because they are objective, and their objectivity depends upon our shared beliefs: Objectivity depends not on the location of an attributed property, or its supposed conceptual tie to human sensibilities; it depends on there being a systematic relationship between the attitude-causing properties of things and events, and the attitudes they cause. What makes our judgments of the “descriptive” properties of things true or false is the fact that the same properties tend to cause the same beliefs in different observers, and when observers differ, we assume there is an explanation. This is not just a platitude, it’s a tautology, one whose

134 truth is ensured by how we interpret people’s beliefs. My thesis is that the same holds for moral values. Before we can say that two people disagree about the worth of an action or an object, we must be sure it is the same action or object and the same aspects of those actions and objects that they have in mind. The considerations that prove the dispute genuine—the considerations that lead to correct interpretation—will also reveal the shared criteria that determine where the truth lies (Davidson 1995b: 47). Here again, objectivity is a by-product of the interpretive situation. The objectivity of values is a projection out of our shared beliefs about others and about ourselves: If, instead of asking where values are, we turn to the problem of understanding what it is like to judge that an act or object or institution is morally desirable or ought to exist or is obligatory, we realize that we must be attributing some property or other to an entity or group of entities. The semantic nature of such judgments is clear: we are classifying one or more things as having a certain property. The thing or things must either have that property or not (assuming the things exist). How do we tell what the content of a particular moral judgment is? This is a question of interpretation, of the understanding by one person of the utterances of another, since there is no other context in which the content of a judgment can be agreed to or disputed. To take up the position of an interpreter is consciously to assume the status anyone with thoughts and attitudes must be in, for the attitudes of a person have a content—are interpretable—only if that person is in communication with others; only interpreters can be interpreted. Thus by explicitly introducing the interpreter we complete in microcosm the social situation which alone gives content to the idea of being right or wrong about a shared public world (Davidson 1995b: 48). If we now apply this conception of ethical and practical values to epistemic norms about belief, such as evidence and truth, what would be Davidson’s position? To my knowledge he never addresses the question in this form, but it is reasonable to predict that he would hold, for instance about the

135 “norm” (in sense (b) above) for belief that our beliefs are such that they are correct only if true, or about the “norm” that our beliefs should respect evidence, that these norms are objective not in the ontological sense, but in the sense that they are part and parcel of what is presupposed by correct interpretation. This could be called, if we needed a label, a form of objectivism without realism about epistemic (and other) norms.

3. Problems for Davidson’s normativism Having tried to describe Davidson’s position about norms, I would like now to point out its difficulties. The first problem is the one which I have already outlined. Davidson’s normativism is limited to (a)-type norms, and excludes (b)-type norms. The normative principles of rationality are not norms in the sense that that they do not prescribe or regulate any actions or beliefs. They could be conceived either as very general features of actions and beliefs or as very general features of what our concepts of actions or belief are. In this sense, as he himself recognises, Davidson’s theory of mind is not normative at all. Now if one intends to limit the use of the notion of “norm” when speaking about the mental to the notion of general framework principles of rationality, the fact that these are not normative in the usual sense should not be a problem and is at best a case of equivocation. After all Davidson himself denies that truth, or any other notion, is a “norm”.6 If, however, one intends to claim, as I do, that the thesis of the normativity of the mental implies a stronger notion of norm, this is a problem, for the notion of normativity with which we end up happens to be fairly shallow. The second difficulty which is raised by Davidson’s version of normativism concerns the uncodifiability of rationality. How far does it go? Interpretationism, the view that the mind is essentially interpretable, presupposes that we have the necessary aptitudes to distinguish, within a certain behavioural or linguistic pattern what is rational or not. But where do these aptitudes come from? They must belong, in one sense or another, to our natural equipment. If they are not to be mysterious, they have to be 6

For similar views, see (Glüer 2000), (Glüer and Wikförss 2006), (Rey 2007).

136 based on certain psychological capacities ((Braddon Mitchell and Jackson 1997), (Engel 1996)). A third problem concerns the ontology of norms according to Davidson. On his view, although they are objective, statements about norms of rationality do not have genuine truth conditions. They are merely posits within an instrumentalistic scheme of interpretation such as the one which is described by him as comparable to a measurement scheme. The uncodifiability of rationality and the subjectivity which affects interpretation imply that Davidson is closer to a non-cognitivist or expressivist view about rational norms. But the combination of non-cognitivism about the norms of the mental and the thesis that the norms of the mental are essential to mental states and content (what I have called above normativism) gives rise to a threat of irrealism or eliminativism about the mental. Why? The problem for non cognitivism about normativity is that if non cognitivism about normativity is correct and if mental contents and states are essentially normative, there is no such thing as satisfying normative constraints, and therefore, it would seem, no such things as beliefs. For to satisfy a constraint is to have the relevant property. If belief is subject to normative constraints, being a believer requires that one has the relevant normative properties. For example if someone believes P and believes that if P then Q, then they have the property of being such that they ought to believe that Q. Again, for any believer, there is a way their beliefs ought to evolve under the impact of putative information. But if non cognitivism is true, there are no normative properties to have or to fail to have (Jackson 2000). For these reasons, Davidson’s normativism about the mental runs the risk of being like Hamlet without the Prince. A theory of the normativity of the mental must give us an account not only of the kind of norms to which various kinds of mental states are subject, but also an account of how these norms regulate the formation of mental states. If there is indeed such a thing as being correct or incorrect in one’s beliefs, and if we want to have an account of the epistemic norms which regulate inquiry, we need to have a more full-fledged conception of these norms. It is quite understandable that philosophers have been reluctant to embrace a more substantial conception of the normativity of the mental and of the norms which rule the epistemic domain. They have feared that talk of norms in these

137 domains could suggest that having a belief with a certain content implies eo ipso that one incurs thereby certain sorts of requirements or imperatives. But certainly when I believe that it rains, or that Nicolas Sarkozy is proud of himself, I do not incur any sort of obligation. The mere fact that beliefs are supposed to be true or false does not create for believers any obligation with respect to their beliefs. So why insist that there are norms of the mind, or epistemic norms? I want to claim that, although these worries—which Davidson himself expresses when he denies that truth is a norm—are ill placed once we understand better what kind of normativity here is in place. It seems to me that it is neither the weak (a)-type normativity of rational principles nor the strong normativity of a categorical ought.

4. Epistemic norms with normativity What would a more full-fledged conception of epistemic norms look like? It seems to me that it should at least satisfy the following desiderata. (1) It should specify what are the norms, and provide their proper formulation (if they are to be couched in terms of oughts, what kinds of oughts should they be?) (2) It should explain what is the point of conforming the standard of correctness that apply to normative statements; in other words it should give us the rationale of the specific norms. This is in large part done when we relate the norms to the most general properties of the states involved (for instance a belief is correct if it is true; but how is the notion of truth related to that of correctness?)7 (3) It must explain how there can be an agreement between the normative standards and the normative prescriptions of rules which follow from the formulation of the norms (there should be an harmony between 7

Although I cannot detail this point here, it presupposes that to any norm there corresponds a general truth (hence a descriptive statement) about a given state, which grounds, in the appropriate way, the prescription which flow from it. In this respect I subscribe to Husserl’s conception in his Logical Investigations according to which a norm exists in virtue of the truth of some independent descriptive statement.

138 the abstract formulation or the truths upon which the norms are based and the prescriptive principles which derive from it, or if one prefers, how the normative connects to the non normative. (4) Hence it must specify how norms regulate reasoning, thinking, acting. This involves in part understanding how the norms can be reasons for us to believe certain things, not simply in the objective or externalist sense of there being reasons to do or think such and such, but also in the internalist sense of there being reasons for us to do or to believe such and such. This is in part explaining how the norms can have some motivating force (I take this to be analogue, for epistemic reasons, of the problem of motivation by moral norms in moral psychology, see (Engel 2005a)). (5) It must specify the appropriate ontology of norms. Normativist theories of the mind—the view according to which normativity is in some sense an essential dimension of the mental—differ in many respects. I do not propose here to do more than a very short review of the main varieties (see Engel to appear for a fuller account). In the first place theories of norms can either be high profile or low profile.8 High profile norms are the norms of rationality which apply to mental states in general in virtue of principles of rationality, such as those of charity and coherence mentioned above. Davidson’s conception of the normativity of the mental is a typically high profile one. Low profile norms are the norms which affect mental contents in virtue of the concepts which figure in them, and the rules which determine these contents. Conceptual role semantics, in the version of this view according to which conceptual content is determined by the rules of inferences that a rational agent would follow, is a low profile conception. For instance such a theory would say that a belief content is subject to a specific norm if it involves the logical concept of conjunction and its usual associated rules of inference ((Peacocke 1992) is a conception of this sort, but (Brandom 1994) is too, in spite of all the differences).

8

This distinction is inspired by the one made by (Bilgrami 1992) about norms of the mind. It has also some affinities with the one proposed by Glüer and Wikforss (to appear) between “content engendered” and “content determining norms”

139 In the second place, normativist conceptions of the mental differ depending upon whether they take the normative load to bear upon the contents themselves (as with the low profile conceptions) or the attitudes that one has to these contents (Boghossian 2003). Davidson is neither a theorist of the former sort (he does not hold that the mental is normative in virtue of the concepts which figure within contents) nor a theorist of the second sort (he denies that norms could be attached to beliefs, desires, or intentions as such). In the third place, normativist conceptions diverge on the nature of the normative load involved in mental contents. Some take this involvement to be a matter of real essence (a mental content would not be a content if it did not have normative properties), others a matter of our understanding of mental contents.9 Davidson is certainly a theorist of the second, and not of the first kind, since he does not hold that mental contents have real properties, apart from their being interpreted. My own version of normativism (see (Engel 1999, 2001, 2005b, forthcoming))—which I can here only outline—locates the normative load at the level of the attitudes and not at the level of the concepts involved in the contents. A belief content, on this view, is not normative because it involves a certain kind of concept which are subject to inferential norms, but because it is subject to certain specific norms attached to the attitude of (for instance, but the same would be true of intentions, desires) belief. Which norms? The primary and the basic one is the truth norm: (B) A belief is correct iff it is true from which one derives a prescriptive norm : (B*) One ought to believe that P only if P is true (B) is a semantic condition, but there are inferential or rational conditions as well such as:

9

For this distinction see (Wedgewood 2006).

140 (R) A belief is correct iff it is rational from which one derives other prescriptions such as : (R*) For any A, B, if B is entailed by A, then avoid believing A while not believing B and similar ones about coherence. A lot more should be said about the proper formulation of (B) and (R), and there is a lot of discussion about this (see (Wedgewood 2002), (Boghossian 2003), (Engel 2003, 2005b), (Shah 2003), (Steglish Petersen 2006), (Hattiangadi 2007)). In a sense, Davidson recognises (B)-type norms on belief. But as we saw, he does not accept (R)-type ones. Like (Peacocke 1992), I distinguish the norms at the level of reference, like (B), and those at the level of sense, like (R), and I take them to be strongly associated (unlike (Brandom 1994), who gives primacy to (R) kind of norms). (B) and (R) are just statements of the norm, which in a sense flow from what belief is and what minimal constraints it obeys, and are meant to say something about questions (1) and (2) of the previous section, but these formulations do not answer questions (3) and (4). In order to answer these questions we need to ask: how is belief regulated by the correctness condition? Here there are several kinds of answers : one can claim that belief is, in some sense, directly regulated by (B) (Wedgewood 2000), or that it is regulated by a specific attitude, the intention to have only true beliefs (or consciously aiming at truth) (Velleman 2000), or that there is a psychological feature of belief, its “transparency” (the fact that the best first person test about our believing that P is answering the question whether P) (Shah 2003). I favour a version of the last position. How can the norm governing belief have motiving force? How can we answer the question which I raised about Davidson’s uncodifiability of rationality thesis: from where, psychologically speaking, do our rational abilities come from? Here I would defend a view similar to Wedgewood’s (Wedgewood 2006). To be able to have certain concepts and attitudes, and to be able to be moved by the norms involved in these attitudes, the subject must have certain dispositions to conform the principles of rationality that

141 feature in the correctness conditions for a given concept of attitude. But we have to be careful in specifying what these dispositions are, and their connections to the norms. There is a certain view, which might be called dispositionalism, but of which functionalism about the mental is but a version, which attempts to define an attitude through the set of set of characteristic dispositions associated to it. Notoriously dispositionalism encounters a problem. Suppose that we say that a normative principle on belief like (R*)—For any A, B, if B is entailed by A, then avoid believing A while not believing B. If we take it to consist in a disposition, how can it be explained by a mere regularity? Can the ideal of rationality expressed by (R*) be explained simply by the value we place upon conforming to such standards? No. And that was in part Davidson’s point about there being no norm of truth to which we aspire and which would regulate our actual believings. If we have to explain the connection between the normative principle (R*) and our behaviour we have to postulate not simply dispositions to follow such rules, but also rational dispositions. In other words the dispositions themselves have to be specified in normative terms. Subjects must have a disposition to conform to the norm of correctness of belief or the implication principle. Why is that so? (Wedgewood 2006) gives the following (in my view convincing) explanation. Every time we formulate a form of reasoning in wholly non normative terms, it will turn out that form of reasoning in question is defeasible, and that there can be circumstances where it is not rational to engage in such forms of reasoning. Defeating conditions have to be specified. But the very notion of defeating condition is itself normative. The rational disposition must be such that it tends not to be manifested in the absence of defeaters. Now someone might find that this is very close to the idea that rationality is uncodifiable, and in so far as this means that we cannot explain rationality through disposition without invoking rationality itself, I agree. But that does not mean that there is no basic disposition corresponding to the norm of correctness for a concept. For instance consider the disposition that we have to follow the logical rule of Modus Ponens. Well known empirical studies of reasoning show that only 72% of subjects accept instances of Modus Tollens, and that the proportion strongly increases with modus ponens, which suggests that Modus Ponens may be the primary rule for “if”, and there is a possible account of this at

142 the psychological level (e.g. in terms of mental models (Johnson Laird and Byrne 1991)). We can say that we have a rational disposition to accept the MP inferences, and not the MT ones. We can also violate these norms. Modus ponens is in turn justified as a rule of logic. So there is an harmony between the justificatory principles and the psychological dispositions (Engel 2001). Of course a lot more should be said, but I have tried to indicate what form a theory of epistemic and mental norms could take.

Conclusion Davidson was almost right. He was right that epistemic norms are primitive and irreducible, and that they are part and parcel of our understanding of the mind, and of the nature of mind. He was, however, wrong to consider that they are uncodifiable. We can specify them, and the psychological disposition which underlie them. So contrary to what anomalous monism claims, the normative principles have an “echo” in the psychology of agents. Contrary to what Davidson claims, the mental and the epistemic norms do not operate at a high profile level of rationality principles only. They are not framework principles about what our mental states and contents are. They are genuine norms, with normative force, and they can regulate our mental states. Moreover they are not mere interpretative principles similar to the principles of measurement. They are objective.

References Bilgrami A. (1992), Belief and Meaning, Oxford University Press, Oxford. Boghossian P. (2003), “Contents and Norms”, Philosophical Issues 13, 3145. Braddon Mitchell D. and F. Jackson (1997), Philosophy of mind and cognition, Blackwell, Oxford. Brandom R. (1994), Making it Explicit, Harvard University Press, Harvard. Jackson F. (2000), “Non Cognitivism, Normativity, Belief”, in: J. Dancy (ed.), Normativity, Blackwell, Oxford.

143 Child W. (1992), Causality, Interpretation and the Mind, Oxford University Press, Oxford. Davidson D. (1970), “Mental Events”, reprinted in (Davidson 1980b). Davidson D. (1975), “Thought and Talk”, reprinted in (Davidson 1984). Davidson D. (1980a), “A Unified theory of thought meaning and action”, reprinted in (Davidson 2004). Davidson D. (1980b), Essays on Actions and events, Oxford University Press, Oxford. Davidson D. (1983), “A coherence theory of truth and knowledge”, reprinted in (Davidson 2001b). Davidson D. (1984), Inquiries into Truth and Interpretation, Oxford University Press, Oxford. Davidson D. (1986), “Deception and Division”, reprinted in (Davidson 2001b). Davidson D. (1990), “Representation and Interpretation”, reprinted in (Davidson 2004). Davidson D. (1991), “Three varieties of knowledge”, reprinted in (Davidson 2001b). Davidson D. (1995a) “Could there be a science of rationality?” reprinted in (Davidson 2004). Davidson D. (1995b), “The objectivity of values”, reprinted in (Davidson 2004). Davidson D. (1999), “Reply to Pascal Engel”, in (Hahn 1999). Davidson D. (2001a), “Comments on Karlovy Vary Papers”, in (Kotatko, Pagin and Segal 2001). Davidson D. (2001b), Subjective, intersubjective, objective, Oxford University Press, Oxford. Davidson D. (2004), Problems of Rationality, Oxford University Press, Oxford. Dummett M. (1959), “Truth”, reprinted in M. Dummett (1978), Truth and other enigmas, Duckworth, London. Engel P. (1994), Davidson et la philosophie du langage, PUF, Paris. Engel P. (1999), “The norms of the mental”, in (Hahn 1999). Engel P. (2001), “Is truth a norm?” in (Kotatko, Pagin and Segal 2000). Engel P. (2002), Truth, Acumen, Bucks.

144 Engel P. (2005a) “Logical Reasons”, Philosophical Explorations 1, march, 21-35. Engel P. (2005b), “Truth and the aim of Belief”, in: D. Gillies (ed.), Models in Science, King’s College Publications, London, 77-97. Engel P. (forthcoming), “Belief and normativity”. Glüer K. (2001), “Dreams and Nightmares. Conventions, Norms, and Meaning in Davidson’s Philosophy of Language”, in (Kotatko, Segal and Pagin 2000). Glüer K. and A. Wikforss (forthcoming), “Against Content Normativity”. Hahn L.E. (1999) (ed.), The Philosophy of Donald Davidson, Open court, La Salle. Johnson Laird P. and R. Byrne (1991), Deduction, Erlbaum, Hove. Kotatko P., P. Pagin and G. Segal (2001) (eds.), Interpreting Davidson, CSLI Press, Stanford, CA. Mackie J. (1977), Ethics, Inventing Right and Wrong, Penguin, London. McDowell J. (1986), “Functionalism and Anomalous Monism”, in: E. Lepore and B. McLaughlin (eds.), Actions and Events. Perspectives on the Philosophy of Donald Davidson, Blackwell, Oxford. Peacocke C. (1992), A Study of Concepts, MIT Press, Cambridge MA. Railton P. (1999), “Normative Force and Normative Freedom”, in: P. Railton (2003), Rules, Values and Norms, Cambridge University Press, Cambridge. Rey G. (2006), “Resisting Normativism in Psychology”, in: Contemporary Debates in Philosophy of Mind, Blackwell, Oxford. Shah N. (2003), “How Truth Regulates Belief”, Philosophical Review 112, 447-482. Schroeder T. (2003), “Davidson’s Theory of Mind is Non Normative”, Philosophers’ Imprint. Velleman D. (2000), “On the aim of Belief”, in: The Possibility of Practical Reason, Oxford University Press, Oxford. Wedgewood R. (2002), “The Aim of Belief”, Philosophical Perspectives 16, 276-297. Wedgewood R. (2007), “Normativism defended”, in: J. Cohen and B. McLaughlin (eds.), Contemporary Debates in Philosophy of Mind, Blackwell, Oxford.

