The Philosophy of Hilary Putnam (Philosophical Topics, Volume 20, Number 1, Spring 1992)

, Unindexed Front Matter , Philosophical Topics, 20:1 (1992:Spring) p.0 THE PHILOSOPHY OF HILARY PuTNAM Issue Editor Ch...

Author: Christopher S. Hill (ed)

242 downloads 1063 Views 8MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

, Unindexed Front Matter , Philosophical Topics, 20:1 (1992:Spring) p.0

THE PHILOSOPHY OF HILARY PuTNAM Issue Editor Christopher S. Hill Contributors David L. Anderson • Akeel Bilgrami Noam Chomsky • James Conant • Burton Dreben • Gary Ebbs Richard Healey • Gerald J. Massey • John McDowell Richard W. Miller • Hilary Putnam • Alan Sidelle

Volume 20

Numberl


PHILOSOPHICAL TOPICS FOUNDING EDITOR: Robert W. Shahan

Address correspondence not pertaining to subscriptions to:

PHILOSOPHICAL TOPICS

Department of Philosophy 318 Old Main University of Arkansas Fayetteville, AR 72701 © Copyright 1993 by the Board of Trustees of the University of Arkansas. Published by The University of Arkansas Press, Fayetteville, Arkansas.


Philosophical Topics VOLUME 20, NUMBER 1 SPRING 1992

THE PHILOSOPHY OF HILARY PUTNAM


Contents

1. Realism and Rational Inquiry Gary Ebbs

2. Putnam on Mind and Meaning

1

35

John McDowell

3. What Is Realistic about Putnam's Internal Realism?

49

David L. Anderson 4. Realism without Positivism Richard W. Miller

5. The Search for Logically Alien Thought: Descartes, Kant, Frege, and the Tractatus

85

115

James Conant

6. Chasing Quantum Causes: How Wild Is the Goose?

181

Richard Healey 7. Explaining Language Use

205

Noam Chomsky

8. Can Externalism Be Reconciled with Self-Knowledge? Akeel Bilgrami

233


9. Identity and the Identity-like

269

Alan Sidelle 10. Putnam, Quine-and the Facts Burton Dreben

293

11. The Indeterminacy of Translation: A Study in Philosophical Exegesis

317

Gerald J. Massey

12. Replies Hilary Putnam

347

Ebbs, Gary, Realism and Rational Inquiry , Philosophical Topics, 20:1 (1992:Spring) p.1

PHILOSOPHICAL TOPICS VOL. 20 NO. I, SPRING 1992

Realism and Rational Inquiryl

Gary Ebbs University of Pennsylvania

Hilary Putnam's wide-ranging contributions to philosophy over the past thirty-five years apparently reflect many different points of view. Even careful readings of his work can leave one with the impression that there is no single set of underlying principles from which his many arguments flow. Some would say that he is like Isaiah Berlin's fox-volatile and brilliant, as opposed to steady and systematic; focused on particular issues, not unifying principles; light-footed in his approach, and quick to pursue a new interest.2 This perception of Putnam is partly sustained by the standard view that his work falls into two main periods, roughly delimited by "Realism and Reason," his 1976 Presidential Address to the Eastern Division of the American Philosophical Association. 3 In the first period, it is generally supposed, Putnam was a paradigm metaphysical realist, who held that truth is "radically non-epistemic" and argued vigorously against all views which tie the contents of our statements to our methods of verifying or falsifying them. In the second period, Putnam is supposed to have abandoned metaphysical realism to embrace internal realism, a view which implies that the contents of our statements are tied to our methods of verifying or falsifying them. Thus it is commonly believed that Putnam turned his back on a nonepistemic conception of truth and adopted a kind of verificationism. His tum towards internal realism is classified as just one, albeit the most dramatic, of many changes in his philosophical beliefs. 1


In addition, there is a temptation to see the supposed metaphysical realism of Putnam's first period as a kind of scientific naturalism, according to which only the sciences can provide us with true descriptions of the world. Putnam's causal "theory" of reference must then be viewed as a preliminary sketch of a rigorous scientific theory of reference. When viewed this way, the causal "theory" is of no value to us if it cannot be made scientifically precise. Putnam's later rejection of naturalistic theories of reference and rationalitY then amounts to an abandonment of his causal "theory" of reference, and must be seen as another fundamental change in view. These interpretations of Putnam's work have some support in his writings, and he seems at times to accept them himself. But I see more continuity in Putnam's thinking than these standard views acknowledge. I will argue that his work is motivated by one underlying philosophical project, which is at odds with both metaphysical realism and scientific naturalism. The starting point for this project is that our participation in everyday and scientific linguistic practices subjects our statements to norms that determine what we are talking about, when we agree, and when we disagree. Putnam's project is to clarify our implicit, practice-based understanding of these norms. He introduces his causal picture of reference in order to deepen our understanding of the norms underlying our linguistic practices, not to provide a metaphysical or scientific foundation for them. Putnam's project leads to the rejection of metaphysical realism. According to his causal picture of reference, there is an essential interdependence between reference and belief. This shows that our conception of the entities we refer to is not independent of the substantive beliefs we rely on in our inquiries. And it implies that we can make no sense of metaphysical realism. The interdependence of our ontological notions and our substantive beliefs is implicit in Putnam's early arguments against logical positivism, and it lies at the heart of internal realism. This interdependence goes hand in hand with the interdependence of our concepts of truth and rational acceptability. But this does not mean that internal realism is a verificationist view according to which truth is defined in terms of justification or rational acceptability. Putnam's project is also at odds with scientific naturalism. Although Putnam at times shows an interest in developing a scientific theory of reference, he does not believe that his causal picture of reference is valuable only if it can be made scientifically precise. His interest in developing a causal theory of reference stems from his desire to clarify the norms underlying our linguistic practices, and does not reflect a commitment to scientific naturalism. So Putnam's rejection of scientific theories of reference and rationality does not mark a fundamental change in his philosophical views. The view I present here has never been fully articulated by Putnam himself, and it is perhaps incompatible with some of what he has written. My

2


goal is to understand why I find some of Putnam's arguments convincing, not to provide a definitive interpretation of his work. I think of the following as a creative reconstruction of Putnam's views, which merits consideration whether or not it is compatible with all of his writings.

1. CARNAP'S VIEW OF RATIONAL INQUIRY Putnam's early views developed in reaction to logical positivism. In particular, his criticism of the analytic-synthetic distinctionS can be fully appreciated only against the background of Carnap' s conception of rational inquiry. Carnap was centrally interested in clarifying the epistemology of the natural sciences and of rational inquiry in genera1.6 The starting point for Carnap's understanding of rational inquiry is displayed in his attitude towards the apparent disputes and controversies found in traditional metaphysics. In his "Intellectual Autobiography," Carnap writes: Even in the pre-Vienna period, most of the controversies in traditional metaphysics appeared to me sterile and useless. When I compared this kind of argumentation with investigations and discussions in empirical science or in the logical analysis of language, I was often struck by the vagueness of the concepts used and by the inconclusive nature of the arguments. I was depressed by disputations in which the opponents talked at cross purposes; there seemed hardly any chance of mutual understanding, let alone of agreement, because there was not even a common criterion for deciding the controversy .... I came to hold the view that many theses of traditional metaphysics are not only useless, but even devoid of cognitive content. 7

This attitude towards apparent disputes in traditional metaphysics reflects a conception of rationality central to Camap's philosophical project. According to this conception, agreement or disagreement between investigators is possible only if they share criteria for determining whether their judgments are correct or incorrect. If they do not share such criteria, then they cannot be genuinely agreeing or disagreeing, even if they appear to be. On Carnap's view the controversies and questions in traditional metaphysics fail to be genuine because there are no criteria for deciding them. Carnap also held that investigators will fail to agree or disagree if they each have different criteria for assessing their judgments. In this case, although the investigators may succeed in raising questions and making meaningful claims, they cannot agree or disagree with each other. Insofar as they do not share a criterion for rational inquiry, they are not really communicating at all. This kind of failure to agree or disagree presupposes that there is more than one criterion of rational inquiry. Unlike Frege and Russell,