145 Wiggins D. (1987), Needs, Values, Truth, Oxford University Press, Oxford. Zangwill N. (2005), “The Normativity of the Mental”, Philosophical Explorations 8 (1).

THE PLACE OF ONTOLOGY IN DAVIDSON’S THEORY OF INTERPRETATION Andrea C. BOTTANI (University of Bergamo, Italy)

1. Introduction Among the most important features of Davidson’s theory of radical interpretation is the close connection the theory establishes between matters of language and matters of fact. On the one hand, radical interpretation is shaped by the interpreter’s ontology. On the other hand, successful interpretation is a criterion of ontological acceptability, and interpretive flop is a criterion of ontological mistake. So, radical interpretation makes ontology and semantics—the theory of what exists and the theory of what sentences mean—go hand in hand. I shall try to clarify how ontology and interpretation are connected in Davidson’s approach and why they are allegedly connected as they are. Certainly, the connection depends in part on the formal constraints placed on radical interpretation—the idea that a theory of interpretation has to take the form of a Tarskian theory of truth. But I shall claim that, contrary to one proposal I shall discuss, the nature of the connection cannot be correctly understood unless one takes into account the empirical constraints placed on radical interpretation especially by the principle of charity. Finally, I shall argue that the alleged role of ontology in interpretation conflicts with other aspects of Davidson’s theory of radical interpretation, being ultimately inconsistent with the assumption that reference is inscrutable.

2. Semantic description and metaphysical agreement At the beginning of one of his most influential articles, Davidson says: In sharing a language, in whatever sense this is required for communication, we share a picture of the world that must, in its large features, be true. It follows that in making manifest the large features

148 of our language, we make manifest the large features of reality. One way of pursuing metaphysics is therefore to study the general structure of our language (Davidson 1977: 199). Although the details of Davidson’s approach may be somewhat elusive, the general method is clear. Suppose we wonder whether events (walkings, stubbings and the like) have a place in the world. All we have to do in order to get an answer is to give a theory of interpretation for our language. If we have to quantify over events in order to interpret what we ordinarily assert, then events have a place in the world, and we are not free, whatever we say, to give up walkings and stubbings, at least no more than we are free to give up our own language. Although Davidson is careful enough to specify that this is not “the sole true method of metaphysics”, he says nothing about what other “true methods” there are and to what extent (if any) they are to be regarded as legitimate. What he says is that no interpretation of a language can fail to involve an account of the most general aspects of reality. So, unless interpretations can be correct qua semantic and incorrect qua metaphysical—which would be very strange—the method of truth is not just one method in metaphysics among many, it is a ultimate and irrefutable method. For, if a theory of interpretation entails certain metaphysical assumptions, they cannot be incorrect unless the theory of interpretation is itself incorrect. Perhaps, some metaphysical problems cannot be addressed at all by studying the general structure of our language. (Davidson says nothing about that, but, if metaphysical problems of this sort exist, theories of interpretation can only be neutral concerning their solution). However, those metaphysical issues that can be addressed by studying the general structure of our language require no further scrutiny. As Davidson emphasizes, the idea of a close connection between semantics and metaphysics—the theory of what sentences mean and the theory of what exists—is not new. There are plenty of venerable precedents, of which Davidson mentions some en passant (Davidson 1977). What is characteristic of Davidson’s approach, however, is not merely the controversial but rather common idea that the best method for doing metaphysics is to study the general structure of ordinary language—

149 the principle “no metaphysics without semantic description”. Rather, it is the principle “no (semantic) description without (metaphysical) agreement”, namely, the idea that we cannot interpret a speaker without putting her largely in agreement with us regarding the general structure of the world—which implies that the large features of reality that emerge from a general theory of our own language must also emerge from a general theory of whatever language. The former idea seems to be grounded in the latter as follows. If successful interpretation requires ontological agreement, nobody can be interpreted as massively disagreeing with us regarding “the large features of reality”. Therefore, the best way to clarify the large features of reality is to study the general structure of our own language (which is the former idea: no metaphysics without semantic description). The latter idea is much less common than the former, but perhaps not completely new, its most striking precedent probably being the scholastic idea, developed in the Grammaticae speculativae from the 12th century onwards, that there is just one grammar, embodied in particular languages with merely accidental modifications—a grammar whose structure can be clarified by analyzing the metaphysical nature of things. (Here, however, the epistemic order of the semantic and the metaphysical is reversed: Davidson’s suggestion is doing semantic to answer metaphysical questions, not the reverse.) Although perhaps not completely unprecedented, the idea has weird consequences. One of them is that no boundary can consistently be drawn between descriptive and revisionary metaphysics. According to Strawson, “descriptive” metaphysics aims at describing the most general features of our thought about the world and is nothing other than an attempt to understand in ontological terms general logical and semantic structures, while “revisionary” metaphysics, on the contrary, aims to discern an allegedly better structure (Strawson 1959). However, if interpretation necessarily requires—by and large—metaphysical agreement, no language can be interpreted as involving a metaphysics significantly divergent from our own. Accordingly, revisionary metaphysics become non-expressible and non-interpretable, just a sort of nonsense. This sounds extremely surprising. Take van Inwagen’s metaphysics, according to which nothing has proper parts except living organisms, so that there are no chairs, no

150 buildings, no stones and so on (van Inwagen 1990). How should we interpret what Peter van Inwagen says in Material Objects? Alternatively, take Sidelle’s “Pure Stuffism”, that is, the thesis that matter, or stuff, is the only real, mind-independent entity of the world and objects are to be disallowed altogether (Sidelle 1998). How should we interpret Sidelle’s metaphysics? Perhaps, Sidelle, van Inwagen and many other “revisionary” metaphysicians are stubbornly trying to say what cannot be said.1 Perhaps it is ourselves who are stubbornly trying to interpret what cannot be interpreted. Perhaps both. The interesting question is: why exactly does Davidson says something so surprising? How should his argument be reconstructed? And is the argument strong enough to support the weird conclusion? In the fourth chapter of his book Semantics, Tense and Time, Peter Ludlow claims that a Tarskian theory of truth—and so a Davidsonian theory of interpretation—entails metaphysical consequences inasmuch as it assigns semantic values to the words of the object language by implicitly quantifying over those values. As interpreters, Ludlow says, “we will be committed to whatever objects serve as a semantic value in a correct Ttheory for natural language” (Ludlow 1999: 66). The only way an interpreter can represent the semantic knowledge of a speaker is to identify the semantic values of the speaker’s words with the entities to which the interpreter is herself committed (those entities the theory of truth quantifies over). Since in a Tarskian theory of truth “to be is to be a semantic value”, a theory of radical interpretation can only project onto the object language its own metaphysical assumptions concerning the general structure of the world. The reason why, according to Ludlow, radical interpretation requires metaphysical agreement, can be roughly summarized as follows. A theory of radical interpretation is a Tarskian theory of truth, and a Tarskian theory of truth includes a scheme of reference (Ludlow does not use this expression)—a pattern of correspondences between words and things that exist according to the interpreter (the “semantic values” of those words). The interpreter’s ontology shapes the semantic theory by shaping the 1

Another example is provided by Heller’s eliminativist version of fourdimensionalism (Heller 1990).

151 scheme of reference. If Ludlow is right, the surprising principle “no semantic description without metaphysical agreement” is a plain consequence of the idea that a theory of interpretation must take the form of a Tarskian theory of truth (that is, a plain consequence of the formal constraint placed by Davidson on radical interpretation). Let me tell something to show that this is not the case. Ludlow starts considering examples of the basic clauses of a T-theory for elementary words (atomic “nonterminal nodes”) of the object language, such as (1) and (2). (1) (∀x) Val (x, Dick) iff x = Dick

(to be read as: For all x, x is a semantic value of ‘Dick’ iff x is identical with Dick)

(2) (∀x) Val (x, red) iff x is red

(to be read as: For all x, x is a semantic value of the predicate ‘red’ iff x is red)

(1) and (2) differ only superficially from their analogues in a standard “scheme of reference” of Davidsonian variety—clauses like “‘Dick’ refers to x iff x = Dick” or “x satisfies ‘Red’ iff x is red”. One formal difference is that in (1) and (2) the quantification is explicit. Another difference is that (1) and (2) (but not their more familiar analogues) have exactly the same form (that is, “for any x, x is a semantic value of the expression e iff…”).2 Ludlow claims that, because the quantification in (1) and (2) is not vacuous, (1) commits us to the existence of Dick, and (2) to the existence of one or more objects that happen to be red. But this seems to be far from clear. Suppose I say that, for all x, x is a unicorn if and only if x is a horned horse, or something like that. Would you be allowed to conclude that, since 2

In Ludlow’s elegant formulation of Tarskian theories of truth, all reference clauses have the same general form, regardless of whether they are about individual constants or predicates. Even the formal difference between “projection rules” and “reference clauses” tends to vanish. A semantic axiom is a projection rule or a reference clause depending on whether it assigns a semantic value to a “terminal” or a “nonterminal” node (Ludlow 1999: 33-36).

152 the quantification in my sentence is not vacuous, I am committed to the existence of unicorns? I suspect not. Since everything is such that it is not a unicorn and it is not a horned horse, the sentence turns out true just because the domain of quantification is not vacuous—although it would be (vacuously) true even if the domain of quantification were vacuous. One might insist that at least (1) commits us to the existence of something, namely, of Dick. For being identical with Dick is not having certain properties (such as being a horned horse or the like), but bearing a certain relation with a particular thing—indeed being that thing. And, granted, if Dick is a particular thing, then there must be something identical with him. So, one might be tempted to conclude that a theory of truth is at least committed to the existence of everything named by a proper name in whatever clause like (1). A consequence of this conclusion is that a theory of truth has ontological bearing inasmuch as it employs genuine proper names in clauses such as (1). But the conclusion is over-hasty, for we have plenty of names that seem to denote nothing really existing—which is the reason why one can say that, for all x, the name ‘Sherlock Holmes’ denotes x iff x = Sherlock Holmes without being committed in any way to the real existence of Sherlock Holmes. It is precisely because the name ‘Sherlock Holmes’ denotes x if and only if x is identical with Sherlock Holmes—one might observe—that the name denotes nothing! Indeed, as Ludlow himself remarks, there are a number of well-known treatments of proper names according to which (1) conveys either no ontological commitment or a very banal and uninteresting one. If proper names are treated as disguised definite descriptions, for example, (1) is in no way more ontologically committal than (2). There are well-known problems with descriptive theories of proper names, but many proposals have been made despite those problems to save the core of a descriptive theory of proper names from rejection. One of them is to treat proper names as rigidified definite descriptions. Another is to treat proper names according to a bidimensional semantics (Stalnaker 1990). If one prefers to treat so-called “empty” names as denoting Meinongian entities—things that neither exist nor subsist but nevertheless there are in some sense of “there are”—(1) really commits us to something called “Dick”, but only in the most general, banal and uninteresting sense in which we can all be said to be committed to Sherlock Holmes, Pegasus,

153 or Santa Claus (Meinong 1904). Alternatively, clauses like (1) can be interpreted as governed by a free logic. In this interpretation, (1) conveys per se no ontological commitment at all. It does not say whether a word of the object language refers or not, but only what are the “conditions of reference” of that word (that is, how something must be in order to be a referent or semantic value of that word), in much the same way as, mutatis mutandis, a T-sentence does not say whether the mentioned sentence is true, but only what are its “conditions of truth”, that is, what has to happen in order for it to be true. In (Sainsbury 2005), a complete treatment of this variety is given for reference clauses like (1) according to which, whenever Dick does not exist, (1) turns out true, but any simple sentence containing the name ‘Dick’ turns out false. One might be tempted to say that, if nothing is red and nothing is identical to Dick, (1) and (2) fail to do their job—namely, to assign to the expressions ‘red’ and ‘Dick’ some semantic value. But this—it might be replied—would be to ask too much of (1) and (2). What semantic axioms such as (1) and (2) have to do is to state clearly necessary and sufficient conditions for being the semantic value of some expression. But it would be too much to ask them to take responsibility for the way the world is (particularly for the real existence of something able to satisfy the necessary and sufficient conditions they specify). It would be too much, because (1) and (2) express no commitment to the existence of red things and to the existence of something identical to Dick. They specify conditions of reference just as T-sentences specify conditions (not values) of truth. In his book, Ludlow discusses briefly Dummett’s distinction between robust and modest theories of meaning (that is, theories that explain and theories that fail to explain the epistemic abilities that underlie our semantic competence) (Dummett 1975, 1991). And he claims that, even if Tarskian theories of truth can at most work as modest theories of meaning, this is in no way a diminution of their semantic value: it is worth distinguishing between modest and robust semantic projects, and both are worth implementing. To the extent that that distinction holds, I would say the same of a parallel distinction regarding the roles of ontology in semantics. There is an ontologically modest semantic project, that merely aims to specify conditions of reference and truth for the expressions of the

154 object language and an ontologically robust semantic project, that aims to state whether the expressions of the object language satisfy the specified conditions of reference and truth or not. The idea is that, if theories of radical interpretation are as ontologically robust as they are, it is in virtue of the empirical constraints placed on them, not in virtue of their formal status as Tarskian theories of truth. If this is true, placing empirical constraints on a Tarskian theory of truth for a language is tantamount to shifting from an ontologically modest semantic project to an ontologically robust semantic project for that language. Let me develop the suggestion in the following section.

3. Ontological charity If neither (1) nor (2) are ontologically committal per se, what reasons are there to think that semantic description demands metaphysical agreement? If any reasons are left, these can only have to do with the empirical constraints placed on interpretation, rather than with the formal nature of a Tarskian theory of truth. According to the well-known “principle of charity”, the only way to interpret what the speaker says on the grounds of how she acts is to suppose that she is largely right about what there is and what there is not. In Davidson’s words: “The more sentences we conspire to accept or reject, the better we understand the rest, whether or not we agree about them” (Davidson 1974: 137).3 This means that the left-hand side of a T-sentence of an interpreter must be true, save exceptions, 3

Statements of the principle abound in Davidson’s articles. Here are a couple of examples. “If all we know is what sentences a speaker holds true, and we cannot assume that his language is our own, then we cannot take even a first step towards interpretation without knowing or assuming a great deal about the speaker’s beliefs. Since knowledge of beliefs comes only with the ability to interpret words, the only possibility at the start is to assume general agreement on beliefs. We get the first approximation to a finished theory by assigning to sentences of a speaker conditions of truth that actually obtain (in our own opinion) when the speaker holds those sentences as true” (Davidson 1974: 196). “Just as too much attributed error risks depriving a the subject of his subject matter, so too much actual error robs a person of things to be wrong about” (Davidson 1977: 200).

155 whenever the speaker’s sentence mentioned in the right-hand side is held true by the speaker. Per se, therefore, (1) and (2) do not commit us— respectively—to the existence of Dick and to the existence of red things, but they do this to some extent inasmuch as the interpreted speaker holds as true such sentences as “There are red things” and “There is something identical to Dick”. For, if the interpreted speaker is by and large right about the world, there is some presumption that ‘Dick’ is not an empty name and that ‘red’ does not have an empty extension. Granted, the interpreted speaker can happen to be wrong here and there on what there is, but she cannot be massively wrong. If charity holds, then, T-theory clauses of the form “For all x, x is a semantic value of the name ‘N’ iff x = N” or “For all x, x is a semantic value of the predicate ‘P’ iff x is P” generally commit us to the existence of N, or to the existence of one or more P, whenever the speaker says “N exists”, or “There are one or more P”. The principle of charity plays a very important role in the theory of radical interpretation, being aimed—in Davidson’s words—“to solve the interdependence of belief and meaning by holding belief constant as far as possible while solving for meaning”. The caveat “no (semantic) description without metaphysical agreement” is just an aspect of the principle of charity. Another aspect is the caveat “no logical description without logical uniformity”, the idea that the metalanguage can discern a logic in the object language only insofar as it recognizes in the object language its own logic (the latter caveat suppresses logical difference, just as the former suppresses metaphysical difference). A further aspect of the principle is its role in Davidson’s famous argument against the very idea of a conceptual scheme. I am not interested here in going deep into the position of the principle of charity in the complicated overall geography of Davidson’s theory of radical interpretation, including such controversial loci as semantic holism, the interdependence of meaning and belief, the criticism of the “third dogma”, the coherence theory of truth and so on. I content myself to emphasize that the principle “no semantic description without metaphysical agreement” might be seen as springing from two assumptions: the principle of charity (“no interpretation without massive agreement”) and the idea that the theory of radical interpretation has to take the form of a Tarskian theory of truth, a theory containing reference and satisfaction clauses such as (1) and (2). For in this case, as we have

156 just seen, the interpreter is necessarily committed to the existence of the greater part of the entities to which—according to the interpreter herself— the speaker’s words refer. If this is true, the weird principle “no semantic description without metaphysical agreement” can be denied by rejecting the principle of charity—perhaps indirectly, by denying some of the assumptions the principle is grounded upon (for example semantic holism). These are, as Dummett often says, “deep waters”, certainly too deep to be explored in this paper. What I intend to do is simply to say a few words on whether the weird principle “no (semantic) description without metaphysical agreement” can be denied even if one grants both the principle of charity and the idea that the theory of interpretation has to take the form of a Tarskian theory of truth. Here, as far as I can see, there are essentially two worries. First, one can believe that a Tarskian theory of truth (that is, a theory both formally correct and materially adequate in Tarski’s sense) can dispense with satisfaction-clauses like (1) and (2)—a doubt normally coupled with the idea that Tarski’s theory is ultimately a disquotational theory of truth (perhaps modulo a relation of translation, when the metalanguage and the object language differ). Second, one can doubt that clauses like (1) and (2) really succeed in mapping the words of the object language onto the entities of the real world: if all that clauses like (1) and (2) can do is to couple the words of the speaker with the words of the interpreter, there cannot be much reason to believe that (1) and (2) commit us to any entity. What is certain is that Davidson’s position regarding the mentioned worries is far from clear. On the one hand, Davidson has always defended the indispensability of satisfaction clauses in the radical theory of interpretation. On the other hand, he has proposed to “give up” reference and so satisfaction. Let me say why I have difficulties in understanding Davidson’s position here. There are in principle a limited number of ways to dispense with satisfaction clauses in a theory of truth. Whenever the object language has a finite number of elementary names and a finite number of elementary predicates, it may have a finite number of atomic sentences too. In such a case, a finite list of T-sentences can be given in the metalanguage, one for each atomic sentence of the object language. Any quantified sentence of

157 the object language can be treated in the Tractatus way (Wittgenstein 1921), as a finite conjunction or a finite disjunction of atomic sentences and the relative T-sentence can be recursively constructed on the basis of the T-sentences giving the truth conditions of the atomic sentences occurring in the conjunction, or disjunction. A theory of truth of this variety does not need satisfaction clauses like 1) and 2). Indeed it is simply a disquotational theory of truth (or a disquotational theory of truth modulo a relation of translation). Davidson has objected that this strategy for dispensing with satisfaction clauses in a theory of truth is unsuitable whenever the object language has the normal expressive powers of a natural language. For ordinary language has a finite number of elementary names and elementary predicates, but not a finite number of atomic sentences, due to the fact that atomic sentences can be built from complex names and complex predicates: Predicates come in any degree of complexity, since they can be built up from connectives and variables; and constant singular terms can be complex (Davidson 1977: 217). Two general strategies to dispense with satisfaction clauses in a theory of truth are nevertheless available, at least in principle, even if the object language can produce an infinite number of atomic sentences. First, the theory of truth can embrace a substitutional interpretation of quantifiers (a well-known example is given in (Kripke 1976)). Second, the theory can take the form—at least in an infinitary metalanguage—of an infinite disjunction like: “x is true iff (x = ‘snow is white’ and snow is white) or (x = ‘grass is green’ and grass is green)…. and so on ad infinitum, with a disjunct for each of the infinite sentences of the object language. As far as I know, Davidson has never analysed in detail the second strategy, mentioned en passant in (Etchemendy 1988), but he argues that if the object language has normal expressive powers, a theory of truth à la Kripke is materially inadequate, that is, unable to yield all the T-sentences required by Convention T.

158 4. Why we cannot live with ontology and without reference As I have just stressed, Davidson has always firmly defended the indispensability of satisfaction clauses like (1) and (2) in a theory of interpretation: a theory of radical interpretation for a language has to take the form of a Tarskian theory of truth for that language, and such a theory must define truth in terms of denotation of names, satisfaction of predicates and fulfilment of functors. In a nutshell: in terms of reference. From this point of view, a Tarskian theory is nothing more than the definition of truth as a particular kind of reference: roughly, the satisfaction of open sentences with zero free variables. On the other hand, the appeal to the notion of satisfaction in a definition of truth is not, according to Davidson, a free option. It is the only way to account for how the truth or falsity of a quantified sentence depends on the words occurring in it and on the way these words are put together (what may be called the logical structure of the sentence). Among the most important philosophical consequences of Davidson’s theory of meaning, on the other hand, there is a certain dissolution of the notion of reference, its apparent elimination from the theory of meaning and, perhaps, its elimination tout court: the idea that we should, to use Davidson’s words, “give up” reference, “get rid of” that notion or “live without” it. Davidson emphasises the difference between recursive and explicit definitions of reference. A Tarskian theory of truth, he says, neither explains nor analyses the concept of reference, since it embraces nothing more than a recursive definition—“a scheme”—of reference. And he insists that a recursive definition of reference is a strictly theoretical posit, necessary in order to implement a theory of interpretation but not susceptible to direct empirical control. Davidson makes a distinction between “explanation within the theory of meaning” and “explanation of the theory”. Within the semantic theory, all that we do is explain the truth conditions of sentences by assigning a reference to their elementary semantic constituents and specifying a set of recursive clauses (projection rules) apt to generate the truth conditions of any sentence, given the references of the singular terms, predicates and functors occurring in it. But there is no way of giving a direct empirical reduction or explanation of the notion of reference: the references of the words of a language are

159 whatever things can determine empirically correct truth conditions for the sentences of that language. They are theoretical posits, “as are the notions of singular term, predicate, sentential connective, and the rest” (Davidson 1990: 300). As Quine and others have convincingly shown via the method of permutations (Quine 1969), however, the crucial point is that it is possible to construct systematic examples of incompatible schemes of reference apt to determine exactly the same truth conditions for all the sentences of a language: according to a scheme, ‘Dick’ refers to Dick, according to another it refers (say) to the shadow of Dick. Quine’s suggestion was to treat as true both the assignments of reference, but only relative to two different manuals of translation. According to Davidson, however, we cannot say, even relatively or arbitrarily, what a word refers to. We cannot even say that a word refers to a certain object because all the words we can use in order to try to say this have in their turn a totally indeterminate or inscrutable reference. So, a sentence such as “the name ‘Dick’ refers to Dick” fails to determine the reference of Dick, even arbitrarily or relatively to a homophonic manual translating ‘Dick’ as ‘Dick’. And the reason is that the English words used in the sentence are no less indeterminate in reference than the word mentioned in it. Davidson’s conclusion is: We don’t need the concept of reference; neither do we need reference itself, whatever that may be. For if there is one way of assigning entities to expressions (a way of characterizing ‘satisfaction’) that yields acceptable results with respect to the truth conditions of sentences, there will be endless other ways that do as well. There is no reason, then, to call any of these semantic relations ‘reference’ or ‘satisfaction’ (Davidson 1977: 224). This does not mean, obviously, that we can build a theory of truth (interpretation) for a language, giving no scheme of reference for its words (no Tarskian theory of truth, hence no Davidsonian theory of meaning, could ever do that). And a scheme of reference can be nothing other than a list of sentences of the form “a such and such expression refers to...”. However, Davidson explains, all that such sentences determine is “the way

160 we answer questions about reference, not reference itself” (Davidson 1979: 239). A scheme of reference, in other words, does not settle matters of reference, it settles only “how I answer all sorts of questions about what the speaker means or refers to by a word or sentence” (Davidson 1979: 238). It is not so easy to understand exactly what Davidson might mean by these words. If what he wants to say is that a sentence like “‘rabbit’ refers to rabbits” may be used to answer one who asks what ‘rabbit’ refers to but not to say what ‘rabbit’ refers to, I would find it extremely difficult— perhaps impossible—to grasp how that could ever happen. To be sure, if the sentence “‘rabbit’ refers to rabbits” does not tell what ‘rabbit’ refers to, neither can it be used to answer one who asks what does ‘rabbit’ refer to. Perhaps, however, it might be used to answer one who utters the question “what ‘rabbit’ refers to?”—provided that person, uttering the question, does not succeed in asking what ‘rabbit’ refers to. I think it is just this that Davidson has in mind: we can neither say what ‘rabbit’ refers to by uttering assertively “‘rabbit’ refers to rabbits” nor can we ask it by uttering the corresponding question. The predicate “refers to” continues to exist, however, together with a number of “schemes of reference”, sequences of sentences of the form “such and such an expression refers to...”. If these sentences fail to say what the words mentioned in them refer to, what might they ever say, supposing they say anything at all? Davidson’s answer is that these sentences do not determine what objects the words of a speaker refer to but merely “what words we can use in our own language to interpret [her] words” (Davidson 1979: 238-239). One might get the impression that nothing substantial to distinguish the relation of reference from that of translation is left and an “ontologically mixed” relation between words and objects gives way to an “ontologically homogenous” relation between words and words. So construed, Davidson’s idea is that, at the pre-sentential level, the theory of interpretation has to be confined to establishing syntactic equations between words, and it is equations like these that sentences of the form “... refers to...” express in a spurious way. The semantic meaning of these formal equations is merely reflected: interacting with a system of “projection rules”, they yield T-sentences having direct semantic meaning.

161 In this perspective, what we improperly name “the reference predicate” should be more correctly described, in Carnap’s terminology, as a spurious counterpart in the material mode of speech of the predicate “is translatable as”. The question is: can the predicate “refers to” be really merely this? The idea is that, when we utter “‘rabbit’ refers to rabbits”, all that we say is that ‘rabbit’ is interpretable or translatable (from English into English) as ‘rabbit’. Should this idea hold, Davidson’s thesis that the sentence “‘rabbit’ refers to rabbits” fails to say what ‘rabbit’ refers to would simply amount to the thesis that the sentence does not tell how ‘rabbit’ should be interpreted or translated. But Davidson does not mean that. He means that the sentence “‘rabbit’ refers to rabbits” fails to determine the reference of ‘rabbit’. How could this be said, however, if all that we have is a disguised predicate of translation? How can we express, with no predicate of reference, the idea that reference is inexpressible? If we have no predicate of reference, how can we say that we have none? And if we cannot say that, how can we know it? If we can neither say nor ask what a word refers to (but merely how we should interpret or translate it), how can we say that it could never be said? What is clear is that Davidson does not stop speaking of reference. For example, he says reference is “relative to languages” (Davidson 1979: 240).4 Once again, however, he specifies that what the relativization determines is only “the way we answer questions about reference, not reference itself” (Davidson 1979: 239). To say that reference is relative to languages is therefore a mere façon de dire. Just a few years earlier, Davidson emphasised that the theory of meaning, although it does without reference, cannot do without ontology, since it relates each singular term to some object or other, and it tells what entities satisfy each predicate. Doing without reference is not at all to embrace a policy of doing without semantics or ontology (Davidson 1977: 223). 4

Davidson never deviated from his “distal” theory of reference (see for example Davidson 2005, ch. III).

162 Davidson maintains that the ontology of a language is “revealed by a theory of truth” (for that language). And he adds that his holistic approach differs from Quine’s in virtue of the direct connection it establishes between issues of truth, issues of ontology and issues of logical form, i.e. in virtue of the idea that a Tarskian theory of truth may work at the same time as a key to the ontology of a language and as a test of logical form. The sense in which a theory of truth can bring to the surface or reveal an ontology is roughly this: if accepting certain entities or kinds of entity (for example individual events or instants of time) into the domain of the variables of a language is necessary in order to assign to the sentences of that language the right truth conditions, then those entities or kinds of entity belong to the ontology of that language. However, in attacking the relativity of reference, Davidson maintains that we cannot connect terms to objects unless our own words (unlike those we try to interpret) have a fully determinate reference. And this is something Davidson—following Quine—denies. All this makes it difficult to see how we can do without reference but go on with ontology. We cannot, to be sure, maintain that nothing exists on the mere ground that there does not exist anything we can refer to. If there does not exist anything we can refer to in one language or another, however, it is no longer so certain there is still something worthy to be called “ontology” of a language. Remember that the question what objects a particular sentence is about, like the questions what objects a term refers to, or what objects a predicate is true of, has no answer (Davidson 1984: xix). In rougher words, if a permutation of the universe interchanging George Bush with a certain chair yields empirically acceptable results as regards the truth conditions of all English sentences, then the sentence “George Bush really exists” can be correctly interpreted as saying that a certain chair really exists. Davidson emphasises that giving up reference is not, at least in his intentions, an idealist or anti-realist move. According to him, all that giving up reference forces us to do is to conceive the basic elements of reality as no longer the references of words but rather as the causes of assertoric utterances. This brings out a substantial difference between

163 Davidson’s method of radical interpretation and Quine’s method of radical translation. Davidson explains: we must, in the plainest and methodologically most basic cases, take the objects of a belief to be the causes of that belief. And what we, as interpreters, must take them to be is what they in fact are. Communication begins where causes converge: your utterance means what mine does if belief in its truth is systematically caused by the same events and objects (Davidson 1984: 318). With plenty of sentences, however, this idea seems to yield unpalatable consequences. Assent to the sentence “the rabbit is running” is not regularly caused by rabbits but by running rabbits. And the assent to the indigenous equivalent of “the book is on the table” is caused neither by books nor by tables, but by books put on tables, or by tables on which books are put—or perhaps by both things together. If I utter “the rabbit is running”, however, my utterance is not about a running rabbit, it is about a rabbit (if it is about anything at all) and says about it that it is running (should there be a non-running rabbit, the sentence would be false because the rabbit the sentence is about is not running, not because there is nothing the sentence is about). At least in such cases, therefore, it is difficult to “take the objects of a belief to be the causes of that belief”, since what causes our assent to a sentence such as “that rabbit is running” cannot be the object or the event the sentence is interpreted as being about but has to embrace as well everything the sentence says about it. In an early article, Davidson observed that the failure of correspondence theories of truth based on the notion of fact traces back to a common source: the desire to include in the entity to which a true sentence corresponds not only the objects the sentence is “about” (another idea full of trouble) but also whatever the sentence says about them. One well-explored consequence is that it becomes difficult to describe the fact that verifies a sentence except by using that sentence itself (Davidson 1969: 49).

164 Unfortunately, what Davidson says about the entity to which a true sentence corresponds can be said without modification about the entity causing us to assent to a sentence. Even though Davidson does not seem aware of it, the idea that the objects and events of the world are not the references of our terms, but rather what causes us to assent to sentences, ends surprisingly by treading in the steps of an old and discredited theory of truth (at least in Davidson’s opinion): the idea that truth is nothing more than a form of correspondence between sentences (propositions, assertions, or something similar) and reality. In his posthumous Truth and Predication (Davidson 2005: chs. IVVII) Davidson still emphasizes the indispensable role of reference in a theory of interpretation, and claims that Tarski’s definition of satisfaction resolves the metaphysical problem of predication, a riddle philosophers have struggled with since the very origins of their discipline. As I have just tried to show, however, the only relation of “satisfaction” that the inscrutability of reference allows us to have seems to be a syntactic relation of correspondence between words. Hence, such a relation can have nothing to do with metaphysics or ontology.

5. Conclusion I have argued for the following theses. (A) The principle “No interpretation without metaphysical agreement” does not follow from the formal constraints placed by Davidson on radical interpretation—that is to say, from the idea that a theory of radical interpretation has to take the form of a Tarskian theory of truth. Indeed, the reference clauses of a Tarskian theory of truth state conditions of reference for the object language words, just as the T-sentences of a Tarskian theory state conditions of truth for the object language sentences. Therefore, no reference clause for an object language word can on its own commit the interpreter to the existence of something the word refers to, just as no Tsentence for an object language sentence can on its own commit the interpreter to the truth of that sentence.

165 (B) The principle “No interpretation without metaphysical agreement” hinges partly upon the empirical constraints placed on radical interpretation, and especially upon the principle of charity. If charity holds, reference clauses of the form “For all x, x is a semantic value of the name ‘N’ iff x = N” generally commit us to the existence of N whenever the speaker says “N exists”. For, if the interpreted speaker is by and large right about the world (which is the case, if charity holds), there must be some presumption that ‘N’ is not an empty name. (C) One can thus deny the principle “No interpretation without metaphysical agreement” by denying the principle of charity. But, even if the principle of charity holds, there can still be doubts that interpretation requires metaphysical agreement. First, one can believe that a Tarskian theory of truth can dispense with reference clauses such as (1) and (2). Second, one can doubt that clauses like (1) and (2) really succeed in mapping the words of the object language onto the entities of the real world: if all that such clauses can do is to couple the speaker’s words with the interpreter’s words, there cannot be much reason to believe that (1) and (2) commit us to some entity. (D) Davidson’s position concerning the above doubts is extremely difficult to understand. On the one hand, he firmly defends the indispensability of reference clauses in a theory of truth (interpretation). On the other hand, he claims that we should “live without” reference. I have argued that we cannot live with ontology but without reference. And I have claimed that the only way we can have reference clauses without reference is to interpret those clauses as stating syntactic relations between words instead of as stating semantic relations between words and the things they are about. But this would deprive the theory of interpretation of any ontological bearing.

References Davidson D. (1969), “True to the Facts”, in (Davidson 1984), 37-54. Davidson D. (1974), “On the Very Idea of a Conceptual Scheme”, in (Davidson 1984), 183-98.

166 Davidson D. (1977), “Reality without Reference”, in (Davidson 1984), 215-26. Davidson D. (1977a), “The Method of Truth in Metaphysics”, in (Davidson 1984), 199-214. Davidson D. (1979), “The Inscrutability of Reference”, in (Davidson 1984), 227-241. Davidson D. (1983), “A Coherence Theory of Truth and Knowledge”, reprinted in: E. Lepore (ed.), Truth and Interpretation. Perspectives on the Philosophy of Donald Davidson, Blackwell, Oxford 1985, 30719. Davidson D. (1984), Inquiries into Truth and Interpretation, Clarendon Press, Oxford. Davidson D. (1990), “The Structure and Content of Truth”, The Journal of Philosophy LXXXVII, 6, 279-328. Davidson D. (2005), Truth and Predication, Harvard University Press, Cambridge, Mass. Dummett M. (1975), “What is a Theory of Meaning?”, in: S. Guttenplan, (ed.), Mind and Language, Oxford University Press, Oxford. Dummett M. (1991), The Logical Basis of Metaphysics, Harvard University Press, Cambridge, Mass. Etchemendy (1988), “Tarski on Truth and Logical Consequence”, The Journal of Symbolic Logic 53, 51-78. Heller M. (1990), The Ontology of Physical Objects: Four-Dimensional Hunks of Matter, Cambridge University Press, Cambridge, Mass. Kripke S. (1976), Is there a Problem about Substitutional Quantification?, in: G. Evans and J. McDowell (eds.), Truth and Meaning, Clarendon Press, Oxford. Ludlow P. (1999), Semantics, Tense and Time, MIT Press, Cambridge, Mass. Meinong A. (1904), Über Gegenstandstheorie, in: A. Meinong (ed.), Untersuchungen zur Gegenstandstheorie und Psychologie, Leipzig, Barth, 1-50. Quine W.V.O. (1969), Ontological Relativity and Other Essays, Columbia University Press, New York. Sainsbury M. (2005), Reference without Referents, Oxford University Press, Oxford.

167 Sidelle A. (1998), “A Sweater Unraveled: Following One Thread of Thought for Avoiding Coincident Entities”, Nous 32, 423-448. Stalnaker R. (1990), “Narrow Content”, in: C. Anderson and J. Owens (eds.), Propositional Attitudes: The Role of Content in Logic, Language and Mind, CSLI, Stanford, 131-146. Strawson P. F. (1959), Individuals, Methuen, London. van Inwagen P. (1990), Material Beings, Cornell University Press, Ithaca. Wittgenstein L. (1921), Tractatus Logico-Philosophicus, Engl. transl. by D.F. Pears and B.F. McGuinness, Routledge and Kegan Paul, London 1961.