3


Camap from the start accepted that there are many different criteria for rationally assessing judgments. He came to believe that the sharing of these criteria must be understood as the sharing of linguistic frameworks. Through a clarification and analysis of the rules for the correct use of linguistic expressions, he believed, we can make precise, for each genuine case of rational agreement or disagreement, the particular linguistic framework which underlies it. Motivated by this framework-relative conception of rational inquiry, Camap's philosophical project was to clarify our understanding of the criteria relative to which inquiry is possible by offering rigorous descriptions of the rules governing the correct use of linguistic expressions. For Camap, an explicit specification of one complete set of such rules constitutes a clarification of a particular linguistic framework. He believed that our understanding of the structure of rational inquiry is deepened as we generalize about the relationships between the rules of linguistic frameworks, and the criteria for detennining the correctness or incorrectness of statements made within particular frameworks. The two most important of these relationships are captured by Camap' s distinction between analytic and synthetic statements. On Camap's view, a statement is analytic if its correctness or incorrectness is detennined solely by the rules of the linguistic framework within which it is made; it is synthetic if its correctness or incorrectness is detennined by those rules only in conjunction with empirical investigations. This classification was meant to be exhaustive: every statement in every linguistic framework is either analytic or synthetic. This distinction between two different ways in which a statement can be correct lies at the very heart of Camap's analysis of the structure of rational inquiry. Camap sharply distinguishes between changes of belief made within particular linguistic frameworks and changes in linguistic framework. The former may involve an element of free choice, but do not involve changes in the rules for detennining the correctness or incorrectness of statements, whereas the latter do involve changes in those rules. Because changes in linguistic framework do not take place within any particular linguistic framework, they cannot be rational or irrational. Camap therefore endorsed a Principle of Tolerance, according to which investigators are free to adopt any linguistic framework which suits their purposes. 8 Investigators may change the frameworks they use as much as it suits them. But they will not succeed in agreeing or disagreeing with each other unless they are working within the same linguistic framework. Only then will they share criteria for deciding the correctness or incorrectness of their statements. Since on Camap' s view rational inquiry is not possible outside of a particular linguistic framework, we have no framework-independent conception of facts or objects. No absolute sense can be made of statements that facts or objects exist. Insofar as they have genuine content, such statements always

4


presuppose a precise criterion of correctness or incorrectness. We cannot even individuate a statement without specifying a linguistic framework. For Camap there simply are no intelligible statements, whatever their subject matter, which are not made from within particular linguistic frameworks.

2. PUTNAM'S CRITIQUE OF THE ANALYTIC-SYNTHETIC DISTINCTION In "The Analytic and the Synthetic" Putnam rejects Camap's model of meaning on the grounds that it misrepresents the norms underlying our rational inquiries. For Putnam, our only grasp of these norms is based in our participation in ongoing commonsense and scientific linguistic practices. His strategy is to use our implicit grasp of the norms underlying these practices to criticize the analytic-synthetic distinction. Despite our incomplete understanding of our linguistic practices, we often have confidence in our pretheoretical judgments about when two investigators are talking about the same thing, when they agree, and when they make claims which cannot both be true. Such judgments are not infallible, but on Putnam's view they are good guides to the norms underlying our linguistic practices. Using carefully chosen examples, Putnam argues that the analytic-synthetic distinction is incompatible with some of our most confident pre-theoretical judgements about when investigators make incompatible claims about the same subject. Along the way he begins to develop a picture of meaning which more accurately describes the norms underlying our everyday and scientific inquiries. Let us look at Putnam's argument against the analytic-synthetic distinction in detail, focusing on just one of his examples. He observes that before Einstein, scientists thought that kinetic energy was correctly described by a particular equation. After Einstein's development of the theory of relativity, kinetic energy must be described by a different equation. The adoption of the theory of relativity is taken by Carnap, and the logical positivists in general, to be a paradigm of change in linguistic framework. Since the two kinetic energy equations are stated in different linguistic frameworks, they cannot make conflicting claims about the same form of energy. Thus on Camap's view, the adoption of the new kinetic energy equation must be viewed as a change in the meaning of the term "kinetic energy," not a change in the scientists' beliefs about a single form of energy. The trouble with this view of the kinetic energy equations, according to Putnam, is that it does not accurately describe our pre-theoretical judgments about the case. Scientists confidently judge that the two equations describe the same form of energy, and that those who accept the relativistic equation for kinetic energy disagree with those who held the pre-relativistic equation

5


for kinetic energy. But if we assume with Camap that the equations are made within different linguistic frameworks, we will conclude that they do not describe the same form of energy, and that the scientists who accept the second equation do not disagree with the scientists who accepted the first one. So if we take our pre-theoretical judgments to be good guides to the norms underlying rational inquiry, we must conclude that Carnap's view is wrong.9 Putnam observes that the obvious way to make sense of the judgment that the later scientists disagree with the earlier ones is to see the term "energy" as referring to the same quantity in both equations. Indeed, the judgment that "energy" has the same reference in both equations seems to go hand in hand with the judgment that those who accept the relativistic equation disagree with those who held the pre-relativistic equation. As a step towards making sense of judgments like these, Putnam begins to develop a new picture of the semantic role of terms like "energy". These are what he calls "law-cluster terms". They occur in statements of many scientific laws, none of which is essential to our identification of the concepts they express. One or more of the laws in which a law-cluster term occurs may be given up without changing the concept expressed by that term. Once we view "energy" as a law-cluster term, we can accept that the reference of the term "energy" did not change when the old equation was given up, and that the earlier equation attributed a property to the quantity energy which we do not now attribute to that quantity. So this picture of the role of terms like "energy" clarifies and further supports our pre-theoretical judgments that the two equations describe the same form of energy, and that the scientists who accept the later equation disagree with those who accepted the earlier one. lO

3. PUTNAM'S VIEW OF RATIONAL INQUIRY The kinetic energy example provides just one simple illustration of Putnam's approach, and the notion of a law-cluster term offers only the barest outlines of a model of meaning that helps to clarify some of our pre-theoretical judgments about when investigators are talking about the same things, when they are agreeing, and when their claims are incompatible. But we are already in a position to make a few observations about the general shape of Putnam's view. Like Camap, Putnam's project is to deepen our understanding of rational inquiry. But unlike Carnap, Putnam takes seriously the judgments we make as participants in ongoing everyday and scientific inquiries; in particular, our judgments about when two investigators are talking about the same thing, and when their claims are incompatible. Camap's model of meaning, if we were to accept it, would misrepresent the norms of our actual practices and would undermine our confidence in these basic judgments. So Putnam

6


urges that we reject Carnap's model, and replace it with one which more accurately characterizes our actual linguistic practices. 11 Putnam's notion of law-cluster terms reflects his view that the norms of rational inquiry are open-ended. He does not begin with a fixed idea of what kinds of criteria two investigators must share in order to agree or disagree. Instead he starts with our pre-theoretical judgments, and develops a model of meaning which helps us to make sense of and clarify those judgments. It turns out that if we want to make sense of our actual linguistic practices, we must acknowledge that the norms underlying these pre-theoretical judgments are more flexible than the logical positivists thought. We often judge that two investigators are making incompatible claims about the same thing even when they do not agree about what procedures to follow in order to determine whose claim is correct. The scientist who accept the relativistic kinetic energy equation have very different grounds for their belief than those who accepted the earlier equation. By Carnap's standards, they do not share a criterion for determining which of these equations is correct. But once we start with the pre-theoretical judgment that the scientists who accept the later equation disagree with those who accept the earlier one, we see that Carnap's conception of agreement and disagreement is too strict. In its place we do not try to state a new set of necessary and sufficient conditions for agreement or disagreement. Instead, we require that our description of the nonns underlying our practices fit with and help us make sense of our pretheoretical judgments about when two investigators are talking about the same thing, when they agree or disagree. A philosophical account of the nonns underlying rational inquiry which meets this requirement will be as open-ended as the ordinary and scientific inquiries in which our pretheoretical judgments about reference, agreement, and disagreement are made. I mentioned earlier that according to Carnap we have no frameworkindependent conception of facts or things. From Carnap's perspective, then, one might object that Putnam's notion of a law-cluster term is obscure. Putnam supposes that in many of the statements in which the term "energy" occurs, it refers to the same quantity, even if the criteria for assessing the correctness or incorrectness of those statements are quite different. But this account of the semantic role of the tenn "energy" presuppo~es that we have a framework-independent conception of energy. Since according to Carnap we have no such conception, we can't make sense of Putnam's notion of a law-cluster term. 12 This objection masks an important similarity between their views. Even though Putnam and Carnap have very different views about how to characterize the nonns of rational inquiry, they both believe that we have no understanding of statements, facts, or things independent of those norms. As we shall see, on Putnam's view our conceptions of the entities to which our