LANGUAGE AND CONCEPTUAL SCHEMES Michele MARSONET (University of Genoa, Italy)

1. Introductory Rremarks A conceptual scheme is, according to some dictionaries of philosophy, “a set of concepts and propositions that provide a framework for describing and explaining items of some subject-matter along with criteria for recognizing which phenomena are to be considered deviant and in need of explanation” (Brown 1995: 146-147) or the general system of concepts with which we organize our thoughts and perceptions. The outstanding elements of our everyday conceptual scheme include spatial and temporal relations, other persons, meaning-bearing utterances of others, and so on. To see the world as containing such things is to share this much of our conceptual scheme (Blackburn 1996: 72-73). It follows from the previous general definitions that, when dealing with conceptual schemes, philosophers take into account the beliefs and assumptions formulated, for example, in science and morality. The comprehensive outlooks on the world generated by some community— when taken together—form an inclusive theory in terms of which the members of that community explain and interpret both their empirical and moral experience. The limits imposed on the term “community”, on the other hand, are determined by philosophers themselves. “Community” may mean in this context a particular society or all members of humankind. The key point of the contemporary debate on conceptual schemes is, however, the following. Given the fact that thought (i.e. the manipulation of concepts) is not possible without the existence of language, conceptual schemes are most of the times (although not always) identified with languages or, even better, with sets of intertranslatable languages. If this is true, learning a language means to acquire the conceptual scheme it embodies and, as a matter of fact, according to this view a conceptual schemes is a language. Moreover, one can ask whether there is only one conceptual scheme or many; if a plurality of conceptual schemes is admitted the problem of

170 relativism arises. To sum up, a conceptual scheme is a “frame of reference”, that is to say the view-point, or set of presuppositions or of evaluative criteria within which a person’s perception and thinking always occur, and which constrains in a selective way the course and outcome of these human activities. It is well known that, according to Donald Davidson, men would be unable to interpret speech from a different conceptual scheme as even meaningful. He claims that, since translation proceeds according to the “principle of charity”, and since it must be possible for an omniscient translator to make sense of what we say and of how we behave, we can be assured that most of the beliefs formed within the commonsense conceptual framework are true. Davidson thus challenges the scheme-content dualism, and his attack to the scheme-content distinction is supported by a set of arguments purporting to reject, first of all, the thesis that totally different conceptual schemes can actually exist. To put things in a sketchy manner, he equates having a conceptual scheme with having a language, so that we face the following elements: (1) language as the organizing force; (2) what is organized, referred to as “experience”, “the stream of sensory experience”, “physical evidence”; and finally (3) the failure of intertranslatability. It follows that: It is essential to this idea that there be something neutral and common that lies outside all schemes [...] The idea is then that something is a language, and associated with a conceptual scheme, whether we can translate it or not, if it stands in a certain relation (predicting, organizing, facing, or fitting) with experience (nature, reality, sensory promptings) (Davidson 1974: 190-191). If this is the situation—Davidson goes on—then we could say that conceptual schemes that are different in a radical way from each other correspond to languages that are not intertranslatable. How can we, however, make sense of a total failure of intertranslatability among languages? For sure “we could not be in a position to judge that others had concepts or beliefs radically different from our own” (Davidson 1974: 197). Davidson’s conclusion is that if one gives up the dualism of scheme

171 and world, he will not give up the world, but will instead be able to “reestablish unmediated touch with the familiar objects whose antics make our sentences and opinions true” (Davidson 1974: 198). On his part Richard Rorty fully endorses Davidson’s stance. Starting from Quine’s rejection of the notion of “meaning” as anything transcending what is contextually defined in predicting the behaviour of other people, Rorty deems it impossible to distinguish an untranslatable language from no language: Once we imagine different ways of carving up the world, nothing could stop us from attributing “untranslatable languages” to anything that emits a variety of signals” (Rorty 1994: 6). For example, we may imagine aliens who endorse a totally different view of the world, so that their language would—in principle—turn out to be untranslatable into our own. This means that they would carve up the world according to a completely different conceptual scheme. However Rorty thinks that: For all we know, our contemporary world is filled with unrecognizable persons. Why should we ignore the possibility that the trees and the bats and the butterflies all have their various untranslatable languages in which they are busily expressing their beliefs and desires to one another? [...] So I think that to rule the butterflies out is to rule out the Galactics and the Neanderthals, and that to allow extrapolation to the latter is to allow for the possibility that the very same beliefs and desires which our Galactic descendants will hold are being held even now by the butterflies (Rorty 1994: 9-10). It follows that any language must, to count as a language, be translatable into our own, and that—quite surprisingly—the large majority of our present beliefs must be true. In other words, “the world” will just be the stars, the people, the tables, and the grass—all those things which nobody except the occasional “scientific realist” philosopher thinks might not exist. The fact that the vast

172 majority of our beliefs must be true will, on this view, guarantee the existence of the vast majority of the things we now think we are talking about (Rorty 1994: 14). By endorsing this line of thought, we no longer need the notion of “the world” conceived of as an “independent reality”, a notion which is endorsed by those thinkers who claim that different conceptual schemes carve up the world differently. Davidson’s and Rorty’s solution is radical, but we are bound to ask at this point what the expressions “reality” and “world” mean for them. Let us assume that they can be identified with the world of common sense which is formed by the familiar objects whose antics—as Davidson says— make our sentences and opinions true or false. These familiar objects are tables, chairs, houses, stars, etc., just as we perceive them in our daily life. One is not entitled to ignore, however, that the current discussion on the problem of scientific realism arise because there is a strong asymmetry between the commonsense view of the world and the scientific one. For instance, the table that we see with our eyes is not the same table that we “see” through the mediation of scientific instruments, and this fact is not trivial. It is rather easy to reach a high level of intersubjective agreement among the individuals present in a room about the colour, size and weight of a table, and it can also be granted that we form our beliefs in this regard by triangulating—in a Davidsonian sense—with our interlocutors and the surrounding environment. Such an agreement, however, becomes problematic when we try to reconcile this vision of the world with what present science tells us about it. So being in touch with such familiar objects as tables, chairs and stars “most of the time”—as Rorty specifies—has a fundamental bearing only on the ontology of common sense, since our actual science shows that quite a different representation of reality can actually be provided. Or, even better, it shows that those objects might not exist as human beings perceive them. Naturally, one can always resort to an objection of the following kind: why should we deem the table viewed as a collection of subatomic particles more important than the table that our eyes see in daily life? After all, we can conduct our life well enough even ignoring what science claims

173 (just like men did for many thousand years). This, however, looks like a serious underevaluation of the scientific enterprise. Davidson, as we said before, associates conceptual schemes with languages, and then adopts linguistic intertranslatability as the identity criterion for conceptual schemes themselves. Subsequently we are told that, in order to call something “a language,” say L0, we must be ready to accept the idea that the statements of L0 can be translated into those of our own language (let us call it L1). It easily follows from this line of reasoning that, if this cannot be done, L0 is not a language at all. According to Davidson, we must conclude [...] that the attempt to give a solid meaning to the idea of conceptual relativism, and hence to the idea of a conceptual scheme, fares no better when based on a partial failure of translation than when based on total failure. Given the underlying methodology of interpretation, we could not be in a position to judge that others had concepts or beliefs radically different from our own (Davidson 1974: 197). One may point out, however, that linguistic intertranslatability cannot be such an absolute criterion, because in certain circumstances we are able to realize that some sort of language is used, even though we cannot translate it into our own language. Larry Laudan has noted in this regard that there is no reason to assume the presence of different world-views only when there are no criteria of intertranslatability among them. He claims, in fact, that: Only with the so-called linguistic turn have philosophers supposed that conceptual schemehood is to be understood in terms of nontranslatability. Aristotle’s cosmos and Einstein’s universe represent very different world-views. With Davidson, I believe that each can be made intelligible to adherents of the other. But only someone as wedded to the translation thesis as Davidson is would imagine that the latter fact (viz., intertranslatability) constitutes grounds for denying that they represent different conceptual schemes (Laudan 1996: 13).

174 The absolute primacy that Davidson places on translatability should thus be rejected, hence Laudan’s proposal to identify conceptual schemes on ontological, axiological and methodological—and not exclusively linguistic—criteria. The fact is that Davidson resorts to a sort of “pansemanticism” which sees linguistic behaviour as the only behaviour that really counts, while Laudan’s approach is more articulated. The above-mentioned pansemanticism endorsed by Davidson clearly transpires when he tells us that: if all we know is what sentences a speaker holds true, and we cannot assume that his language is our own, then we cannot take even a first step towards interpretation without knowing or assuming a great deal about the speaker’s beliefs. Since knowledge of beliefs comes only with the ability to interpret words, the only possibility at the start is to assume general agreement on beliefs (Davidson 1974: 196). Here we have a good reason for claiming that Davidson is far less “postanalytic” than Rorty depicts him to be. The overemphasis placed upon linguistic behaviour is, in fact, a typical trait of the analytic school, which tends to forget the fact that, after all, man came first and language later. Language is a relatively recent factor in the history of our evolution, as science shows us, and many parts of our behaviour are guided by nonlinguistic criteria. We can avoid the aforementioned analytic overemphasis only by recognizing that language is not the whole of reality, but a social product devised essentially for practical purposes.

2. Ajdukiewicz on conceptual apparatuses Strangely enough, when one gets involved in the contemporary debate on conceptual schemes he very rarely finds mention of Kazimierz Ajdukiewicz.1 However, the theses of the Polish philosopher in this regard 1

With the exception, of course, of such texts dealing specifically with Polish philosophy of the past century such as (Skolimowski 1967) and (Wolenski 1989).

175 are quite important, and we deem his contribution to be no less original— and no less controversial—than those of Quine, Davidson and Rorty. Let us note from the onset that even for Ajdukiewicz conceptual schemes are languages. Furthermore, we would like to point out that Ajdukiewicz wrote his main essays on this topic in the 1930’s, i.e. half a century before the debate on conceptual schemes started following the publication of Davidson’s paper in 1974. This confirms, we believe, the originality of his approach. Wolenski rightly notes that Ajdukiewicz’s approach is linguistic all the way down, and that he devised no evolutionary structure in language. But this, of course, is no reason for denying the importance of his theses for the contemporary debate: Ajdukiewicz’s limits, after all, are the limits of analytic philosophy itself. Jan Wolenski writes in this regard that: Ajdukiewicz treated cognitive processes as inseparably connected with language: we always think in some language, and our statements are meanings which are attributes of sentences in some language L. Hence cognition, or, to put it more rigorously, cognition as a product, can be identified with the meaning of sentences. This is the essential methodological intention of Ajdukiewicz’s semantic epistemology (Wolenski 1989: 199). In the Polish philosopher’s works the expression “conceptual apparatus” replaces “conceptual scheme”. Ajdukiewicz’s notion of conceptual apparatus is strictly tied to “close” and “connected” languages, and thus has a more technical connotation than what is meant today by the expression “conceptual scheme”. Wolenski tells us that the class of meanings of a closed and connected language was termed by Ajdukiewicz the conceptual apparatus of that language. It follows from the appropriate definitions that two conceptual apparatuses are either identical or have no element in common. And an important consequence is that: If two conceptual apparatuses have at least one element in common, then they are identical. Thus conceptual apparatuses never overlap. Ajdukiewicz held that every meaning belongs to some conceptual

176 apparatus. Hence open languages are mixtures of various conceptual apparatuses (Wolenski 1989: 204-205). Let us note, at this point, that Ajdukiewicz’s theses are subject to the same criticisms we previously addressed to Davidson and Rorty. Laudan’s remark that only a full endorsement of the linguistic turn’s main tenets may explain why so many philosophers insist on equating conceptual schemehood to languagehood apply to Ajdukiewicz as well. Only now, following the rise of post-analytic philosophy and the rediscovery of pragmatism, the basic tenet according to which linguistic behaviour is the sole behaviour that really matters has openly been challenged.2 We all know, of course, that Ajdukiewicz’s conception of language is an autonomous one, since he took language to be a product which is independent of action. Wolenski reminds us that: The problem is clarified immediately when we consider the fact that Ajdukiewicz was not interested in the origins of a language, but in language as a product. The thesis on the autonomy of language acquires meaning when we bear in mind the difference between actions and their products (taken over by Ajdukiewicz from Twardowski). An objective assignment of meanings to expressions is possible only when language is treated as a product” (Wolenski 1989: 205). Yet, this fact does not distinguish Ajdukiewicz’s ideas from the main stream of the linguistic philosophy of the past century. What makes Ajdukiewicz’s thought so appealing is the fact that he cleverly anticipated, in the 1930’s, many theses that are commonly discussed today. So we find out that his “radical conventionalism”, despite its several shortcomings, has many precious insights too, because in a famous paper dating back to 1934 he wrote: Of all the judgements which we accept and which accordingly constitute our entire world-picture, none is unambiguously determined 2

See, for example, (Devitt 1991) and (Rescher 1994).

177 by experimental data; every one of them depends on the conceptual apparatus we choose to use in representing experiential data. We can choose, however, one or another conceptual apparatus which will affect our whole world picture (Ajdukiewicz 1934: 67). Subsequently Quine became famous for saying more or less the same thing, while Ajdukiewicz’s contributions are still ignored by most Western philosophers. Quine claims that there are many implicit background assumptions which make all the difference to how we interpret our experiences, and how we make our final evaluation of statements. This means that we cannot simply get meaning from experience, since there are no “neutral” observations available to men. And it is precisely because our conceptual judgements meet experience as a body that we must allow for possible revisions at any place within that body, so that “no statement is immune to revision” (Quine 1954: 43). And, if this is right, we must even allow for the possibility of changes in our verdicts on what is experienced itself. Now compare Quine’s statements with the following by Ajdukiewicz: No articulated judgement is absolutely forced on us by the data of experience. Experiential data do indeed force us to accept certain judgements if also we are based on a particular conceptual apparatus. However, if we change this conceptual apparatus, we are freed of the necessity of accepting these judgements despite the presence of the same experiential data (Ajdukiewicz 1934: 72). It is clear that what Quine defines as our “conceptual sovereignty” plays a key role even in this context, although the words used by the two authors are not the same. In any event, the striking similarity between the two philosophers is clearly detectable, once again, in the following statements by Ajdukiewicz (written in 1935): Even the epistemologist cannot speak without a language, cannot think without a conceptual apparatus. He will thus make his decision as to truth in a way which corresponds to his world-perspective (Ajdukiewicz 1935: 117).

178 Not only that: even logic is, according to the Polish philosopher, “relative to” a particular conceptual apparatus, and a change in the conceptual apparatus means a change in logic, too (Wolenski 1989: 208). On his part, Quine claims that revision even of the logical law of the excluded middle has been proposed as a means of simplifying quantum mechanics; and what difference is there in principle between such a shift and the shift whereby Kepler superseded Ptolemy, or Einstein Newton, or Darwin Aristotle? (Quine 1954: 43).

3. Final remarks In the final analysis we deem it necessary to point out that conceptual schemes are neither born out of nothing nor established on aprioristic bases. Their aim is to provide us with means for thinking about—and for speaking of—a reality which includes ourselves. We can add four partial definitions of “conceptual schemes” to the ones provided in the opening section. In a first sense they are: (A) sets of socially codified beliefs, that is to say belief-structures that are warranted by social use. In a second sense conceptual schemes are (B) sets of logically interconnected beliefs, i.e. structures in which our conceptual sovereignty above nature plays an essential role. In a third sense conceptual schemes are (C) world-views, i.e. interpretations of the world. In a fourth sense they are (D) operational perspectives on the world, i.e. means by which human beings interact with the surrounding environment. In our view they are primarily tied to the dimension of human action, and must be seen as elements of the agent/environment interaction. As a matter of fact data concerning non-verbal action and behaviour can lead us to ascribe beliefs in quite a plausible way. No doubt translatability helps a great deal, but certainly it is not an a priori condition for ascribing beliefs. These remarks pave the way towards understanding what conceptual schemes really are. They are a sort of practical metaphor which is supposed to convey the outcome of our categorization of reality.

179 One should always be careful not to ascribe to them any metaphysical or self-subsistent feature: in other words, we must produce no reification of conceptual schemes, because their real nature is practical and functional. In order to understand what a conceptual scheme is we must not have recourse to abstract idealizations, because the comprehension of its nature can only be achieved by looking at how it works. Dewey’s idea that our explanatory mechanisms are themselves the products of inquiry, in turn, opens the door to another key notion: “conceptual innovation”. If we look at the history of science, for example, it is easily understandable that we form our conception of the sun in quite different terms from those of Aristotle, or our conception of the heart in terms very different from those of Galen. The presence of different conceptual schemes may thus be explained by the process of conceptual innovation which—at least thus far—never came to an end in human history. We should thus challenge Davidson when he says that “we get a new out of an old scheme when the speakers of a language come to accept as true an important range of sentences they previously took to be false” (Davidson 1974: 188). The point at stake is in fact different, since a change of scheme is not just a matter of saying things differently, but rather of saying different (in the sense of new) things.3 In other words, a scheme A may be committed to phenomena that another scheme B cannot even foresee: Galenic physicians, for instance, had nothing to say about viruses because those entities lay totally beyond their conceptual dimension. This means that our classical logic based on the principle of bivalence is not much help in such a context. Some assertions that are deemed to be true in a certain scheme may have no value in another scheme, so that we need to formalize this truthindeterminacy by having recourse, say, to a Lukasiewicz-style manyvalued logical system in which, besides the classical T and F, a third (Indeterminate) value I is present. We have, in sum, a much more complex picture than the one contained in Davidson’s paper. It is important to note, once again, some hints contained in Ajdukiewicz’s works. Jerzy Giedymin writes in this regard that:

3

See for example (Rescher 1980).

180 If different world-pictures cannot be compared either logically [...] or experimentally, are they equally good or can they not be compared in any way whatever?—They can be compared and evaluated in the process of “human understanding” [...] or from an “evolutionary” point of view (Giedymin 1978: xi).4 In other words, we can understand Galenic medicine, but a Galenic physician would lack the conceptual apparatus for understanding ours. So, to deny that different conceptual schemes exist is a little absurd. Of course, as we said previously, the expression “conceptual schemes” is a metaphor: we cannot see or touch them as we do with physical objects. Their presence, however, is detectable from human behaviour, and this means that they are tied to the dimension of human action. Conceptual schemes, in sum, evolve, because they are processes and not immutable structures. One should always take into account the broader models (conceptual schemes, cultural traditions) by means of which we judge our sentences— including, for example, the mythological ones—to be true or false. They are part of the “framework of conceptual thinking”5 and, as long as human beings are concerned, they can think because they are able to measure their thoughts by having recourse to standards of correctness and of relevance. The aforementioned “framework of conceptual thinking” somehow transcends the individual thought of individual thinkers. This explains why there is truth and error with respect to it, even though we may talk of entities which do not exist in the physical world. There is indeed a correct and an incorrect way to describe this framework.

References Ajdukiewicz K. (1934), “The World-Picture and the Conceptual Apparatus”, in (Ajdukiewicz 1978).

4 5

Here Giedymin refers to (Ajdukiewicz 1935). For a definition of this expression see (Sellars 1963).