7


terms refer are not available prior to or independently oft 3 the norms which underlie our judgments about when investigators are talking about the same thing, when they agree, and when they make incompatible claims. Putnam's introduction of the notion of a law-cluster concept is intended to help clarify our implicit understanding of those norms, not to provide a metaphysical foundation for them. So although Putnam has a more flexible, practicebased conception of the norms of rational inquiry, he also is implicitly committed to the Camapian idea that our understanding of statements, facts, and things is based in the norms underlying rational inquiry, and cannot be fully detached from them. This aspect of Putnam's view has far-reaching consequences, as we shall see later.

4. PUTNAM'S CRITICISM OF THE ANALYTIC-SYNTHETIC DISTINCTION CONTRASTED WITH QUINE'S I have been emphasizing that for Putnam our participation in ongoing inquiries subjects our statements to norms which it is the task of philosophy to describe and cl¢fy. So Putnam's criticism of the analytic-synthetic distinction ultimately rests on our perspective as participants in everyday and scientific inquiries. This aspect of Putnam's project is best appreciated when contrasted with the view of language which underlies Quine's criticism of Camap's analytic-synthetic distinction. For Camap, as we have seen, genuine agreement or disagreement between investigators is possible only if they share criteria for assessing the correctness or incorrectness of their statements. Camap believed that the sharing of these criteria must be understood as the sharing of linguistic frameworks, which consist in precise and determinate rules for the correct use oflinguistic expressions. On Camap's view, a statement is analytic if its correctness or incorrectness is determined solely by the rules of the linguistic framework within which it is made; it is synthetic if its correctness or incorrectness is determined by those rules only in conjunction with particular empirical investigations. Quine's criticism of this conception of rational inquiry begins with his requirement that there be an objective scientific basis for the attribution of a particular linguistic framework to an investigatorY On Camap's view, one investigator can legitimately correct or criticize the statements made by another investigator only if they are both working within the same linguistic framework. If they are not working within the same linguistic framework, they do not share criteria for assessing their statements, and so neither investigator is in a position to evaluate the statements of the other. So, Quine reasons, on Camap's view intersubjective rational criticism is not possible

8


unless there is an objective basis for the attribution of a particular linguistic framework to an investigator. If there is no way to determine objectively which linguistic framework an investigator is using, there is no way to determine objectively whether or not his claims are correct. Quine is a scientific naturalist. He applies his naturalism to the question of whether there is an objective basis for attribution of linguistic frameworks to investigators. QUine's naturalism is implicit in "Two Dogmas of Empiricism,"15 his celebrated attack on the analytic-synthetic distinction. In this paper Quine rejects Carnap's attempts to clarify the distinction between analytic and synthetic statements on the grounds that they all make use of terms, like "synonymy" and "semantical rule", which are just as obscure as the analytic-synthetic distinction itself. Quine's criticisms in "Two Dogmas of Empiricism" are often misunderstood, because he does not say what kind of clarification would satisfy him. In retrospect we can see that for Quine, Carnap's use of semantical terms is legitimate only if there is an objective naturalistic basis for using those terms to characterize an investigator's linguistic behavior. Quine's central criticism is that there is no such basis for using semantical terms in this way. From his naturalistic point of view, these terms are literally without any determinate application. This is exactly how he puts the point in chapter two of Word and Object, where he presents his thesis that translation is indeterminate. 16 Here Quine's naturalism is fully explicit. His aim is to show that such notions as meaning, semantical rule, synonymy, and analyticity cannot be understood in terms of the natural facts about language use. Quine assumes that the natural facts about language and meaning are exhausted by speakers' dispositions to respond to queries under various prompting stimulations, described in neurophysiological terms. From within this naturalistic picture, a claim about which linguistic framework a speaker is using is objective only if it can be understood as determined by the physical behaviors which underlie the speaker's use of his words. Thus, Quine writes, . . . there are no meanings, nor likenesses nor distinctions of meaning, beyond what are implicit in people's dispositions to overt behavior. For naturalism the question whether two expressions are alike or unlike in meaning has no determinate answer, known or unknown, except in so far as the answer is settled in principle by people's speech dispositions, known or unkpown.1f by these standards there are indeterminate cases, so much the worse for the terminology of meaning and likeness of meaning. 17

In chapter 2 of Word and Object, Quine argues that what a speaker "means" by her words is not fully determined by the facts about how she uses them. Compatibly with all the facts about her linguistic behavior, a speaker's utterances can be translated in a number of "inequivalent" ways. So, Quine argues, Carnap's distinction between utterances whose meanings are analytic, and

9


those whose meanings are synthetic, cannot be understood in naturalistic terms. Quine concludes that this distinction, and the related notion of a linguistic framework, must be abandoned. On Quine's view Camap's conception of rational inquiry is all make-believe, based on distinctions which have no objective basis. From this brief sketch we can see that Quine's criticism of the analyticsynthetic distinction is fundamentally different from Putnam's. The most basic difference is that Quine does not take our participation in ongoing rational inquiries at face value. Instead, he treats an investigator's use of a language as an object of scientific investigation. His requirement that there be an objective basis for the attribution of a linguistic framework to an investigator reflects this perspective; his naturalistic description of language use deepens and clarifies it. In contrast, on Putnam's view we do not require a scientific foundation for our linguistic practices of agreeing or disagreeing with one another. Our participation in these practices is all the foundation we need. Philosophy can help us to clarify the norms which we must acknowledge as participants in rational inquiries, but it should not try to provide us with a foundation for our judgments as to when we are talking about the same things, and when we agree or disagree. If we begin with the proper perspective on language use, by taking our participation in rational inquiries at face value, we will not be tempted to ask Quine's questions, or to accept his answers to them. 18 This difference between Putnam and Quine has consequences which I will explore further below.

5. THE INTERDEPENDENCE OF REFERENCE AND BELIEF I noted earlier that despite fundamental differences, there are important similarities between Camap's and Putnam's views of rational inquiry. In particular, Putnam is implicitly committed to the Camapian idea that our understanding of statements, facts, and things is grounded in the norms underlying rational inquiry. This aspect of Putnam's approach has been widely misunderstood. One reason for this misunderstanding is that interpreters have focused on Putnam's causal picture of the references of naturalkind words. When Putnam's discussions of natural-kind words are read out of context, it can seem as though on Putnam's view there is a conception of natural kinds which is available independently of the norms which underlie our inquiries. It is then natural to think that he is trying to answer the question of how we succeed in referring to natural kinds, conceived in that independent way. It is generally supposed that Putnam's answer to this question is the causal "theory" of reference: our natural-kind words refer to natural kinds, conceived independently of our scientific inquiries, in virtue of causal

10


relations we bear to those kinds. But this interpretation, though initially tempting, is incorrect. For on Putnam's view we have no conception of facts or things which is prior to, or available independently of, the norms underlying our actual commonsense and scientific inquiries. This is reflected by the fact that on his view there is an essential interdependence of our concepts of reference and belief. Putnam's causal picture of reference is a development of his notion of law-cluster terms. He takes another step towards the causal picture in "Dreaming and Depth Grammar," where he criticizes Norman Malcolm's criterial view of meaning. Here is one of Putnam's most compelling counterexamples to the criterial view: Consider the following case: there is a disease, multiple sclerosis, which is extremely difficult to diagnose. The symptoms resemble those of other neurological diseases; and not all of the symptoms are usually present. Some neurologists believe that multiple sclerosis is caused by a virus, although they cannot presently specify what virus. Suppose a patient, X, has a 'paradigmatic' case of multiple sclerosis. Then Malcolm's view is that, no matter what we find out later, X has multiple sclerosis because that is what we presently mean. In particular, if we later identify a virus as the cause of multiple sclerosis, and this patient's condition was not caused by that virus, he still had multiple sclerosis.19