181 Ajdukiewicz K. (1935), “The Scientific World-Perspective”, in (Ajdukiewicz 1978). Ajdukiewicz K. (1978), The Scientific World-Perspective and Other Essays, 1931-1963, Reidel, Dordrecht. Blackburn S. (1996), “Conceptual scheme”, in: S. Blackburn (ed.), Oxford Dictionary of Philosophy, Oxford University Press, Oxford. Brown H.I. (1995), “Conceptual scheme”, in: T. Honderich (ed.), The Oxford Companion to Philosophy, Oxford University Press, OxfordNew York. Davidson D. (1974), “On the Very Idea of a Conceptual Scheme”, reprinted in (Davidson 1984), 183-198. Davidson D. (1984), Inquiries into Truth and Interpretation, Clarendon Press, Oxford. Devitt M. (1991), Realism and Truth, Blackwell, Oxford, 2nd ed. Giedymin J. (1978), “Editor’s Introduction” in: (Ajdukiewicz 1978). Laudan L. (1996), Beyond Positivism and Relativism, Westview Press, Boulder-San Francisco-Oxford. Quine W.V.O. (1954), “Two Dogmas of Empiricism”, in: W.V.O. Quine, From a Logical Point of View, Harvard University Press, Cambridge, MA 1980, 2nd ed. Rescher N. (1980), “Conceptual Schemes”, in: P.A. French, T.E. Uehling Jr. and H.K. Wettstein (eds.), Midwest Studies in Philosophy, vol. V: Studies in Epistemology, University of Minnesota Press, Minneapolis, 323-345. Rescher N. (1994), “The Rise and Fall of Analytic Philosophy”, in: N. Rescher, American Philosophy Today and Other Philosophical Studies, Rowman & Littlefield, Lanham, 31-42. Rorty R. (1994), “The World Well Lost”, in: R. Rorty, Consequences of Pragmatism, University of Minnesota Press, Minneapolis. Sellars W. (1963), “Philosophy and the Scientific Image of Man”, in: W. Sellars, Science, Perception and Reality, Routledge & Kegan Paul, London-New York, 1-40. Skolimowski H. (1967), Polish Analytic Philosophy, Routledge & Kegan Paul, London. Wolenski I. (1989), Logic and Philosophy in the Lvov-Warsaw School, Kluwer, Dordecht-Boston.

DAVIDSON’S NATURALISM Mario DE CARO (University of Roma Tre, Italy)

One of the most relevant reasons why Donald Davidson’s philosophy has had, and still has, such a lasting influence is that it occupies a strategic position in the debate concerning the form that a suitable philosophical naturalism should take. On one side of this debate, there are the advocates of the so-called scientific naturalism, which today is probably the majority’s view, according to whom (i) the only reality is the world as described by the natural sciences, and (ii) philosophical inquiries have to be continuous with scientific inquiries (the great fortune of many naturalization projects today is a clear consequence of this view). On the other side, there are the defenders of a more liberal naturalism, who argue for a more subtle and pluralistic view of reality, and deny the continuity of philosophy with science. Davidson’s philosophy is at the centre of the battlefield between scientific and liberal naturalists, since some of its features deeply appeal to the former group, while other features appeal to the latter. This ambivalence is particularly clear in the case of anomalous monism, the highly influential Davidsonian conception of the mental. Anomalous monism is the view that every mental event is identical to a specific physical event, whereas mental properties supervene on (but cannot be reduced to) physical properties, and cannot instantiate strict laws either. This view has been widely discussed because many philosophers have seen it as a promising way of preserving what is important in naturalism (that is, ontological monism), while denying a reductionist view of the mental. There are also philosophers, however, who have tried to refute anomalous monism. Ted Honderich, Jaegwon Kim and several others, for example, have accused such a view of being nothing more than epiphenomenalism in disguise (a very bad thing to be for any respectable philosophical theory, of course).1 I will not discuss this issue here, however. My point, instead, 1

See several of the essays in (Heil and Mele 1993). There are also philosophers who do not see epiphenomenalism as a “Bad Thing”: see, for example, (Bieri 1992).

184 will be that another relevant problem affects anomalous monism, that is, the difficulty of conjugating three claims: the “token-identity” thesis, the supervenience of mental properties on physical properties, and semantic externalism.

§ 1. In Mental Events, Davidson wrote that, Mental characteristics are in some sense dependent, or supervenient, on physical characteristics. Such supervenience might be taken to mean that there cannot be two events exactly alike in all physical respects but differing in some mental respects, or that an object cannot alter in some mental respects without altering in some physical respects (Davidson 1970: 214). In this quotation Davidson remained prudently silent about what the subvenient basis of the supervenience relation was. However, this is a legitimate question one can ask—and later he asked it himself. Do mental properties supervene on the brain properties of the individual alone (that is, is each mental event identical to a specific brain event)? Or, as many externalists would say today, does the subvenient physical basis of mental events essentially include some physical properties external to the individual? This is an important question, especially with regard to the philosophy of mind and to the possibility of investigating the mind by investigating the brain. As we will see, my claim will be that Davidson was uncertain on this regard, and switched his mind between these two views—arguably because both have important pros, and important cons. Too bad, then, that it looks like one of those sad cases in which one can’t have it both ways. In my view, Davidson’s ambivalence on this issue depended on the questions that the so-called “externalist revolution” raised for his view. Let us consider this issue, then.2

2

The following part of the paper partially overlaps with some parts of (De Caro 1999a).

185 As is well known, externalists refuse the classic internalist assumption that mental content is entirely determined by the non-relational, internal properties of the mind/brain. According to them, on the contrary, content is determined, at least partially, by factors external to the mind/brain. Many versions of externalism have been proposed which can be usefully classified in different ways. The various externalist conceptions can be distinguished depending on the pervasiveness they attribute to the phenomenon of the externalist determination of content. This phenomenon can be seen either as limited to a few classes of expressions (proper names, natural kind words, and indexicals) or as an-across-the-board semantic feature.3 Davidson vigorously defends the latter view. The dependence of meaning on factors outside the head […] is ubiquitous, since it is inseparable from the social character of language. […] [I]t is a perfectly general fact about the nature of thought and speech (Davidson 1988: 167). Most typically, however, the different kinds of externalist view are distinguished by referring to the kinds of external determinants of content to which they respectively appeal. According to causal externalism mental content is essentially determined by the causal relations between the subject, on one hand, and the objects and events in the external world, on the other hand.4 According to social externalism content is determined by the sociolinguistic context in which the speaker is placed. Davidson explicitly accepts the fundamental assumption of causal externalism, stating that “causality plays an indispensable role in determining what we say and believe” (Davidson 1983: 435). He, however, criticizes all the most common versions of social externalism, since these views conceive of content as depending on social usage and linguistic conventions of which the speaker could be completely unaware—an utterly unpalatable

3 4

For this distinction, see (Bilgrami 1992), passim. The loci classici for this view are (Kripke 1972), (Putnam 1975).

186 hypothesis for Davidson.5 He, however, developed his own kind of social externalism. In order to understand the latter point, it is useful to compare Davidson’s externalism with the one proposed by the later Wittgenstein.

§ 2. First of all, one should notice that in Davidson’s externalism, the term “social” peculiarly refers more to an intersubjective context (one in which triangulation can take place because two, or more, subjects can causally interact with each other, and with their shared environment) than to a community of speakers who are bound by a common language and share a “form of life” (to use the Wittgensteinian term). Then we can notice that both Davidson and Wittgenstein resolutely refute Cartesianism and advocate an externalist view of thought and meaning; that both defend a holistic conception of mental content; that both Wittgenstein and Davidson—notwithstanding their respective unequivocal criticisms of internalism—recognize that there is an important asymmetry between the way in which we know our own thoughts, and the other forms of our knowledge. However, as a matter of fact, Davidson refuses the main tenet of Wittgenstein’s later philosophy of language and mind—this is the fundamental normative function of social rules and shared languages practices in the determination of mental content. According to Davidson, all we need to explain the supposedly puzzling Wittgensteinian notion of “following a rule” is the idea of the “joint expectations” of the speaker and the interpreter: “the hearer expects the speaker to go on as he did before; the speaker expects the hearer to go on as before” (Davidson 1994a: 233). At any rate, even if it is true that Davidson refuses the classic versions of social externalism—beginning with the Wittgensteinian one –, his conception encompasses some assumptions that clearly make mental content dependent on the existence of a intersubjective context without 5

On Putnam’s “linguistic division of labour”, see (Davidson 1994b: 5-6); on Burge’s thesis according to which communal standards of correctness are prior to individual ones, see (Davidson 1990b: 197), (1990c: 310); on Kripke’s skeptical interpretation of the Wittgensteinian notion of “following a rule”, see (Davidson 1992, 1994b).

187 which a subject could never acquire the ability of thinking and talking. In fact, Davidson’s “triangular externalism” can be reasonably described as a very personal synthesis of causal externalism and (a peculiar version of) social externalism.

§ 3. Before considering Davidson’s proposal in more detail, however, it is worth noticing that he formulated his version of externalism by covering a very personal route, one substantially independent of both the main sources of externalism in contemporary philosophy—that is, Putnam’s (Putnam 1975) and Kripke’s (Kripke 1972) views on meaning, on the one hand, and Wittgenstein’s later philosophy, on the other hand. More precisely, triangular externalism can be seen as a natural outcome of Davidson’s reflection because, since its very beginning, it has embodied many externalist motifs. Davidson himself was aware of this point, [That] the contents of our thoughts and sayings are partly determined by the history of causal interactions with the environment [...] comes naturally to someone like me who has for some thirty years been insisting that the contents of our earliest learned and most basic sentences (“Mama”, “Doggie”, “Red”, “Fire”, “Gavagai”) must be determined by what it is in the world that causes us to hold them true. It is here, I have long claimed, that the ties between language and the world are established and that central constraints on meaning are fixed; and given the close connections between thoughts and language, analogous remarks go for the contents of the attitudes (Davidson 1990b: 198-99). Beyond these semantic views (which clearly recall causal externalism), there are also other reasons for thinking that Davidson’s philosophy has always been externalist in spirit. A clear example is offered by the distinctive Davidsonian thesis according to which thoughts can be attributed only to creatures who speak a language and have the ability to interpret each other (a view that clearly makes content dependent on the existence of a social context). An even clearer example of Davidson’s early externalism is given by his unremitting polemic against the tenets of

188 internalism—that is, the dualism of conceptual scheme and empirical content (the so-called “third dogma of empiricism”); the idea that there are epistemic intermediaries and special objects of thought; the representational conception of the mind; the correspondence theory of truth; the possibility of Cartesian skepticism. To sum up: triangular externalism represented the full development of many of Davidson’s previous views, even though many of these theses already were clearly externalist in character.

§ 4. As I noted before, sometimes Davidson states that externalism should hinge upon our actual linguistic practices, not on some bizarre thought experiment. Indeed, the argument to which Davidson appeals more frequently in expounding his own externalist conception concerns a very ordinary situation—one that involves a child who is taught his or her first language by an adult. In such a situation, a fundamental triangular relationship is very clearly at work. A child (first apex of the triangle) learns a language and forms intentional states while simultaneously confronting the adult-teacher (second apex) and the objects and events that cause the utterances of the teacher (third apex). As a result of this triangular process, the child will be conditioned to hold true, in determinate circumstances, some basic sentences; and, in such a way, he will also learn the conditions of applicability of the words of which these sentences are composed. A point on which Davidson insists strongly is that this triangular process is possible only because evolution and learning have made the ways in which different human beings classify stimuli very similar. In the case discussed here, the learner and the teacher share “similarity responses” to external objects and events (Davidson 1991a). According to Davidson, without these shared responses to common stimuli, thought and language would not have any content. It takes two points of view to give a location to the cause of a thought, and thus to define its content [...] If the two people [...] note each other’s reactions (in the case of language, verbal reactions), each can correlate these observed reactions with his or her stimuli from the

189 world. The common cause can now determine the content of an utterance and a thought. The triangle which gives contents to thought and speech is complete (Davidson 1991b: 159). It should be noticed that, in this “essential triangle”, causal and social externalism are strictly interwoven. The external objects and events by which the child is causally stimulated will determine the content of sayings and thoughts; but this will happen only if an actual communicative interaction between the child and (at least) another speaker is in play. It is this communication that makes it possible for the child to understand that the adult is responding to the same object. What is fundamental here is that only actual communication can make an individual aware of the idea of objectivity, which is an essential condition for propositional thought (only a creature that knows the difference between truth and falsity can be said to have thoughts).6 As Davidson says, “[o]nly those who […] share a common world can communicate; only those who communicate can have the concept of an intersubjective, objective world” (Davidson 1994: 234). Davidson insists particularly on the example of the learning child— which has the advantage of making the explanatory virtues of triangular externalism very clear. However, it is not difficult to see that, mutatis mutandis, this kind of explanation can also be applied to any situation that involves interpretation (the triangle, in this case, will be formed by the interpreter, the interpreted speaker and the circumstances to which the speaker responds).7 According to a fundamental Davidsonian thesis, interpretation presupposes the application of the principle of charity, i.e., the necessary attribution to the speaker of both a certain degree of logical consistency and a large amount of true beliefs. The latter attribution implies that the interpreter cannot help assuming that the speaker is

6

It is important to point out that also the essential social factor to which Davidson refers is causal in character (the child is caused to hold some sentences true by the teacher’s behaviour). In this way, it is easy for Davidson to conclude that there is no contrast between social and causal externalism. 7 Actually, according to Davidson’s reconstruction of ontogenesis of language and thought, the child also behaves as a sort of interpreter (a protointerpreter, we could say).

190 respondent to the same features of the world to which the same interpreter would respond in similar circumstances (Davidson 1991b: 158). In this way, again, we can see how the two externalist forms of determination of content are inextricably engaged. The causal relationship between the external world and the speakers is essential to the determination of their mental content, but this relationship presupposes that some kind of intersubjective standard is shared by the speakers. It follows from the nature of correct interpretation that an interpersonal standard of consistency and correspondence to the facts applies to both the speaker and the speaker’s interpreter, to their utterances and their beliefs (Davidson 1990b: 211). According to Davidson8, given this externalist framework, the prestige of many of the venerable enigmas that have traditionally worried philosophers, substantially diminishes—if it does not vanish completely (think of the doubts concerning the existence of the external world and other minds). However, new quandaries are raised that will not leave philosophers jobless. It should not be assumed that when we cease to be bullied or beguiled by the scheme-content and subjective-objective dichotomies, all the problems of epistemology will evaporate. But the problems that seem salient will change (Davidson 1988: 166). Externalism, however, caused some troubles for Davidson’s philosophy. One problem, for example, regards the supervenience relation between the mental and the physical, and consequently, the token-identity thesis between the mental and the physical—and anomalous monism in general.

§ 5. A good way of evaluating the implications of these externalistic views for the overall Davidsonian philosophical project is by referring to a thought experiment presented by Davidson in “Knowing One’s Own 8

See, for example, (Davidson 1983).

191 Mind” (Davidson 1987) as well as in other places. That thought experiment concerned the so-called Swampman, a perfect physical duplicate of Davidson himself created by accident ex nihilo. The philosophical question that the Swampman raises was whether, besides being Davidson’s perfect physical duplicate of, it was also his perfect psychological duplicate (it should be noticed that Davidson always referred to the Swampman with the pronoun “it”). As I said, the Swampman is a perfect physical duplicate of Davidson, created ex nihilo by random natural forces. Its behaviour seems to fit perfectly our standards of interpretability (people mistake it for the real Davidson); ex hypothesis, however, there has been no adequate causal interaction between the Swampman and its environment. Therefore, according to Davidson, its words and thoughts have no content at all. On this basis, Davidson concludes that there is no way for us to say that the Swampman really has any thoughts (even if, of course, he could acquire them, by properly interacting with the environment and the human beings). In Davidson’s words, My replica can’t recognize my friends; it can’t re-cognize anything, since it never cognized anything in the first place. It can’t know my friends’ names (though of course it seems to), it can’t remember my house. It can't mean what I do by the word ‘house’, for example, since the sound ‘house’ it makes was not learned in a context that would give it the right meaning—or any meaning at all. Indeed, I don’t see how my replica can be said to mean anything by the sounds it makes, nor to have any thoughts […] In one word, the Swampman’s trouble was that his brain stored nothing that could connect his sounds with the world, nothing to give his thoughts and words a semantics, a content (Davidson 1987: 443-444). Prima facie, all this seems to confirm the classic externalistic slogan that “Meanings [and more generally, thoughts] are not in the head”. Despite seeming to advocate the standard externalist view, on several occasions Davidson also criticized it. He in fact wrote that—even if one accepts externalism—there still is a sense in which mental events actually are in the head, since they are identical to brain events. It is true that, in the

192 interpretive process, the interpreter has to consider the features of the social and physical environment in which the interpreted subject has learned how to use the words he or she uses. According to Davidson, however, this only means that mental events are in part identified by their causes. The same happens, he writes, when we identify a disease as snow blindness or a particular bodily state as sunburn: in both cases, in order to identify them we can refer to the causal processes that generated those states. A determinate sunburn, for example, is identical to a certain state of the body, even if the very same bodily state could have been caused by something different from exposure to sunlight, let’s say a wound. That sunburn and that wound would instantiate the same bodily state; numerical identity aside, they would be token-identical to the same physical state of the body. However, they would have different causal stories to which we should refer to when we try to explain them. For Davidson, this is true also of the relationship between the mental and the physical. He writes, [T]he correct interpretation of what a speaker means is not determined solely by what is in his head; it depends also on the natural history of what is in the head (Davidson 1988: 164). This point, however, does not have, in itself, any ontological import: every mental event is identical to a physical event, since how we describe and identify events and states has nothing directly to do with where those events are (Davidson 1988: 170). This is clearly an epistemological point. In interpreting a speaker, we do describe mental events by referring to the external environment—and this is the sense in which the causal history of these events is relevant. In themselves, however, mental events are identical to physical events in the brain of the subject. In Davidson’s words, I think that [mental] states are inner, in the sense of being identical with states of the body, and so identifiable without reference to objects or events outside the body (Davidson 1987: 444).

193 At this stage, therefore, it seems that Davidson wants to uphold the standard interpretation of the token-identity theory, according to which each particular mental state token is identical to a physical (brain) state token, even if he recognized that there is an externalist story to be considered if one wants to interpret correctly that mental state. Several things should be noticed here. First of all, in this way, Davidson seems to think that there are two ways of identifying mental events: (i) we can identify mental events with reference to the causal history that determined what is in the head (a form of identification that is crucial for interpretation); (ii) we can identify mental events “without reference to objects or events outside the body”, as he says in the quotation we just read. The latter form of identification explains, according to Davidson, a phenomenon that dualists used to explain very well, but more orthodox externalists have lot of troubles with: the authority of first-person knowledge. If mental states are also “identifiable without reference to objects or events outside the body” one can account for the special authority speakers have on their own mental states, and this (besides being an important epistemological issue in itself) is crucial for Davidson’s theory of interpretation. In fact, according to him, in order to interpret a speaker, we have to assume that the speaker knows what he or she thinks.

§ 6. However, this view also generated several difficulties for Davidson, the most important of which was connected with the correct interpretation of his claim that the mental supervenes on the physical. For a long time— and, it seems to me, even when he introduced the Swampman on the philosophical scene—Davidson advocated the claim that the mental is in a relation of weak individual supervenience with the physical. The mental weakly supervenes on the physical since, in one world, two physically identical beings would also be psychologically identical; however, this would not be true across all possible worlds, which would be so if the case was one of strong supervenience. Incidentally, the reason

194 why Davidson refuses strong supervenience is clear: if physically identical individuals were mentally identical in all possible worlds, this would imply, as noticed by Jaegwon Kim, the existence of psychophysical laws, which is inconsistent with Davidson’s thesis of the anomalousness of the mental. (In addition, as noted by (McLaughlin and Bennett 2005), Davidson owes us “an explanation of why mental and moral properties weakly supervene on physical [properties]—and it must be an explanation that does not entail that strong supervenience holds as well”9.) However, for Davidson the mental also individually supervenes on the physical, since any two individuals in the same brain state would be in the same psychological state. At this stage, Davidson thought that the mental states of an individual were in a relation of weak individual supervenience with his or her brain states (and this is still the way in which Davidson’s anomalous monism is generally interpreted). This view, however, was at odds with the externalist constrains Davidson had to accept when he started to think about the Swampman. The analogy between the sunburns and the mental states, to begin with, was not an adequate one. It is true that there is a sense in which a sunburn and a wound, if physically indistinguishable, are identical (in the sense of qualitative, not numerical identity, of course). But can we say the same for intentional mental states, for states with a content? I believe that in front of me is a glass of water. My counterpart on the Twin-earth believes that in front of him there is a glass of twater. These two mental states cannot be identical; therefore, for the property of transitivity, they cannot be identical with the same brain state. As Andrew Woodfield puts it, [An]… interesting consequence [from externalism] is that all versions of the mind-brain identity theory are false. No de re mental state about an object that is external to the person’s brain can possibly be identical with a state of that brain, since no brain state pre-supposes the existence of an external object. Any state which did incorporate an environmental object would not be a state of the brain, but would be rather a state of the brain-environment complex (Woodfield).

9

Cf. (Davidson 1985) and (1993: especially 4, n. 4).

195 At some point, probably Davidson started to think along the same lines. In fact, later in his career he seemed to have changed his mind about the kind of supervenience relating the mental to the physical. His new view was that mental properties do not supervene on the brain properties of the agent alone, since the external physical environment is also necessary for individuating—not just for describing—mental events. In this new view, externalism does not only play an epistemological role, it also plays an ontological role, since the reference to the physical world is seen as a condition of identity, not as a mere criterion of identification of mental events. In “What is Present to the Mind”, for example, Davidson wrote, [...] subjective states are not supervenient on the state of the brain or nervous system: two people may be in the same physical state and yet be in different psychological state. This does not mean, of course, that mental states are not supervenient on physical states, for there must be a physical difference somewhere if psychological states are different [Notice: Davidson does not say that the states are described differently; he says they are different!]. The physical difference may not be in the person; like the difference between water and twater, it may be (we are supposing) elsewhere (Davidson 1989: 12-13). In this way it is not just a question of identifying and describing a state (which in itself is identical to a brain state), by referring to its external causes. On the contrary, it is the mental state itself, with its specific content, that is identical to a physical state whose conditions of identity essentially refer to the external. No doubt, this (wider) definition of the supervenience of the mental on the physical fits perfectly with externalism. However, this clearly implies that, contrary to the above-mentioned quotation, it is correct to say that “thoughts are not in the head”. At that point Davidson had finally come to terms with the idea that there can be a psychological difference between two individuals without a difference inside their respective bodies. His new view implied a relation of supervenience between mental states and physical states which was partially different from the one we saw earlier—it was not a relation of

196 weak individual supervenience anymore, but one of weak global supervenience.10 Davidson’s externalist view still implied weak supervenience, since it stated that two physically identical beings in a particular world will also be psychologically identical within our world, but not across all possible worlds (as it would be if strong supervenience held). However, differently from the view seen earlier, this view also implied global supervenience, since two physically identical individuals may differ psychologically if the spatiotemporal properties in the respective environments are different. It is important to notice that Davidson insisted that the token-identity theory, encompassed by his anomalous monism, had not been compromised by his new view of supervenience. In fact, now he read the token-identity theory as merely saying that every psychological state is identical to a physical state, but not necessarily to a brain state. At any rate, Davidson tried to convince us that his new view on supervenience was good news for anomalous monism—and in particular for the idea that the mental is essentially anomalous. His idea was that if the subvenient basis of psychological states includes physical states external to the subjects then the idea that there are not psychophysical strict laws is strengthened. If mental properties are supervenient not only on the physical properties of the agent but in addition on the physical properties of the world outside the agent, there can be no hope of discovering laws that predict and explain behaviour solely on the basis of intrinsic features of agents (Davidson 1995: 157).11 (Also on this issue Davidson’s reasoning is not very convincing. His argument for the anomalism of the mental is a priori, and it is aimed at 10

See Paul (Teller 1995: 485), according to whom the notion of “global supervenience” depends on the comparison of “cases comprised by the way the whole world might be: two counterfactual scenarios describing whole world histories which agree in all subvening respects agree in all supervening respects”. 11 On this point, see also (Loar 1993).

197 proving that psychophysical laws are impossible. In itself, on the contrary, the acceptance of global supervenience only makes very implausible that we will ever be able to detect those laws, if they exist, but does not seem to add much to the issue whether they objectively exist or not.)

§ 7. This new view on supervenience does do justice, in my view, to the externalist intuitions that generated it. However, the claim that mental events supervene also on extraindividual physical events caused big problems for Davidson’s philosophy. First, as I just said, the original interpretation of the token-identity thesis, for which mental events supervene of brain events, was proven wrong (but this is also a big problem for classic cognitive science, as is well known). Second, if my mental states supervene on external physical states of which I can easily be unaware, my first-person authority is dramatically shaken (a problem to which Davidson was very sensitive). Third, as has often been noticed, given this externalist view, the project of radical interpretation also seems to encounter a serious difficulty. In fact, as Davidson himself notices, a correct interpretation also requires that the interpreter account for the externalist determination of the content of the intentional states of the interpreted speaker. However, in fact, the interpreter may remain utterly unaware of this kind of determination—for example, such determination could concern the interpreted speaker’s past. Fourth, in this scenario there is a problem for Davidson with regard to the principle of individuation of the states that are actually token-identical. At some point, in fact, under Quine’s pressure, Davidson came to admit that his original attempt to individuate events by referring to their causes was not acceptable, because this kind of individuation is unpredicative (this is because causes themselves are events, and so the causal individuation of a particular event would unpredicatively quantify over the entire class of events). So Davidson had to swallow the idea that states are individuated by their spatial-temporal locations (two states with the same spatial-temporal location are identical). Now, however, we can see that the pressure of externalism forced Davidson to come back to the individuation of states in causal terms. Unpredicativity was back.

198 In general, the original interpretation of the token-identity thesis given by Davidson—for which mental states supervene of brain states—was proven inadequate (but it should be noticed that this is also a general problem for classic cognitive science). In the first place, identity no longer concerned single states, but very complicated causal processes (“the natural history of what is in the head”), which might involve, for example, the causal influence of either water or twater on the thinker. In the second place, in this scenario the supervenient mental level seems to have a character that the subvenient physical level does not share, so that the token-identity, besides not holding for the final mental and brain states, does not hold for the respective causal stories either. This is because for individuating a mental state, it appears necessary to consider its causal history (i.e., my belief that there is water in front of me has the content it has also because of its causal history), whereas this is not true of the physical level (every physical state can be individuated whatever causal history it has; and the same physical state could have been produced by different sufficient causes). It seems, therefore, that causal individuation at the mental level has a modal character that causal individuation at the physical level does not share; so by Leibniz’s law of the indiscernibility of identicals, they cannot be identical. If this true, the idea that mental states, or processes, supervene on physical processes seems rather unstable in the Davidsonian scenario. These were very serious difficulties for the Davidsonian projects as a whole. For this reason, perhaps, Davidson reconsidered his attitude toward the most proximate cause of these difficulties—the Swampman. And probably it had been the scenario of the Swampman (with the other counterfactual, Twin-earth based scenarios) that had driven Davidson to conclude that it was unreasonable to keep saying that mental events supervene on brain events alone. It was probably also for this reason that Davidson took the Draconian decision of disposing of his creature. Be this as it may, in (Davidson 1999) we can read that the Swampman, as a bad child, was “embarrassing” his creator—who in return abandoned it to its destiny. The whole scenario involving the Swampman, he wrote, was actually misleading. In philosophical reflection we should only appeal to our ordinary intuitions. The counterfactual scenarios so loved by philosophers are conceived exactly to confuse those precious intuitions.

199 Therefore, Davidson now said, the right question philosophers should ask themselves is “how do we tell when a creature is thinking, when we use our ordinary intuitions and knowhow?” The answer to that question was easy, according to Davidson, and no bizarre philosophical fiction—such as the Swampman—should stop us from answering the query correctly: “If we can communicate with the creature or robot on a range of topics in our natural environment, it is conscious and it is thinking”. In my view, however, the Swampman was innocent of the charge, so its execution by its creator was a terrible injustice. The Swampman was the legitimate product of conceptual analysis, which correctly refers to counterfactual, sometimes bizarre, scenarios—not only to ordinary intuitions. It was a legitimate question to ask if a creature with the causal story of the Swampman would have been able to think and talk meaningfully. Moreover, some reflections proved that the Swampman would not be thinking or talking at all. If, however, this conclusion really compromises our intuitive trust in first-person authority, as well as in our theories about interpretations and the relation between the mental and the physical, well, too bad for that authority and for those theories.

References Bieri P. (1992), “Trying Out Epiphenomenalism”, Erkenntnis 36, 1992, 283-309. Bilgrami A. (1992), Belief and Meaning, Blackwell, Oxford. Burge T. (1979), “Individualism and the Mental”, in: P. French, T. Uehling and H. Wettstein (eds.), Metaphysics, Midwest Studies in Philosophy 4, University of Minnesota Press, Minneapolis, 73-121. Davidson D. (1970), “Mental Events”, in: L. Foster and J. W. Swanson (eds.), Experience and Theory, University of Massachusetts Press, Amherst, 79-101, reprinted in (Davidson 1980a), 207-27. Davidson D. (1980), Essays on Actions and Events, Clarendon Press, Oxford.

200 Davidson D. (1983), “A Coherence Theory of Truth and Knowledge”, in: D. Henrich (ed.), Kant oder Hegel, Klett-Cotta, Stuttgart, 423-38, reprinted in (Lepore 1986), 307-19. Davidson D. (1985), “Replies to Essays X-XII”, in: B. Vermazen and M.B. Hintikka (eds.), Essays on Davidson Actions and Events, Clarendon Press, Oxford, 242-252. Davidson D. (1987), “Knowing One’s Own Mind”, in: Proceedings and Addresses of the American Philosophical Association 61, 441-58. Davidson D. (1988), “The Myth of the Subjective”, in: M. Benedikt and R. Burger (ed.), Bewustein, Sprache und die Kunst, Edition S, Verlag der Österreichischen Staatsdruckerei, Wien, 45-54, reprinted in M. Krausz (ed.), Relativism. Interpretation and Confrontation, University of Notre Dame Press, Notre Dame, 159-72. Davidson D. (1989), “What is Present to the Mind?” in: J. Brandl and W.L. Gombocz (eds.), The Mind of Donald Davidson, Grazer Philosophische Studien 36, Rodopi, Amsterdam, 3-18. Davidson D. (1990b), “Epistemology Externalized”, Análisis Filósofico 10, 1-13 (in Spanish), Engl. transl. in Dialectica 45, 1991, 191-202. Davidson D. (1990c), “The Structure and Content of Truth”, Journal of Philosophy 87, 297-328. Davidson D. (1991a), “Meaning, Truth and Evidence”, in: R. Barrett and R. Gibson (eds.), Perspectives on Quine, Blackwell, Oxford, 68-79. Davidson D. (1991b), “Three Varieties of Knowledge”, Philosophy 66, 156-66. Davidson D. (1992), “The Second Person”, in: P. French, T. Uehling, and H. Wettstein (eds.), The Wittgenstein Legacy, Midwest Studies in Philosophy 17, 255-67. Davidson D. (1993), “Thinking Causes”, in (Heil and Mele 1993), 3-17. Davidson D. (1994a), “Self-Portrait”, in: S. Guttenplan (ed.), A Companion to the Philosophy of Mind, Blackwell, Oxford, 231-36. Davidson D. (1994b), “The Social Aspects of Language”, in: B. McGuinness and G. Olivieri (eds.), The Philosophy of Michael Dummett, Kluwer, Dordrecht, 1-16. Davidson D. (1995), “Could There Be a Science of Rationality?” International Journal of Philosophical Studies 3, 1-16, reprinted, with an afterward, in M. De Caro and D. Macarthur (2004) (eds.),

201 Naturalism in Question, Harvard University Press, Cambridge, MA, 152-169. Davidson D. (1999), “Interpretation: Hard in Theory, Easy in Practice”, in (De Caro 1999b), 31-44. De Caro M. (1999a), “Davidson in Focus”, in (De Caro 1999b), 1-29. De Caro M. (1999b) (ed.), Interpretations and Causes, Kluwer, Dordrecht. Heil J. and A. Mele (1993) (eds.), Mental Causation, Oxford University Press, Oxford. Kripke S. (1972), Naming and Necessity, in: G. Harman and D. Davidson (eds.), Semantics of Natural Languages, Reidel, Dordrecht, 253-355, reprinted, with a new introduction, by Harvard University Press, Cambridge, MA, 1980. Kripke S. (1982), Wittgenstein on Rules and Private Language, Blackwell, Oxford. Lepore E. (1986), Truth and Interpretation. Perspectives on the Philosophy of Donald Davidson, Blackwell, Oxford. Loar B. (1983), “Can we confirm supervenient properties?” in: E. Villanueva (ed.), Naturalism and Normativity, Ridgeview, Atascadero, 74-92. McLaughlin B. and K. Bennett, (2005), “Supervenience”, Stanford Encyclopaedia of Philosophy: http://plato.stanford.edu/entries/supervenience. Putnam H. (1975, “The Meaning of ‘Meaning’”, in: K. Gunderson (ed.), Language, Mind and Knowledge, Minnesota Studies in the Philosophy of Science 7, University of Minnesota Press, Minneapolis, reprinted in: H. Putnam (1975), Minds, Language and Reality, Philosophical Papers, 2, Cambridge University Press, Cambridge, 215-71. Teller P. (1995), “Supervenience”, in: J. Kim and E. Sosa (eds.), A Companion to Metaphysics, Blackwell, Oxford, 484-6.

DAVIDSON, SELF-KNOWLEDGE, AND SKEPTICISM M. Cristina AMORETTI (University of Genoa, Italy)

Externalism concerning the contents of mental states, generally, is the thesis that the contents of at least some of our propositional attitudes depend partly on objects and events external to the subject who entertains them. In other words, externalists maintain that at least some of our mental contents do not merely supervene upon the internal physical states and intrinsic properties of individuals. Many critics, however, have argued that, while externalism could seem a fascinating and attractive theory, it must be rejected because it brings with it some counterintuitive epistemological consequences. In particular, “incompatibilists” argue that externalism leads to skepticism about privileged self-knowledge. Moreover, following (McKinsey 1991), they also claim that when we try to deny this conclusion, we are then forced into a reductio: the combination of externalism and privileged selfknowledge would actually imply the controversial acceptance of some empirical propositions about the external world purely on a priori grounds (Reductio Argument). In this paper I shall not consider the achievement problem—that is the challenge of determining if externalism and privileged self-knowledge can really be compatible—but I will confine myself to the consequence problem.1 More specifically, in §1, I will attempt to characterize Davidson’s externalism and, in §2, the notion of self-knowledge that he wishes to defend. Then, in §3, I will try to ascertain whether Davidson’s kind of externalism is able to reply to the objection raised by the Reductio Argument. Finally, in §4, I shall outline some consequences of the above results for the anti-skeptical argument that Davidson hopes to derive from his own theory.

1

This terminology—achievement vs. consequence problem—was introduced by (Davies 1998). In fact, even after the achievement problem has been solved, the consequence problem can raise further objections against compatibilism.

204 1. Davidson’s triangular externalism In my opinion, the real argument that Davidson exploits to establish his own externalism derives from the theory of triangulation.2 “Triangular externalism” (as I shall call it from now on) is quite different from other traditional kinds of externalism, like those proposed by Putnam or Burge, since it can be considered as an attempt to combine both causal and social externalism. It is a form of causal externalism to the extent that the content of (some3) propositional attitudes depends, at least in part, on objects and events in the external world. Actually, referring to triangulation, it is very important to underline that dependence on external factors (or rather dependence on the causal history of the subject) is only partial. The mere connection between the subject and the world, in fact, is not at all sufficient to determine the objective content of her propositional attitudes. Thus a third element, the interpreter, is also necessary. According to Davidson, one single creature is not able to find out which is the right external cause that constitutes the objective content of her mental states, since there must be at least another creature whose perceptual reactions and concept-forming abilities are innately substantially similar to those of the subject, so that she can pick out the right cause, which is actually the external common cause. On this basis— as the objective content is necessary to forming any propositional attitude, and such objective content can be determined only by the mutual interaction of (at least) two creatures sharing similar responses to a common external environment as well as similar responsiveness to the sharing of those very responses—it is clear that the presence of the interpreter is necessary for thought. The role of the second subject is 2

The notion of “triangulation” was introduced for the first time in (Davidson 1982). While the theory of triangulation has become a central notion in his whole philosophy, it is particularly crucial to characterize Davidson’s kind of externalism concerning the contents of mental states. Unfortunately, triangulation is somewhat difficult to interpret and not all critics will agree with my own understanding. 3 Since Davidson thinks that one person’s mental contents are holistically interconnected, it is possible to affirm that, at least indirectly, all of our mental states depend on objects and events of the external world.