Malcolm's view does not leave room for scientific investigation and discovery, for what "we could find out later." So it does not accurately describe the norms which underlie the neurologists' use of statements in which the term 'multiple sclerosis' occurs. If we are to describe these norms accurately, ... we should have to say that we reject the view that scientists who accept our hypothetical (future) virological criterion are talking about a different disease when they use the term 'multiple sclerosis' .20

An accurate picture of the meaning of 'multiple sclerosis' would enable us to see how the discovery of the virus that causes mUltiple sclerosis leads to a change in our beliefs about multiple sclerosis, not a change in the reference of the term 'multiple sclerosis', or in the concept it expresses. As a step towards developing such an account, Putnam suggests th~t we think of the reference of 'multiple sclerosis' in the following way: ... there is (we presume) in the world something-say, a viruswhich normally causes such-and-such symptoms. Perhaps other diseases occasionally (rarely) produce these same symptoms in a few patients. When a patient has these symptoms, we say he has 'multiple sclerosis' -but, of course, we are prepared to say that we were mistaken if the etiology turns out to have been abnorma1. 21

11


Putnam's suggestion is that the reference of 'multiple sclerosis' is the disease which is causally responsible for certain symptoms. Since the cause of the symptoms remains constant, the reference of 'multiple sclerosis' does not change when the neurologists discover that multiple sclerosis is caused by a virus. This helps to clarify our practice-based judgment that the neurologists' discovery leads to a change in belief, and not a change in the meaning of 'multiple sclerosis'. Putnam's proposal that the reference of 'multiple sclerosis' is the disease which is causally responsible for certain symptoms, together with the view of "law-cluster concepts" he introduced in "The Analytic and the Synthetic," points towards a new way of thinking about reference. This new picture of reference clarifies our implicit understanding of the norms for agreement or disagreement in our linguistic practices. In the context of Putnam's philosophical project, further clarifications of these norms can only result from detailed investigations of our practice-based judgments about reference, elicited by descriptions of various actual and possible linguistic situations. The most dramatic development in Putnam's pichlre of reference comes in "The Meaning of 'Meaning' ," where he presents thought experiments which suggest that the references of our terms depend on the nature of our physical environments. 22 The best-known of these thought experiments involves the natural kind water. Putnam imagines that there exists a planet called Twin Earth, the same as Earth in all ordinarily discernible ways, but different from Earth in one crucial respect: where there is water on Earth, there is another substance, twinwater, on Twin Earth. Twinwater is in all ordinary contexts indistinguishable from water, but it has a fundamentally different molecular structure, XYZ. My twin, a molecule-for-molecule duplicate of me, lives on Twin Earth. He has had the same kind of linguistic conditioning, and his neuro-physiological dispositions are just like mine. Putnam observes that even if neither of us knows the molecular structure of the stuff we respectively call 'water', our words refer to different natural kinds. When my twin uses the word 'water', he is speaking about twinwater, not water. He refers to that kind of stuff to which he typically applies his word 'water'. It does not matter that he is unaware that the molecular structure of twinwater is XYZ. His term still refers to twinwater, not water. Similarly, my word 'water' refers to water, not twinwater, even if I am ignorant of its molecular structure. The reference of my word 'water' is different from the reference of my twin's word 'water' because he and I typically apply the expression 'water' to different liquids. On the basis of examples and thought experiments like these, it is tempting to think that on Putnam's view we have a conception of natural kinds like water which is prior to, and available independently of, the norms underlying our commonsense and scientific inquiries. This is reflected in our natural

12


temptation to try to adopt a detached perspective on linguistic behavior, viewing it from outside any particular linguistic practice. From this detached perspective, it can seem that we have a conception of natural kinds which is available independently of any linguistic practice. We do not consider the source of our own conception of those natural kinds. We just focus on the question of the relationship between the speaker's linguistic behavior and the natural kinds in the world "as it really is," and we simply assume that we are able to form a conception of those natural kinds independently of any particular linguistic practice. There are passages in which Putnam seems to take the same view of his examples. For instance, in "The Meaning of 'Meaning'" he writes: It is beyond question that scientists use tenns as if the associated criteria were not necessary and sufficient conditions, but rather approximately correct characterizations of some world of theoryindependent entities, and that they talk as if later theories in a mature science were, in general, better descriptions of the same entities that earlier theories referred to. 23

Putnam's use of the phrase "theory-independent entities" in this passage apparently suggests that he thinks that some of our ontological notions are available prior to, and independently of, the norms underlying our inquiries. This seems to support the idea that we can stand back from linguistic practice and imagine a relationship between linguistic behavior and entities which are conceived independently of any of the beliefs we hold as participants in ongoing linguistic inquiries. But this does not make sense from Putnam's point of view. Even in this passage, Putnam gives content to the phrase "theory-independent entities" by noting that scientists "talk as iflater theories in a mature science were, in general, better descriptions of the same entities that earlier theories referred to," thus linking our notion of theoryindependent entities with the normal evolution of scientific theories. Despite first appearances, Putnam's conception of "theory-independent entities" is essentially based in the norms underlying our practices. He does not endorse the tempting thought that we can step outside our linguistic practices and imagine a relationship between our language use and entities conceived independently of norms underlying any particular linguistic practice. On his view, we are always workingfrom within a given linguistic practice, trying to clarify our understanding of that practice, or another one. Our ontological conceptions are not available independently of the norms underlying the linguistic practices in which we participate. One way to appreciate this is to see that on Putnam's view there is an essential interdependence between our concepts of reference and belief. Our beliefs are individuated in part by the causal relations we bear to our environment, and the individuation of the entities to which our terms refer is dependent upon the beliefs expressed with the help of those terms.

13


Putnam's picture of reference and belief is based in our pre-theoretical understanding of norms which determine what people are talking about, and when they agree or disagree. On this picture there is no way to separate sharply the norms which determine what people are referring to from those which determine what they believe. And so there can be no clarification of reference which is not also a clarification of belief. There are several places in "the Meaning of 'Meaning'" where Putnam emphasizes features of our linguistic practices which illustrate the interdependence of our beliefs about when a term is correctly applied and the reference of the term. For example, he warns the reader not to interpret his causal picture of reference as implying that every natural-kind term must pick out a single natural kind, with one underlying microstructure. This would be to read a kind of linguistic atomism into his picture, one which does not acknowledge the interdependence between our use of a term and its reference. He writes that [aJ ... misunderstanding which should be avoided is the following: to take the account we have developed as implying that the members of the extension of a natural-kind word necessarily have a common hidden structure. It could have turned out that the bits of liquid we call 'water' had no important common physical characteristics except the superficial ones. In that case the necessary and sufficient condition for being 'water' would have been possession of sufficiently many of the superficial characteristics. 24

Putnam illustrates this possibility with the example of jade: An interesting case is the case of jade. Although the Chinese do not recognize a difference, the term 'jade' applies to two minerals: jadeite and nephrite. Chemically, there is a marked difference. Jadeite is a combination of sodium and aluminum. Nephrite is made of calcium, magnesium, and iron. These two quite different microstructures produced the same unique textural qualities !25

Here Putnam is suggesting that the necessary and sufficient condition for the correct application of the term 'jade' to some mineral stuff is that it possess sufficiently many of the appropriate textural qualities. The same reasoning applies, in certain counterfactual cases, to a natural-kind term like 'water': ... if H 20 and XYZ had both been plentiful on Earth, then we would have had a case similar to the jadeite/nephrite case: it would have been correct to say that there were two kinds of 'water'. And instead of saying that 'the stuff on Twin Earth turned out not to really be water', we would have to say 'it turned out to be the XYZ kind of water' .26

What these examples illustrate is that the references of some of what we take to be natural-kind words need not be a natural kind. In determining the