205 precisely the social element that Davidson introduces in his account of meaning and mental content. I would introduce the social factor in a way that connects it directly with perceptual externalism, thus locating the role of society within the causal nexus that includes the interplay between persons and the rest of nature (Davidson 1990: 201). The situation, however, is slightly more complicated. The presence of a second subject and the actual intersection of at least two causal chains leading to similar shared responses are necessary, but not sufficient, to finding out the common “relevant cause” and thus to determining objective content. In fact, we need to be aware of the objectivity of content and thus of the very possibility of error (which alone can explain the normative character of our judgments). In order to do this, it is not sufficient that two or more creatures simply interact, but it is also necessary that they actually communicate with each other. To put it another way, triangulation and language cooperate not only to fix the cause of our beliefs and to determine their objective content, but also to provide us with the concept of objectivity and that of error. The identification of the objects of thought rests, then, on a social basis. Without one creature to observe another, the triangulation that locates the relevant objects in a public space could not take place. […] [T]he presence of two or more creatures interacting with each other and with a common environment is at best a necessary condition for such a concept. Only communication can provide the concept, for to have the concept of objectivity, the concept of objects and events that occupy a shared world, of objects and events whose properties and existence is independent of our thought, requires that we are aware of the fact that we share thoughts and a world with others (Davidson 1990: 202). Over and above this, external causes—or rather, a subject ’s own causal history—only partially individuate the content of her beliefs, as internal coherence is also required. In other words, since the content of one single

206 belief necessarily depends on the contents of other beliefs, there are some holistic inferential constraints that must be considered both for individuating and attributing mental content. When a subject learns new words and concepts, in fact, she does not do so in an atomistic way. Conversely, she immediately connects meanings, concepts, beliefs and other propositional attitudes in a holistic framework created by those real interactions which, in an intersubjective linguistic environment, have tied in her mental states with external objects and events, other subjects and other thoughts. Hence, content is also determined by its inferential relations with a vast and coherent array of other beliefs, desires and so on. In a nutshell, the theory of triangulation certainly leads to a form of externalism concerning mental content. However, triangular externalism is weak. In fact, even though content depends on external objects and events, this dependence is not the only component, since social and holistic factors are equally fundamental. That is to say, there are two basic elements in Davidson’s notion of content: a referential element (introduced by the history of actual interactions with external factors) and an inferential one (introduced by holistic constraints). The former connects content to the external world, while the latter connects a single belief to other beliefs entertained by the subject.

2. Davidson’s self-knowledge Before analyzing the Reductio Argument, we shall determine which kind of self-knowledge Davidson wishes to defend. Of course, when we talk about “a priori” knowledge, we are not referring to “innate” knowledge. Experience does play a fundamental role in the acquisition of privileged or a priori knowledge, but no further empirical evidence is required to justify such knowledge. Self-knowledge, therefore, can be obtained without investigating the world further and, moreover, with special authority too.4 Discussing self-knowledge, Davidson never talks about a priority or privileged access explicitly, but he still believes there is an evident 4

First person authority is the thesis according to which there is a strong presumption that second order beliefs are predominantly true.

207 asymmetry between the first person and the third person points of view.5 First person access is not in fact a kind of access one may have to the mind of another subject. For example: It is seldom the case that I need or appeal to evidence or observation in order to find out what I believe; normally I know what I think before I speak or act (Davidson 1987: 15). […] [P]eople generally know without recourse to inference from evidence, and so in a way that others do not, what they themselves think, want, and intend (Davidson 1988: 48). While I can only know someone else’s beliefs and other propositional attitudes from what she says and how she behaves (i.e. from empirical investigation), I may know my own beliefs and other propositional attitudes in a way that is not based on any evidence or observation. As Davidson points out, these basic claims do not imply that self-knowledge is infallible, or complete (involving all the subject’s propositional attitudes), or even incorrigible. Thus, this picture is compatible with occasional failures of self-knowledge (such as self-deception and wishful thinking). Nevertheless, even if we do not characterise first person access as infallible and incorrigible, it still remains a priori in a weaker sense: more specifically, a priori simply means “obtained independently of empirical investigation” or “from the armchair”. What is the source of such knowledge? Self-knowledge cannot be based on inference (since it is immediate precisely in the sense of noninferential), but it cannot be based on inward looking either, because according to Davidson we must reject any representationalist theory of mind where self-knowledge is imagined as an inner eye looking at shadowy figures acting in a strange theatre inside the mind. Given all this,

5

The asymmetry is not substantial, because Davidson does not believe that subjective knowledge is more basic or fundamental compared to objective and intersubjective knowledge. Nevertheless, he does think that this asymmetry exists and it is perfectly compatible with his own externalism.

208 perhaps self-knowledge cannot be considered as an authentic cognitive achievement. Many critics often distinguish between two notions of a priority, namely strong and weak a priori. In the first case, the justification does not rest on any empirical assumption and it is not open to challenge on a posteriori grounds; in the second case, the justification does not rest on any empirical assumption either, but it is actually open to challenge on a posteriori grounds. It seems clear to me that Davidson’s own interpretation of self-knowledge implies the notion of weak a priori. At any rate, so as to remain within the scope of this article, it is sufficient to recognize that Davidson is committed to accepting privileged access, since he acknowledges that, although empirical evidence is required (or even necessary) to form beliefs about the world (and beliefs in general), a subject can know that she has that belief without further empirical investigation or from the armchair. More precisely, she can acquire this kind of knowledge non-inferentially and non-empirically (i.e. a priori in McKinsey’s terms6).

3. Reductio Argument The Reductio Argument was introduced by (McKinsey 1991)7 and it immediately generated great debate. The simplest formulation of the argument runs as follows:8 6

“[T]he idea is that we can in principle find out about these states in ourselves ‘just by thinking’, without launching an empirical investigation or making any assumptions about the external physical world. I will call knowledge obtained independently of empirical investigation a priori knowledge. And I will call the principle that it is possible to have a priori knowledge of one’s own neutral cognitive attitude states, the Principle of Privileged Access, or just ‘privileged access’ for short” (McKinsey 1991: 175). 7 See also (McKinsey 1997, 2002, 2003). 8 See for instance (Boghossian 1997), (Brewer 2000), (Brown 1995, 1999, 2003, 2004), (Brueckner 1992, 2000, 2002, 2003, 2004a), (Davies 1998, 2000, 2003), (Ebbs 2003), (Falvey 2000), (McLaughlin 2003), (McLaughlin & Tye 1998), (Noonan 2000), (Noordhof 2004), (Nuccetelli 1999), (Raffman 1998), (Wright 2000, 2003).

209 (P1) S can know a priori that she believes that water is wet; (P2) S can know a priori that if she believes that water is wet, then her environment contains water; (C) S can know a priori that her environment contains water. (P1) is just a direct consequence of the assumption of privileged selfknowledge according to which a subject can acquire knowledge of her own beliefs merely on a priori grounds (non-inferentially and non-empirically). According to incompatibilists, (P2) is taken to be an upshot of externalism. If externalism is true, then S can know that having certain beliefs necessarily entails particular conditions on her environment, namely that her environment contains water. Moreover, externalism—and thus the above relation—is knowable a priori, because it is established from the armchair by mental experiments or other philosophical arguments, which do not need any further empirical investigation. The consequence (C), however, is quite unlikely since the possibility of knowing an empirical truth like “the environment contains water” on a priori grounds seems to many extremely implausible and even absurd. If we agree with this conclusion, then the argument effectively represents a reductio of the joint hypothesis of externalism and privileged self-knowledge. There are several possible ways to try to dismiss the Reductio Argument.9 At the outset, Davidson might challenge the first premise, saying that a subject cannot know his own beliefs and other propositional 9

It is beyond the scope of this paper to analyze all possible replies to the argument. Briefly, the main strategies are: (1) denying (P1) arguing that the premise cannot be known completely on a priori grounds—see for instance (Brewer 2000) and the critique presented in (Brueckner 2004a); (2) denying (P2) arguing that the premise cannot be known on a priori grounds or, alternatively, that its a priori knowledge does not imply what is explicitly stated in the premise itself or any equivalent conditionals—see for instance (Brown 2004), (Brueckner 1992), (McLaughlin & Tye 1998); admitting that both (P1) and (P2) are correct, but denying the transmission of warrant to the conclusion—see (Davies 2000) and (Wright 2000) (and, for a further discussion, (Beebee 2001), (Brown 2003, 2004), (Brueckner 2004b), (Davies 2003), (McKinsey 2003), (McLaughlin 2003), (Pritchard 2002), (Warfield 2004), (Wright 2003)); accepting the whole argument, but denying the implausibility of its conclusion—see (Sawyer 1998) and (Warfield 1998).

210 attitudes completely a priori. Indeed—following triangular externalism— privileged self-knowledge actually depends on previous empirical investigation. Though this line of reasoning may apparently seem plausible, I agree with (Davies 2000) and other critics that it is not the right move to make. We have already seen that although experience plays a pivotal role in the acquisition of self-knowledge, the distinctive feature of such knowledge is that it does not require any further empirical investigation to get a justification. In other words, it is uncontroversial that the subject already knows that her environment contains water on empirical grounds; yet the contentious issue is supposing that, in addition, she can also know this on a priori grounds, i.e. without further evidence or observation. Moreover, if Davidson contested this premise, he would simply beg the question denying the mere possibility of privileged selfknowledge (or, alternatively, he would come back to the achievement problem since the Reductio Argument simply considers what consequences would follow from a compatibilist approach). In point of fact, I feel that the best way to demonstrate that Davidson’s theory is immune to the objection raised through this very argument is to attack (P2), showing that this second premise is not implied by triangular externalism. There are two strategies to obtaining such a result. The simpler one consists in demonstrating that triangular externalism cannot be known a priori; alternatively, we can accept that triangular externalism can be known a priori, but deny that such externalism actually leads to the kind of entailment between someone’s beliefs and the external world required by the argument. First of all, could triangular externalism be established merely on a priori grounds? In other words, we need to determine whether Davidson’s externalism can be derived exclusively from philosophical arguments (that is from the armchair10). As I see it, the answer is probably no. Davidson’s 10

The ambiguity of the concept of a priori is worth stressing (Nuccetelli 1999): in the first premise it indicates non-inferential knowledge, while in this second premise it is exactly a case of inferential knowledge. Nevertheless, we can skip over this objection by identifying a more general notion of a priori, which encapsulates all non-empirical knowledge (both inferential and non-inferential). See also (Miller 1997) and the reply by (Brueckner 2000).

211 approach, in fact, is explicitly different from all traditional methods of introducing and justifying perceptual and social externalism. (Putnam 1975), for instance, found his own causal externalism on the famous Twin Earth thought experiment. In this case—since a thought experiment does not require any empirical investigation—it is quite obvious that such an externalism can be known a priori.11 On the contrary, Davidson clearly denies the necessity for any thought experiment and—as I showed in the first section—bases his triangular externalism exclusively on the actual practice of communication between two or more speakers. I have a general distrust of thought experiment that pretends to reveal what we would say under conditions that in fact never arise. My version of externalism depends on what I think to be our actual practice (Davidson 1990: 199). If what Davidson says is correct, then his triangular externalism—and consequently (P2)—cannot be known merely on a priori grounds: in order to define its characteristics, we need to carry out an empirical investigation of the actual communicative practices of human beings. Nevertheless, we may still ignore Davidson’s explicit words and intentions and assume that his own triangular externalism can also be known purely on a priori grounds12 (that is to say by means of philosophical arguments or just by reflecting on the nature of mental content), and hence suppose that to that extent incompatibilists are right. Now, we are faced with the following question: does triangular externalism involve the precise kind of entailment between somebody’s beliefs and the external world, which is necessary to make the Reductio Argument properly work? I explained that incompatibilists justify (P2) saying that it is a direct outcome of externalism. In other words, they maintain that if we can know externalism on a priori grounds, then we must know on a priori grounds 11

A similar thing can be said about Burge’s social externalism (Burge 1979, 1982). According to (McLaughlin 2003), just strong a priority makes the argument run, while Davidson does not think that his triangular externalism is knowable strongly a priori. Hence, if McLaughlin and Davidson’s claims were both right, there would be no reason to look for a further reply to the Reductio Argument. 12

212 that the proposition that S believes that water is wet necessarily depends on the fact that her environment really does contain water. To see if triangular externalism is able to dismiss that objection, we should answer the following question: what does a subject, who knows triangular externalism merely on a priori grounds, truly know on a priori grounds? Knowing triangular externalism a priori means knowing the argument from triangulation a priori. Let us imagine that all the following premises (which have been discussed in the first section) are knowable from the armchair: – Objective content is necessary for there to be any propositional attitude. – Finding out the relevant cause is necessary (but not sufficient) in determining objective content. – At least two creatures sufficiently similar to triangulate on the background of a mutual external world (made of material macroscopic objects) are necessary to finding out the relevant cause. – The process of triangulation guarantees that the external world is more or less as I think it is, i.e. if I have a certain set of beliefs, then that set is largely true. This is why we can infer that if we have any propositional attitude, then there is an external world—more or less as we think it is—containing objects, and at least another creature like us. Given all this, I maintain that a priori knowledge of triangular externalism does not allow the inference stated in (P2). More precisely, though the belief that water is wet is partly13 individuated by some environmental condition like the actual presence of water, mere knowledge of the argument for triangular externalism (which is now supposed to be a priori) does not enable a subject who believes that water is wet to know a priori that if she believes that water is wet then her environment contains water.

13

We should remember that the causal element is only one part of the content, because holistic constraints play an equally important role.

213 Davidson characterizes his externalism in a counterfactual way: if the world had been different, then our mental states would have been different. In his own words: What I propose is a modest form of externalism. If our past—the causal process that gave our words and thoughts the content they have—had been different, those contents would have been different, even if our present state happens to be what it would have been had that past been different (Davidson 1999: 165). For our present purpose we can rephrase the above thesis as follows: if the subject had triangulated twin-water instead of water, then she would have entertained twin-water-beliefs and lacked water-beliefs. Consequently, a critic may be tempted to argue that counterfactual dependence leads precisely to the kind of implication needed by the Reductio Argument: if the subject has a water-belief (for example the belief that water is wet), then her environment contains water. Such a conclusion is indeed false. First of all, we must remember that Davidson thinks that the contents of one’s propositional attitudes are partly determined by one’s causal history, that is the history of real causal interactions between the subject and the outside environment during her whole life.14 Now, let us imagine that a subject has water-beliefs because, living on Earth, she had acquired the concept of “water” by triangulating water with other subjects. Then, at a certain point in her life, she unwittingly becomes a victim of a switch: she does not know, but she is moved to Twin Earth, which is apparently identical to our Earth, but instead of water there is another substance, twinwater. Since she has never interacted with twin-water, she still has water thoughts (due to her history of causal interactions with water): she thinks that water is wet, but her actual environment does not contain water.15 14

See (Davidson 1987). More precisely: “[P]eople who are in all relevant physical respects similar (or ‘identical’ in the necktie sense) can differ in what they mean or think. […] But of course there is something different about them, even in the physical world; their causal histories are different” (Davidson 1987: 33). 15 Maybe, we can also admit that, after a while, she will acquire twin-water thoughts since her causal history has changed.

214 Even though the subject cannot infer the conclusion that her environment currently contains water, Davidson’s externalism is, alas, still committed to the following problematic disjunction: her environment contains or has contained water. However, we should consider that Davidson does not think that all of our mental contents directly depend on external factors, although the external element influences the whole thought due to the holistic relations among our propositional attitudes: Not that all words and sentences are this directly conditioned to what they are about […] The claim is that all thought and language must have a foundation in such direct historical connections (Davidson 1987: 29). Hence, in order to infer the existence of water from her water-thoughts, the subject should also know that water is one of those concepts which actually “hook” language and mind to the world. But that further knowledge, not surprisingly, cannot be conceivable merely on a priori grounds. Nevertheless, (Brown 2004) correctly points out that there is another way out for incompatibilists: Perhaps a subject could use anti-individualism to gain a priori knowledge of conditional principles of the form: if a subject is in a certain psychological condition and certain other conditions hold, then her environment must be some particular way. If the subject could know a priori that she meets the antecedent of one of these conditionals, then perhaps she could use it to gain a priori knowledge of some specific connection between her thoughts and the environment (Brown 2004: 274). Two different steps are necessary to check if this strategy could really be effective. More precisely, we need to determine which conditional principles can be established by triangular externalism, and whether a subject can know a priori the antecedents of those conditional principles. According to triangular externalism, contents are individuated partly by the

215 history of causal interactions between the subject and the external objects of her environment in the process of triangulation, but also partly by their holistic relations with other contents. Let us suppose that a subject has a belief about an external object, for example the belief that water is wet. How may she have acquired such a belief? She may have systematically interacted with water in the process of triangulation, or her water-belief may be holistically connected to other beliefs that she had acquired by systematically interacting with external objects in the process of triangulation. However, since a water-belief is a belief about an external object, if S has a water-belief someone else must have systematically interacted with water in the process of triangulation. Hence we can propose the a priori knowledge of the following conditional principle16: (T) If S believes that water is wet, and S has systematically interacted with water in the context of triangulation, or someone else has systematically interacted with water in the context of triangulation, then S’s environment contains water. Since we are supposing that a subject could have a priori knowledge of the argument from triangulation, she could also have a priori knowledge of the above principle (that we inferred from that very argument). Hence, if she could have a priori knowledge of its antecedent, then she could use it to gain a priori knowledge about some contingent particular facts of her environment, more specifically that her environment contains water. Again, such knowledge would represent a reductio of compatibilism. The argument would run as follows: (P1) S can know a priori that she believes that water is wet; (P2*) S can know a priori triangular externalism; (P2’) S can know a priori that if she believes that water is wet, and she has systematically interacted with water in the process of triangulation, or someone else has systematically interacted with water in the process of triangulation, then her environment contains water; (C) S can know a priori that her environment contains water. 16

Limited to those objects that can effectively be triangulated.