14


references of our tenns, we must investigate the things to which they are typically applied. The pattern of our applications of a word is at least as important in determining what it refers to as the microstructures of the things to which we apply it. The pattern of our applications of a word reflects our practice-based judgments about when it is correctly applied. So these examples show that in determining the content and application of our words, there is no principled way to distinguish between the contribution of our practice-based judgments about when our words are correctly applied, and the contribution of the references of our words. And this means that our conception of the references of our words is inextricably tied to the nonns underlying our rational inquiries. 27 The interdependence of belief and reference is further illustrated by Putnam's application of his new picture of reference to artifact words like 'pencil', 'chair', 'bottle', etc. He argues against the "traditional view" that these words are defined by clusters of properties. On this view, statements like pencils are artifacts would be analytically trUe, true in virtue of the cluster definition of 'pencil'. Against this, Putnam tells the following story: Imagine that we someday discover that pencils are organisms. We cut them open and examine them under the electron microscope, and we see the almost invisible tracery of nerves and other organs. We spy on them, and we see them spawn, and we see the offspring grow into full-grown pencils. We discover that these organisms are not imitating other (artificial) pencils-there are not and never were any pencils except these organisms. 28

Putnam argues that if this is conceivable, then it is not analytically true that pencils are artifacts. It could tum out that what we call 'pencils' are in fact organisms. This is like his thought experiment in "It Ain't Necessarily SO,"29 where he imagines that we discover that cats are robots controlled from Mars. In each case, he suggests, we would not say that the subject has changed. We would say that we discovered that pencils are not artifacts, and that cats are not animals: our beliefs would change, but not the references of our tenns 'pencil' and 'cat' .30 On Putnam's view there is no simple way in which the references of our words are determined. We can meaningfully refer to pencils even though they are in fact artifacts, not natural kinds. But that is not to say that a cluster of descriptions determines the meaning of our word 'pencil'. The reference of the word 'pencil' is determined in part by the nonns governing our use of that word. If, as it happens, those nonns pick out a natural kind, then that is the reference of 'pencil' . But if they do not, the word 'pencil' still has a reference, based in the nonns for its correct application in the linguistic community. Even if 'pencil' is an artifact term, it is not synonymous with any cluster of descriptions. Because of this there is no way to determine the reference of 'pencil' without also determining many of our beliefs about

15


pencils. Hence our conception of the entities we call "pencils" is not available independently of our beliefs about pencils. These examples illustrate the deep interdependence of our concepts of reference and belief in our commonsense and scientific practices. They show that the nonns underlying our linguistic practices involve simultaneous and interconnected judgments about what investigators believe, and what they are referring to. Putnam's philosophical project of clarifying our understanding of these practices requires that he acknowledge this interdependence. The community-shared nonns, together with the environment we are in, detennine the references of our words. There is no useful way to distinguish between those parts of the community-shared nonns which depend on our beliefs, and those which solely concern the references of our words. And so our conception of the entities to which our words refer is essentially tied to our beliefs about those entities. The interdependence of reference and belief shows that our ontological notions are not independent of the nonns underlying our rational inquiries. We can't step outside our linguistic practices and imagine a relationship between our language use and entities conceived independently of nonns underlying any particular linguistic practice))

6. THE DISSOLUTION OF METAPHYSICAL REALISM We are now in a position to see that implicit in Putnam's early papers on meaning is a rejection of metaphysical realism, according to which there is a conception of the way things stand in the world, completely independent of any of our beliefs. From Putnam's point of view, the trouble with metaphysical realism is that we can't make sense of it. The metaphysical realist presupposes that we are able to conceive of the world-the entities it contains and the relationships which obtain between them-independently of the nonns underlying our commonsense and scientific inquires. But we have seen that on Putnam's view our ontological notions are not available independently of the nonns underlying our rational inquiries. And if our ontological notions are not available independently of these nonns, we can't make sense of the notion of representation on which the metaphysical realist's alleged conception ofthe world depends. I assume that metaphysical realism amounts to what Bernard Williams and Thomas Nagel have called the absolute conception of the world, a conception of how things stand in the world, completely independent of any of our beliefs about it. 32 Williams and Nagel believe that we understand the absolute conception of the world as the limit of a dialectic which progresses through ever-expanding circles of representations. This dialectic begins with the assumption that our beliefs result from interactions with an independently existing world. From time to time we discover that some of our beliefs are

16


limited or distorted. The discovery of limits or distortions in our beliefs requires that we have a more encompassing representation of the world, one which includes an explanation of the cognitive processes which limited or distorted our beliefs. When this happens locally, within our commonsense or scientific representation of the world, it does not necessarily lead to a conception of how things are completely independent of any of our beliefs. But Williams and Nagel maintain that we can extend our understanding of the dialectical process of overcoming the limitations in our beliefs far beyond any of our substantive beliefs about the world. At the limit, it seems that this dialectic leads to the idea of an absolutely objective representation of the world, without any subjective elements, from which all other representations can be understood. This is the absolute conception of the world. From Putnam's point of view, the initial impression that the absolute conception makes sense dissolves under scrutiny. The reason is that it depends on the assumption that we can conceive of a complete representation of the world which is radically detached from all of our beliefs. In order to conceive of such a representation, our conception of the entities to which our words refer would have to be available independently of all of our beliefs about those entities. But we have seen that our conception of the entities to which our words refer is not available independently of all of our beliefs about them. In order to make sense of the norms governing agreement and disagreement in our commonsense and scientific practices, we must think of reference as involving causal relations with things in our environment. Our understanding of the contents of beliefs is inextricably bound up with our beliefs about the causal relations we bear to things in our environment. To accept Putnam's picture is to see the contents of beliefs as individuated in part by the references of the terms which express them. Hence there is no way to conceive of a belief, or the content of a possible belief, unless we have some idea of the social and physical environment on which its individuation depends. We have no understanding of the notion of representation apart from our understanding of the notion of the content of a (possible) belief. So we are unable to give any genuine content to the metaphysical realist's absolute conception of the world.

7. QUINE'S SCIENTIFIC NATURALISM The conclusion we have just reached is the result of thinking of ontological notions as essentially tied to the norms we are subject to as participants in our commonsense and scientific inquiries. Putnam's participant perspective on these inquiries has other important consequences as well. I noted earlier that it contrasts starkly with Quine's naturalistic perspective on language. I

17


want now to explain in more detail what is wrong with Quine's scientific naturalism from Putnam's point of view. As I mentioned above, Quine's criticism of intentional notions like meaning, reference, and belief starts with his requirement that there be an objective scientific basis for using such notions to describe an investigator's linguistic behavior. His challenge to the assumption that these notions have objective application is set out in chapter two of Word and Object. The central aim of this chapter is "to consider how much of language can be made sense of in terms of its stimulus conditions," where these are primarily exhausted by the stimulus meanings of sentences of the language. 33 The stimulus meaning of a sentence S for a given speaker A (at time t) is defined as the ordered pair consisting of the class of all irradiation patterns of the eye (and other sense modalities) which would prompt A's assent to S, and the class of all irradiation patterns which would prompt A's dissent from S. Assent and dissent are assumed to have a completely behavioral characterization. 34 The question of the determinacy of meaning and reference is raised as a question about the determinacy of translation from one language into another. Quine reasons that if the translation of sentences and words of one language into those of another is not determined by stimulus meaning, then translation is objectively indeterminate. Quine argues that the only sentences whose translations are (fairly well) determined by speech behaviors are observation sentences, and truth functions of these. 35 The references of the predicates (what he calls "terms") of the language, however, are not uniquely determined by speech behaviors. Quine's well-known example involves a native expression, 'Gavagai', which he assumes to have the same stimulus meaning as our one-word sentence, 'Rabbit'. Since stimulus meaning is defined only for sentences as wholes, the behavioral facts about the use of 'Gavagai' do not determine whether this sentence contains a predicate true of rabbits, rabbit stages, or undetached rabbit parts, to mention just three possibilities. The argument for this has two parts. First Quine points out that the stimulus meaning of 'Gavagai', taken in isolation from the stimulus meanings of other sentences of the language, does not uniquely fix what a speaker is referring to: ... a whole rabbit is present when and only when an undetached part of a rabbit is present; also when and only when a temporal stage of a rabbit is present. If we are wondering whether to translate a native expression "gavagai" as "rabbit" or as "undetached rabbit part" or as "rabbit stage," we can never settle the matter simply by ostension-that is, simply by repeatedly querying the expression "gavagai" for the native's assent or dissent in the presence of assorted stimulations. 36

Second, Quine argues that translation of the particles and constructions which make up the speaker's "apparatus of individuation"37-plural endings,

18


pronouns, numerals, the "is" of identity, the words "same" and "other"does not uniquely determine the reference of terms either. Translation of the particles and constructions of individuation involves "analytical hypotheses" which are not uniquely determined by speech dispositions. Thus even if we could decide that the native is referring to rabbits, once we worked out a translation of his apparatus of individuation, this would not show that reference is uniquely determined by speech behaviors. For there may be other acceptable translations, according to which the native is referring to undetached rabbit parts or rabbit stages: ... if one workable overall system of analytical hypotheses provides for translating a given native expression into "is the same as," perhaps another equally workable but systematically different system would translate that native expression rather into something like "belongs with." Then when in the native language we try to ask "Is this gavagai the same as that?" we could as well be asking "Does this gavagai belong with that?" Insofar, the native's assent is no objective evidence for translating "gavagai" as "rabbit" rather than "undetached rabbit part" or "rabbit stage."38

When appropriately generalized, this reasoning implies that the references of a speaker's words are not uniquely determined by her speech dispositions. Since according to Quine speech dispositions are the only facts relevant to translation, his conclusion is that the reference of a speaker's words is not uniquely determined by any facts, known or unknown. In short, reference is

inscrutable. According to Quine, the inscrutability of reference holds for our own language, as well as any foreign language which we try to translate. This means that the translation of our own words is not uniquely determined by speech dispositions. I can say "I am talking about rabbits, not rabbit stages," but there is more than one acceptable translation of that sentence. My entire language could be translated into itself in such a way that all my speech dispositions are preserved, but "rabbit" is translated as "rabbit stage". Thus on Quine's view there is simply no fact ofthe matter about what I am referring to with my word "rabbit". Of course, I don't have to translate my words, I can simply use them. I can assert, for example, that rabbits are not rabbit stages, nor are they undetached rabbit parts. This invplves what Quine calls "acquiescing in our mother tongue and taking its words at face value."39 According to Quine, we can take our words at "face value" in this way, while at the same time believing that there is no objective translation of them, even into our own language. This part of Quine's view is very difficult to understand. From Putnam's point of view, it is easy to see why. Given the interdependence of belief and reference, the inscrutability of reference implies that our beliefs can't have determinate truth conditions. And if our beliefs don't have determinate truth

19


conditions, then agreement or disagreement among investigators is not possible. So if we treat an investigator's language as an object of scientific inquiry, and judge the determinacy of reference and belief by Quine's scientific standards, we must conclude that rational inquiry is not possible: our participation in commonsense and scientific practices does not subject our statements to norms which determine when we agree and disagree. 4o This is to reject Putnam's starting point that our participation in these practices does subject our statements to norms which determine when we agree and disagree. Putnam's criticism of Quine's view of language begins with the observation that on Quine's view . . . a complete account of our understanding of our language would simply be a description of the noises we utter together with a description of the actual processes by which we produce those noises (or subvocalizations).41

As participants in commonsense and scientific inquiries, we see ourselves as subject to norms which determine when we agree, and when we disagree. Putnam's criticism is that QUine's behavioral description of our language use undermines the very possibility of agreement and disagreement: On such an account, we cannot genuinely disagree with one another: if I produce a noise and you produce the noise "No, that's wrong," then we have no more disagreed with each other than if I produced a noise and you produced a groan or a grunt. Nor can we agree with each other any more than we can disagree with each other: if I produce a noise and you produce the same noise, then this is no more agreement than if a bough creaks and then another creaks in the same way.42

So from our perspective as participants in ongoing inquiries, Quine's behavioral description of language use cannot be a complete account of our linguistic practices. To accept Quine's scientific naturalism about language, and his conclusion that reference is inscrutable, would be to abandon our perspective as participants in ongoing linguistic practices. From Putnam's point of view, to abandon this perspective is to abandon rational inquiry itself.

8. THEORIES VERSUS PICTURES OF REFERENCE This conclusion about Quine's view of language is apt to strike many as out of character with Putnam's early work. Putnam at times seems to believe that his sketches of reference and belief are valuable only if they can be developed into a rigorous scientific theory. This appearance is reinforced by the work of

20


two of his students, Hartry Field and Michael Devitt. Both Field and Devitt maintain that if it turns out that according to our best scientific account of language, the references of our words and the contents of our beliefs are not uniquely determined, then we must conclude that there is no fact of the matter about what our words refer to, and what we believe. 43 I have argued that for Putnam the determinacy of reference and belief in our commonsense and scientific practices is immune from any general challenge of this kind. But how is this compatible with his interest in a scientific theory of reference'?44 In order to address this question, we must distinguish between a theory of reference and a picture of reference. Following Kripke, I assume that a theory of reference is a non-circular statement of necessary and sufficient conditions for a word to have a particular reference.45 A theory of reference might be given in purely scientific terms, or it might be stated using nonscientific notions taken from our ordinary practices. The important point is that a theory of reference will give necessary and sufficient conditions, without employing the notion of reference, or any notion which implicitly presupposes the notion of reference. 46 A picture of reference, on the other hand, does not state necessary and sufficient conditions for a term to have a particular reference. Instead, it relates our concept of reference to other concepts, like truth, belief, agreement, and disagreement. A particular picture of reference is valuable to us to the extent that it to clarifies our implicit understanding of reference. Since many aspects o~ our implicit understanding of reference are quite sketchy, a picture of reference need not be precise in order to be valuable. Some philosophers believe that reference is not an objective relation unless there is a true scientific theory of reference. They believe that the criterion for the determinacy of reference is scientific. This is what I call scientific naturalism about reference. Both Quine and Field are scientific naturalists in this sense. They disagree about whether there is a true scientific theory of reference, but they each accept a scientific criterion for the determinacy of reference. A belief in scientific naturalism is one motivation for trying to develop a scientific theory of reference. But one need not be a scientific naturalist about reference in order to be interested in developing a scientific theory of reference. In particular, Putnam's interest in developing a s~ientific theory of reference is not due to an underlying belief in scientific naturalism. He never believed that the criterion for the determinacy of reference and belief is scientific. Instead, his interest in developing a scientific theory of reference is a natural result of his desire to clarify our understanding of the norms implicit in our linguistic practices. Our understanding of these norms will be greatly clarified if someday we develop a scientific theory of reference. So given his underlying project, it is not surprising that Putnam sometimes shows an interest in developing such a theory. Moreover, the same underlying project

21


motivates an interest in developing a clarifying picture of reference. Putnam's causal picture of reference contributes to his underlying project even if that picture cannot be developed into a scientific theory. This is why Putnam does not show any genuine concern about whether his picture of reference can be developed into a scientific theory of reference. For example, in "Language and Reality," a paper which falls in what is standardly regarded as his first period, Putnam comes close to saying that the notion of reference must be made scientifically precise, or it is unusable. After noting the importance of the reference relation in his view of language, Putnam writes: ... unless we can say something informative about this relation, our entire philosophy of language rests comfortably in cloudcuckoo-land .... if we don't know what referring to is, we may assert that Bohr was referring to electrons when he used the word 'electron' or deny that he was; since it is unclear just what relation between Bohr's word 'electron' and the particles in question is being affirmed or denied, a methodology for such affirmation and denials is a methodology for a science which, however valuable and important its results, still rests upon unclear notions. 47

If Putnam was a scientific naturalist about reference when he wrote this passage, one would expect him to show genuine concern about whether there is a true scientific theory of reference. But he seems completely unconcerned about this. Instead of trying to sketch a scientific theory of reference, he describes a thought experiment which shows how we might come to view a group of speakers as referring to things in their environment. The thought experiment begins with what Putnam calls "primitive reference". His account of how we come to see the speakers as primitively referring to various objects in their environment is not a statement of necessary and sufficient conditions for reference. He explicitly disavows any intention to provide such a statement: Let me emphasize that I am not, repeat not, trying to give necessary and sufficient conditions for reference, even primitive reference. I am trying to describe a fairly understandable situation in which we can employ a primitive notion of reference. 48

This passage makes good sense on my interpretation of his project, but it would be hard to understand if we assume that Putnam endorsed scientific naturalism when he wrote it. Moreover, in all of his early papers, as far as I know, Putnam displays a similar lack of concern about whether his causal picture of reference can be developed into a scientific theory. I conclude that Putnam never accepted a scientific criterion for the determinacy of reference. 49 His interest in developing a scientific theory of reference was motivated instead by his project of clarifying our understanding of the nonns underlying our commonsense and scientific practices.