216 (P2’) is the conditional principle that we can infer by our a priori knowledge of triangular externalism (P2*), and as a consequence it is knowable a priori, too. However, the conclusion (C) does not follow from the above argument because the antecedent of (P2’) evidently is not knowable merely on a priori grounds. In fact, the argument lacks a further premise: (P3) S can know a priori that she has systematically interacted with water in the process of triangulation, or someone else has systematically interacted with water in the process of triangulation. But the subject cannot know a priori that she or anyone else has actually interacted with water in the process of triangulation. On the contrary, further empirical investigation is obviously required to know that water or any other empirical object has actually been triangulated with by (at least) another creature sharing a mutual environment with the subject. If the above reasoning is correct, once more the Reductio Argument cannot imply the undesired conclusion (C). Nevertheless, incompatibilists have one more last move: they could argue that triangular externalism yields the conclusion that the whole thought—directly or indirectly—depends on contingent and empirical relations with the external world. And, following this reasoning, they could also say that this very inference is knowable a priori. More precisely, let us suppose that a subject has a belief of some kind. If she knows a priori triangular externalism, then she also knows a priori the following implication: if someone has some propositional attitudes, then there must be at least another creature sufficiently similar to triangulate with her on the background of a mutual external world which is more or less as she thinks it is. Then the argument will run as follows: (P1*) S can know a priori that she believes that p; (P2*) S can know a priori triangular externalism; (P2’) S can know a priori that if she believes that p, then there must be at least another creature sufficiently similar to triangulate with her on the background of a mutual external world which is more or less as she thinks it is;

217 (C) S can know a priori that there must be at least another creature sufficiently similar to triangulate with her on the background of a mutual external world which is more or less as she thinks it is. In the reasoning above, all premises can easily be known from the armchair and thus the Reductio Argument runs. At this very point, however, the argument has radically changed its original structure, and now it is unclear to me if it still constitutes a reductio. In fact, such a new formulation no longer has its initial strength, and thus the argument loses any pretence at dismissing the compatibility of externalism and privileged self-knowledge. As we have already seen, according to triangular externalism the link between our own minds and the external world is an intrinsic and basic feature of our very capacity to have beliefs and other propositional attitudes. It is straightforward that if we admit the a priori knowledge of triangular externalism, we also admit the a priori knowledge of its fundamental characteristics, namely that the external world and other creatures are necessary to having any thought at all. And, as a corollary, we must agree that our general picture of the world cannot be radically different from how the external world really is. Accepting this formulation of the argument, it is not possible to infer knowledge about particular empirical objects of the outside world merely on a priori grounds. But what about the more general conclusion we have just analyzed? Is it still unlikely? If we reflect on the latter interpretation of the Reductio Argument, perhaps we might transform the objection into a reasonable argument in favour of externalism: if externalism and privileged self-knowledge can truly stand together, then a tough antiskeptical argument could be advanced. In other words, we might agree with (Peacocke 1996): [The above argument] does not seem to me an objection, but rather to point positively to a promising line for developing contemporary forms of transcendental arguments (Peacocke 1996: 297).

218 4. The anti-skeptical argument revisited In this final section I shall attempt to outline some consequences of the above results for the argument against skepticism that Davidson hopes to derive from his own kind of externalism. [I]t does seem to me that if you accept perceptual externalism, there is an easy argument against global skepticism of the senses of the sort that Descartes, Hume, Russell and endless others have thought requires an answer. […] If anything is systematically causing certain experiences (or verbal responses), that is what the thoughts and utterances are about. This rules out systematic error. If nothing is systematically causing the experiences, there is no content to be mistaken about […] Any one who accepts perceptual externalism knows he cannot be systematically deceived about whether there are such things as cows, people, water, stars and chewing gum. Knowing why this is the case, he must recognize situations in which he is justified in believing he is seeing water or a cow. In those cases where he is right, he knows he is seeing water or a cow (Davidson 1990: 200201). Davidson wants to find an argument against the global skepticism of the senses, namely against the thesis that we are not justified in believing that there is an external world, nor that our image of the world is largely correct. He claims that his theory can do the job: if we know the argument for triangular externalism, then we must recognize that we are justified (by that very argument) in believing what we actually believe and thus, when those beliefs are true, they also count as knowledge. In order to summarize Davidson’s anti-skeptical argument, let us suppose that it is possible to know a priori the following premises: (P1) Triangular externalism is true. (P2) If a subject S has a certain set of beliefs, then there is an external world, more or less as S thinks it is, and there are also other rational creatures. (P3) S has a certain set of beliefs.

219 (C) There is an external world, more or less as S thinks it is, and there are also other rational creatures. Now, the argument clearly runs, hence we should verify if its premises are actually true and justifiable from the armchair, that is if they count as a priori knowledge. (P1) is the assumption of triangular externalism; (P2) is just a consequence of (P1); and (P3) is the thesis of privileged selfknowledge (that no traditional skeptic about the external world would set aside). In the earlier section I explicitly accepted the a priori knowledge of (P1) just for argument’s sake, but now I want to consider how triangular externalism may be known a priori. Of course, a priori may stand for non-empirically, but certainly it cannot mean non-inferentially, since the knowledge of triangular externalism is not as “direct” and “immediate” as the knowledge of our own mental states. Hence, a priori knowledge of triangular externalism must depend on philosophical reflection: thinking and reflecting about the nature of mental content, one may formulate the theory of triangulation. Moreover, she may take it to be a good a priori justification to think that content necessarily depends on relations with external empirical objects and other rational creatures, which are more or less as she thinks they are. To put it another way, “The agent has only to reflect on what a belief is, to appreciate that most of his basic beliefs are true” (Davidson 1983: 153). It should be stressed that Davidson is not exploiting any empirical investigation, even though further evidence could dismiss the a priori justification in favour of triangular externalism.17 Moreover, mere philosophical reflection could generate other arguments able to explain the nature of mental content, and those arguments could easily constitute evidence against his theory, or at least they could be incompatible alternatives to it. How can a choice be made between two different 17

It is worth stressing that Davidson denies that his own externalism could be known strongly a priori. For example, the process of triangulation plainly necessitates that the two triangulating creatures respond in a similar way to similar stimuli that both of them perceive as similar. The effective similarity in their responses, however, cannot be judged on a priori grounds. Hence further empirical investigation could dismiss triangular externalism.

220 accounts of mental content merely on a priori grounds? For these reasons, evidence against triangular externalism could derive both from further empirical investigation and from different a priori arguments. Now, exploiting the terminology used by (Pryor 2004), there may be the following three possibilities: – If I have strong evidence against triangular externalism, then triangular externalism must be rejected along with its anti-skeptical argument; – If I just believe or suspect that there could be such evidence against triangular externalism, that very possibility does not undermine the justification for triangular externalism, but it can actually undermine my rational commitment to believing (P1) and then prevent me from believing the conclusion of the argument; – If I do not recognize such evidence against triangular externalism, then I am justified in believing that triangular externalism is true and, since the anti-skeptical argument is valid, I am also justified in denying global skepticism about the external world. Hence, the anti-skeptical argument has only conditional validity: if triangular externalism happens to be true, then from the mere fact that we have thoughts and other propositional attitudes we can infer that there is an external world (more or less as we think it is) and there are also other rational creatures sufficiently similar to us. In fact, the general picture of knowledge presented by Davidson connects together subjective knowledge, objective knowledge, and intersubjective knowledge: none of these kinds of knowledge can actually stand alone, nor be reduced to the others. Therefore, if I have subjective knowledge, then I also have knowledge of the external world and of other minds. In particular, I can know that there is an external world, more or less as I think it is, and there are also other rational creatures like me. This conclusion may sound amazing. However, the argument can dismiss only “genuine” skeptical scenarios, namely those which admit the following possibility: I would be in a psychological state indistinguishable from my present psychological state even if the external world has not existed at all or if it has always been completely different from how I think it is. In other words,

221 Davidson’s argument applies only to global radical skepticisms like Descartes’s deceiving God or the permanent envatment. On the contrary, a temporary dream or hallucination cannot be ruled out: triangular externalism does not assure us as to anything in our particular empirical knowledge. Of course, there can be no doubt that it is not a serious problem, but another scenario may be much worse. Now, let us imagine a subject who was born and has always lived on Earth, has systematically triangulated external objects with other rational creatures, and thus has thoughts and beliefs that are largely true of Earth. Then, at a certain instant, her brain is detached from her body and immediately envatted. Following triangular externalism and the argument above, since her causal history is still intact, she has no reason to doubt the truth of her current thoughts and beliefs. In fact, Davidson’s argument does not rule out the possibility of a recent envatment, that is of an envatment carried out after an otherwise normal life made of causal systematic interactions with empirical objects and other rational creatures in the context of triangulation. Although Davidson may add that, in such a scenario, our deception cannot last long, I feel it significantly weakens his anti-skeptical argument.

Conclusion In this paper I have tried, on the one hand, to ascertain whether Davidson’s triangular externalism may be able to dismiss the objection raised by the Reductio Argument and, on the other, to test the anti-skeptical power of his very externalism. More precisely, I argued that the best way to demonstrate that Davidson’s theory is immune to the objection raised through the Reductio Argument is to attack its second premise, showing that it cannot be inferred from triangular externalism. I proposed two different reasons for this assumption: (i) triangular externalism cannot be known a priori; (ii) even if we concede the a priori knowledge of triangular externalism, such externalism does not imply the kind of entailment between someone’s beliefs and the external world necessary to make the argument run properly. Finally, I outlined some consequences of the above results for Davidson’s anti-skeptical argument. My conclusion is that, even if

222 triangular externalism is compatible with privileged self-knowledge, it is not an authentic answer to the skeptic about the external world.*

References Beebee H. (2001), “Transfer of Warrant, Begging the Question and Semantic Externalism”, The Philosophical Quarterly 51, 356-372. Boghossian P. (1997), “What the Externalist Can Know a Priori”, Proceedings of the Aristotelian Society 97, 161-175. Brewer B. (2000), “Externalism and a Priori Knowledge of Empirical Facts”, in (Boghossian and Peacocke 2000), 415-433. Brown J. (1995), “The Incompatibility of Anti-Individualism and Privileged Access”, Analysis 55, 149-156. Brown J. (1999), “Boghossian on Externalism and Privileged Access”, Analysis 59 (1), 52-59. Brown J. (2003), “The Reductio Argument and Transmission of Warrant”, in (Nuccetelli 2003), 117-130. Brown J. (2004), “Wright on Transmission Failure”, Analysis 64 (1), 5767. Brueckner A. (1992), “What an Anti-Individualist Knows a Priori”, Analysis 52, 111-118. Brueckner A. (2000), “Externalism and the a Prioricity of SelfKnowledge”, Analysis 60 (1), 132-136. Brueckner A. (2002), “Anti-Individualism and Analyticity”, Analysis 62 (1), 87-91. Brueckner A. (2003), “Two Transcendental Arguments Concerning SelfKnowledge”, in: (Nuccetelli 2003), 185-200. Brueckner A. (2004a), “Brewer on the McKinsey Argument”, Analysis 64 (1), 41-43. Brueckner A. (2004b), “Strategies for Refuting Closure for Knowledge”, Analysis 64 (4), 333-335. Burge T. (1979), “Individualism and the Mental”, Midwest Studies in Philosophy 4, 73-122. *

I wish to thank Mario De Caro for reading and commenting a draft of this paper.

223 Burge, T. (1982), “Other Bodies”, in (Woodfield 1982), 97-121. Davidson D. (1982), “Rational Animals”, reprinted in (Davidson 2001), 95-106. Davidson D. (1983), “A Coherence Theory of Truth and Knowledge”, reprinted in (Davidson 2001), 137-153. Davidson D. (1987), “Knowing One’s Own Mind”, reprinted in (Davidson 2001), 15-38. Davidson D. (1988), “The Myth of Subjective”, reprinted in (Davidson 2001), 39-52. Davidson D. (1990), “Epistemology Externalised”, reprinted in (Davidson 2001), 193-204. Davidson, D. (1999), “Reply to Stroud”, in (Hahn 1999). Davies M. (1998), “Externalism, Architecturalism, and Epistemic Warrant”, in (Wright et. al. 1998), 321-362. Davies M. (2000), “Externalism and Armchair Knowledge”, in (Boghossian and Peacocke 2000), 384-414. Davies M. (2003), “The Problem of Armchair Knowledge”, in (Nuccetelli 2003), 23-56. Ebbs G. (2003), “A Puzzle About Doubt”, in (Nuccetelli 2003), 143-168. Falvey K. (2000), “The Compatibility of Anti-Individualism and Privileged Access”, Analysis 60, 137-142. McKinsey M. (1991), “Anti-Individualism and Privileged Access”, reprinted in (Ludlow & Martin 1998). McKinsey M. (1997), “Accepting the Consequences of AntiIndividualism”, Analysis 54, 124-128. McKinsey M. (2002), “Forms of Externalism and Privileged Access”, Philosophical Perspectives 16, 199-224. McKinsey M. (2003), “Transmission of Warrant and Closure of Apriority”, in (Nuccetelli 2003), 97-116. McLaughlin, B. (2003), “McKinsey’s Challenge, Warrant Transmission and Skepticism”, in (Nuccetelli 2003). McLaughlin, B. and Tye, M. (1998), “Externalism, Twin-Earth, and SelfKnowledge”, in (Wright et. al. 1998), 285-320. Miller R. (1997), “Externalist Self-Knowledge and the Scope of the a Priori”, Analysis 57, 67-75.

224 Noonan H.W. (2000), “McKinsey-Brown Survives”, Analysis 60 (4), 353356. Noordhof P. (2004), “Outsmarting the McKinsey-Brown Argument”, Analysis 64 (1), 48-56. Nuccetelli S. (1999), “What an Anti-Individualist Cannot Know a Priori”, Analysis 59 (1), 48-51. Peacocke C. (1996), “Our Entitlement to Self-Knowledge: Entitlement, Self-Knowledge, and Conceptual Redeployment”, Proceedings of the Aristotelian Society 96, 117-158. Pritchard D. (2002), “Mckinsey Paradoxes, Radical Scepticism, and the Transmission of Knowledge across Known Entailments”, Synthese 130, 279-302. Pryor J. (2004), “What’s Wrong with Moore’s Argument?” reprinted in (Sosa and Villanueva 2004). Putnam H. (1975), “The Meaning of ‘Meaning’”, in: K. Gunderson (ed.), Language, Mind and Knowledge, University of Minnesota Press, Minneapolis. Raffman D. (1998), “First-Person Authority and the Internal Reality of Beliefs”, in (Wright et. al. 1998), 363-370. Sawyer S. (1998), “Privileged Access to the World”, Australasian Journal of Philosophy 76, 523-533. Warfield T.A. (1998), “A Priori Knowledge of the World: Knowing the World by Knowing Our Minds”, Philosophical Studies 92, 127-147. Warfield T.A. (2004), “When Epistemic Closure Does and Does Not Fail: A Lesson from the History of epistemology”, Analysis 64 (1), 35-41. Wright C. (2000), “Cogency and Question-Begging: Some Reflections on Mckinsey’s Paradox, and Putnam’s Proof”, Philosophical Issues 10, 140-163. Wright C. (2003), “Some Reflections on the Acquisition of Warrant by Inference”, in (Nuccetelli 2003), 57-78.

LIST OF CONTRIBUTORS

M. Cristina AMORETTI is Research Fellow at the University of Genoa (Italy). Andrea BOTTANI is Professor of Theoretical Philosophy at the University of Bergamo (Italy). Mario DE CARO is Associate Professor of Moral Philosophy at the University of Roma Tre (Italy) Pascal ENGEL is Professor of Contemporary Philosophy at the University of Geneva (Switzerland). Jennifer HORNSBY is Professor of Philosophy at Birkbeck, University of London (UK), and a director of the rational agency section of the Centre for the Study of Mind in Nature, University of Oslo (Norway). Michele MARSONET is Professor of Philosophy of Science at the University of Genoa (Italy). Massimo PIATTELLI PALMARINI is Professor of Cognitive Science at the University of Arizona (USA) Eva PICARDI is Professor of Philosophy of Language at the University of Bologna (Italy). Mark SAINSBURY is Professor of Philosophy at the University of Texas (USA). Nicla VASSALLO is Professor of Theoretical Philosophy at the University of Genoa (Italy).

EpistemischeStudien

MichaelEsfeld • StephanHartmann • MikeSandbothe (Eds.)

Vol. 1

Principles of Truth Volker Halbach / Leon Horsten ISBN 3-937202-45-5

Vol. 6

Wissenschaftlicher Realismus Christian Suhm Eine Studie zur RealismusAntirealismus-Debatte in der neueren Wissenschaftstheorie ISBN 3-937202-48-X ca. 250 Seiten Hardcover € 69,00

2. Aufl., 228 pp Paperback € 49,00

Vol. 2

Theoriebeladenheit und Objektivität Zur Rolle der Beobachtung in den Naturwissenschaften Matthias Adam ISBN 3-937202-11-0 274 Seiten Hardcover € 59,00

Vol. 7

Deskriptive oder normative Wissenschaftstheorie? Bernward Gesang ISBN 3-937202-69-2 ca. 240 Seiten Hardcover € 76,00

Vol. 3

Was ist wirklich? Neuere Beiträge zur Realismusdebatte in der Philosophie Christoph Halbig / Christian Suhm ISBN 3-937202-28-5 446 Seiten / Paperback € 32,00

Vol. 8

Persistence Through Time, and Across Possible Worlds Jiri Benovsky ISBN 3-937202-99-4 281 pp. / Hardcover, € 84,00

Vol. 4

Pragmatisch denken André Fuhrmann / Erik J. Olsson ISBN 3-937202-46-3 321 Seiten Hardcover € 84,00

Vol. 9

Naturalismus Thomas Sukopp Kritik und Verteidigung erkenntnistheoretischer Positionen ISBN 3-938793-13-9 348 pp. / Hardcover, € 84,00

Vol. 5

Sprache, Einstellung und Rationalität Pedro Schmechtig Eine Untersuchung zu den Rationalitätsbedingungen von Einstellungs-Zuschreibungen ISBN 3-937202-56-0 330 Seiten / Hardcover € 89,00 Frankfurt • Paris • Lancaster • New Brunswick P.O. Box 1541 • D-63133 Heusenstamm bei Frankfurt www.ontosverlag.com • [email protected] Tel. ++49-6104-66 57 33 • Fax ++49-6104-66 57 34

ontos verlag International Publisher P.O. Box 15 41 63133 Heusenstamm Tel. ++49 6104 66 57 33 Fax ++49 6104 66 57 34 [email protected] www.ontosverlag.com

EpistemischeStudien

MichaelEsfeld • StephanHartmann • Albert Newen (Eds.)

Vol. 10

Non-locality and Possible Worlds Tomasz F. Bigaj ISBN 978-3-938793-29-9

Vol 13

286pp. Hardcover € 89,00

Vol. 11

Reductionism in the Philosophy of Science Christian Sachse ISBN 978-3-938793-46-6 330pp. Hardcover € 89,00

Vol. 12

Von der Physik zur Metaphysik Physikalische Vereinheitlichung und Stringansatz ISBN 978-3-938793-47-3 394 Seiten Hardcover € 119,00

ontos verlag International Publisher P.O. Box 15 41 63133 Heusenstamm Tel. ++49 6104 66 57 33 Fax ++49 6104 66 57 34 [email protected] www.ontosverlag.com

Frankfurt • Paris • Lancaster • New Brunswick P.O. Box 1541 • D-63133 Heusenstamm bei Frankfurt www.ontosverlag.com • [email protected] Tel. ++49-6104-66 57 33 • Fax ++49-6104-66 57 34

Der blinde Fleck Kognitive Fehler in der Wissenschaft und ihre evolutionsbiologischen Grundlagen Ulrich Frey ISBN 978-3-938793-51-0 331 Seiten Hardcover € 89,00