22


9. INTERNAL REALISM, TRUTH, AND RATIONAL ACCEPTABILITY Despite his implicit rejections of metaphysical realism and scientific naturalism, Putnam was a realist in his early period, and he still is. In order to help clarify our understanding of the norms underlying our linguistic practices, Putnam observes that we refer to entities which exist independently of our beliefs about them.50 In this sense, Putnam is a realist. However, the standard view that Putnam was a metaphysical realist in his early period is based on a common but egregious slide from the claim that the entities to which we refer exist independently of our beliefs about them, to the conclusion that our conception of those entities is independent of all of our substantive beliefs' about them. I argued above that on Putnam's view we cannot make sense of metaphysical realism because our understanding of representation is inextricably bound up with our beliefs about the causal relations we bear to things in our environment. Another side of the same basic point is that our ontological notions are not independent of the norms underlying our rational inquiries. Since there is no way to characterize these norms without appealing to our substantive empirical beliefs, we have no conception of the references of our words which is completely independent of all of those beliefs. So, contrary to the standard view of Putnam's early work, Putnam's argument for realism was at the same time a rejection of metaphysical realism. Internal realism is not fundamentally different from the realism Putnam defended in his early papers. What Putnam characterizes as his shift to internal realism is better understood as a change in focus. In earlier papers Putnam was primarily concerned to overthrow certain entrenched philosophical conceptions of reference and belief. His tum towards internal realism reflects his growing appreciation of the subtlety required in order to make sense of the realism implicit in his earlier view. Putnam's own confusion on this point has led to some unfortunate claims about what internal realism amounts to. My interpretation of his underlying project simultaneously helps us to see what is attractive about internal realism, and enables us to criticize some of Putnam's less careful characterizations of it. Putnam has characterized internal realism in a number of apparently inequivalent ways. The central point underlying these various characterizations is that our notions ofJact and truth are not available independently of the norms underlying our linguistic practices. Putnam has emphasized that we need to develop a more refined understanding of truth, and his remarks about truth are often accompanied by comments about facts. For example, in the preface to Reason, Truth and History, he writes that The view which I defend holds, to put it very roughly, that there is an extremely close connection between the notions of truth

23


and rationality; that, to put it even more crudely, the only criterion for what is a fact is what it is rational to accept. 51

Here a criterion for "what is a fact" follows one about truth. Our understanding of facts is intimately connected with our understanding of truth, according to Putnam. Putnam suggests that the connecting link is the notion of rational acceptability implicit in our everyday and scientific inquiries. For example, later in Reason, Truth and History he claims that " ... truth itself gets its life from our criteria of rational acceptability. "52 He would say the same about our concept of fact. Partly on the basis of claims like this, internal realism is standardly thought to be a kind of verificationism. In some passages Putnam seems to hold that truth can be defined in terms of rational acceptability or justification. In the preface to Realism with a Human Face, for example, he writes that According to my conception, to claim of any statement that it is true, that is, that it is true in its place, in its context, in its conceptual scheme, is, roughly, to claim that it could be justified were epistemic conditions good enough. 53

An epistemic condition is "good enough" in Putnam's sense if it is "ideal". In a later passage, he offers an example of what he means by an "ideal" epistemic situation for the rational assessment of the sentence "There is a chair in my study": If I say "There is a chair in my study," an ideal episternic situation would be to be in my study with the lights on or with daylight streaming through the window, with nothing wrong with my eyesight, with an unconfused mind, without having taken drugs or been subjected to hypnosis, and so forth, and to look and see if there is a chair there. 54

In these two passages Putnam is apparently claiming that truth is to be defined in terms of rational acceptability. The suggestion is that "There is a chair in my study" is true just in case I would affirm it under the "ideal" conditions he describes. Putnam seems to be endorsing the following generalization about the relationship between truth and justification: (C) For every statement S, to say that S is true is to say that if episternic conditions were ideal, we would be justified in affirming S.

But before we accept this interpretation, let us consider how (C) fits with Putnam's project of clarifying the norms underlying our commonsense and scientific practices. If (C) does not accurately reflect those norms, then it does not accurately reflect (what ought to be) Putnam's view of the relationship between truth and rational acceptability.

24


Surely some situations are better than others for determining whether or not there is a chair in my study. In the best of these situations, like the one described above by Putnam, the truth of my utterance of the sentence "There is a chair in my study" can be readily determined on the basis of the available evidence. But this observation does not by itself support (C). It may be that for some statements truth and rational acceptability are not so closely connected. More specifically, (C) is false if there is at least one statement S such that any situation in which S is true is a situation in which we would not be justified in affirming S. Consider the statement (*) Last night my brain was transferred into a vat of nutrients, and my neural receptors are now stimulated in just the way they would have been stimulated had my brain not been removed.

Any situation in which (*) is true is one in which I would not be justified in believing it. 55 This example shows that a speaker can entertain a thought about himself even if there is no situation in which he would be justified in believing it. A speaker's understanding of the truth of a statement is not in every case directly tied to his understanding of situations in which he would be able to justify it. Thus it seems that (C) is false. 56 Putnam may not have intended to claim that the truth of a statement S made by a person P must be defined in terms of situations in which P would be justified in accepting S. Perhaps Putnam meant to make the weaker claim that the truth of a statement S made by a person P is to be defined in terms of situations in which S could be verified by someone or other, not necessarily P. And this condition seems to be met by (*), if we take a liberal view of what (*) actual states. Suppose we say that the content of (*) does not depend essentially on the indexicals it contains, but is equivalent to: (**) Last night Gary Ebbs's brain was transferred into a vat of nutrients, and his neural receptors are now stimulated in just the way they would have been stimulated had his brain not been removed.

Then if (*) were true, it might be that someone else - the doctor who performed the brain transfer, for example - would have justification for believing it. So perhaps we can interpret (C) in such a way that (*) does not provide a counter example to it. In any case, (*) shows that there is no simple relationship between the truth of a statement S made by a person P and the situations in which P would be justified in accepting S. So we should not assume that internal realism involves the view that the truth of a statement S made by P can be defined in terms ofthe conditions under which it would be rational for P to affirm S.

25


The most we can say is that our understanding of what it means to say that a statement S is true is essentially tied to our conception of situations in which S would be correctly affirmed by someone or other. But this tells us very little about the precise relationship between the truth of particular statements and the situations in which such statements would be correctly affirmed. For most statements, our present conceptions of when they would be correctly affirmed are either subject to fundamental revision, or extremely vague, or both. Thus we are virtually never in a position to equate the truth of a particular statement with its verifiability in situations which we can precisely describe. All we can say with confidence is that there is an interdependence between the notions of truth and rational acceptibility, because they are both rooted in the norms underlying our everyday and scientific inquires. Putnam has explicitly endorsed this open-ended view of the relationship between truth and rational acceptability: In Reason, Truth and History, I explained the idea thus: "truth is idealized rational acceptability." This formulation was taken by many as meaning that "rational acceptability" ... is supposed (by me) to be more basic than "truth"; that I was offering a reduction of truth to epistemic notions. Nothing was farther from my intention. The suggestion is simply that truth and rational acceptability are interdependent notions. 57

On this more sophisticated view, only careful, context-sensitive investigations of the norms underlying our rational inquires can shed light on the complex relationship between our concepts of truth and rational acceptability. There is no reason to think that there are any informative generalizations about the relationship between these concepts. Hence the standard view that internal realism involves the definition of truth in terms of rational acceptability is incorrect. Philosophers who feel that they understand metaphysical realism will not doubt find that even on this more sophisticated interpretation of internal realism, truth is too closely tied to rational acceptability. This reaction is encouraged by Putnam's unfortunate use of the word "internal" to characterize his realism. For this word suggests that there is an "external" alternative to internal realism, a legitimate perspective from which internal realism looks like an optional view of the relationship between our concepts of truth and rational acceptability. But this reaction rests on a failure (or refusal) to understand Putnam's realism. As we have seen, from Putnam's point of view, metaphysical realism is a thesis which has no genuine content. There is no legitimate perspective from which internal realism looks like an optional view of the relationship between our concepts of truth and rational acceptability. So the metaphysical realist's charge that internal realism ties truth too closely to rational acceptability is empty.

26


NOTES I. 2.

3. 4.

5. 6.

7. 8.

9.

10.

For helpful comments on earlier drafts, I am grateful to Peter Dillard, Sally Haslanger, Mark Kaplan, Scott Kimbrough, Tom Ricketts, Jay Wallace, and Joan Weiner. Berlin contrasts the fox with the hedgehog in the first paragraph of his essay titled "The Hedgehog and the Fox: an Essay on Tolstoy's View of History" (London: Weidenfeld and Nicolson, Ltd., 1953). Here is the most relevant part of that paragraph: ... there exists a great chasm between those, on one side, who relate everything to a single central vision, one system less or more coherent or articulate, in terms of which they understand, think and feel-a single, universal, organizing principle in terms of which alone all that they are and say has significance-and, on the other side, those who pursue many ends, often unrelated and even contradictory, connected, if at all, only in some de facto way, for some psychological or physiological cause, related by no moral or aesthetic principle; these last lead lives, perform acts, and entertain ideas that are centrifugal rather than centripetal, their thought is scattered or diffused, moving on many levels seizing upon the essence of a vast variety of experiences and objects for what they are in themselves, without, consciously or unconsciously, seeking to fit them into, or exclude them from, anyone unchanging, all embracing, sometimes self-contradictory and incomplete, at times fanatical, unitary inner vision. The first kind of intellectual and artistic personality belongs to the hedgehogs, the second to the foxes .... (2) Reprinted in Meaning and the Moral Sciences (London: Routledge and Kegan Paul, 1978), 123-40. For example, in such papers as "Why there Isn't a Ready-made World," and "Why Reason Can't Be Naturalized," both in Putnam's third volume of collected papers, titled Realism and Reason (Cambridge: Cambridge University Press, 1981). In "The Analytic and the Synthetic," reprinted in Mind, Language and Reality (Cambridge: Cambridge University Press, 1975). The essentials of the following interpretation of Carnap are due to Thomas Ricketts. See his paper "Rationality, Translation, and Epistemology Naturalized," Journal of Philosophy 79: 117-36. Rudolph Carnap, "Intellectual Autobiography," in Paul A. Schilpp, ed. The Philosophy of Rudolph Camap (La Salle: Open Court, 1963),44-45. This might suggest that Carnap's conclusion that the traditional metaphysical disputes are not genuine was a violation of his own Principle of Tolerance. But as Carnap saw it, the problem with those disputes is not that they result from the acceptance of different linguistic frameworks, or that they arise only within linguistic frameworks Carnap does not accept. The problem is more fundamental: according to Carnap, we are unable to specify a linguistic framework within which these disputes can be assessed. The reader may wish to dispute Putnam's claim that scientists who accept the later equation disagree with those who accepted the former one. Whether or not Putnam is right about this particular case, there will be others which do make his point. I am primarily interested here in his philosophical methodology, and not in the question of whether his· particular examples must be understood in precisely the way he presents them. Putnam contrasts law-cluster terms with what he calls one-criterion words, like "bachelor". He observes that In the case of a law-cluster term such as 'energy', anyone law, even a law that was felt to be definitional or stipulative in character. can be abandoned, and we feel that the identity of the concept has, in a certain respect, remained .... But 'All bachelors are unmarried' cannot be rejected unless we change the meaning of the word 'bachelor' and not even then unless we

27


11.

12. 13.

14.

15. 16. 17. 18.

28

change it so radically as to change the extension of the tenn 'bachelor'. ('The Analytic and the Synthetic," 53.) According to Putnam, we can understand our different judgments about these two cases by noting that there is only one criterion for being a bachelor. Consequently, if we give up this criterion, we essentially change the meaning of "bachelor". But there is an openended set of laws containing the tenn "energy", and our own practice shows us that we may abandon any of these laws without necessarily changing the reference of the tenn "energy". On the logical positivists' model of language, all tenns are in effect treated as one-criterion tenns. This model of language does not fit with many of our confident judgments about when investigators are making incompatible statements about the same subject. There is some similarity between Putnam's reasons for rejecting Carnap's model of meaning and J. L. Austin's rejection of the analytic-synthetic model of meaning. This is clear in the following passage from Austin's paper "The Meaning of a Word" in J. o. Unnson and G. J. Warnock, eds., J. L. Austin: Philosophical Papers (Oxford: Oxford University Press, 1961),54-75: ... if 'explaining the meaning' is really the complicated sort of affair that we have seen it to be, and if there is really nothing to call 'the meaning of a word' -then phrases like 'part of the meaning of the word x' are completely undefined .... We are using a working-model of meaning which fails to fit the fact that we really wish to talk about . ... it is when we are required to give a general definition of what we mean by 'analytic' or'synthetic', and when we are required to justify our dogma that every judgement is either analytic or synthetic, that we find we have, in fact, nothing to fall back upon except our working model. (62-63) Thanks to Peter Dillard for suggesting that I address this objection. The phrase "available prior to or independently of' is adapted from similar phrases, used to make a related point about Frege, in Thomas Ricketts's paper "Objectivity and Objecthood: Frege's Metaphysics of Judgement" in L. Haaparanta and J. Hintikka, eds. Frege Synthesized (Dordrecht: D. Reidel, 1986). The following interpretation of Quine's critique of Carnap is essentially due to Thomas Ricketts. See his paper "Rationality, Translation, and Epistemology Naturalized." W. V. Quine, "Two Dogmas of Empiricism" in From a Logical Point of View (New York: Harper and Row, 1963),20-46. W. V. Quine, Word and Object (Cambridge, Mass.: MIT Press, 1960). "Ontological Relativity" in Ontological Relativity and Other Essays (New York: Columbia University Press, 1969),29. This way of viewing the difference between Putnam's project and Quine's is inspired by Sir Peter Strawson's article "Freedom and Resentment," Proceedings of the British Academy, vol. xlviii: 1-25. Strawson contrasts detached and non-detached attitudes toward others. His discussion of agency and responsibility depends on our taking seriously the non-detached "participant" attitudes towards others: What I want to insist on is the very great importance that we attach to the attitudes and intentions towards us of other human beings, and the great extent to which our personal feelings and reactions depend upon, or involve, our beliefs about these attitudes and intentions. (64) These attitudes invite us to view others as agents who are responsible for their actions. One of Strawson' s central points is that our concepts of free agency and responsibility get their significance from, and are essentially bound up with, our non-detached reactive attitudes towards others. He contrasts these attitudes with the detached attitudes that we typically associate with a more "objective" point of view on others: What I want to contrast is the attitude (or range of attitudes) of involvement or participation in a human relationship, on the one hand, and what might be called the objective attitude (or range of attitudes) to another human


19. 20. 21. 22.

23. 24.

being, on the other .... To adopt the objective attitude to another human being is to see him, perhaps, as an object of social policy; as a subject for what, in a wide range of sense, might be called treatment. ... If your attitude towards someone is wholly objective, then though you may fight him, you cannot quarrel with him, and ... you cannot reason with him. (66) I do not like Strawson's use of the word "objective" here, since it implies that the participant attitudes, and the concepts of agency and responsibility which go with them, are not really objective. But I find the contrast in points of view illuminating. I believe that intentional notions like agreement and disagreement are like the concepts of agency and responsibility: they get their significance, and are essentially bound up with, our nondetached participation in commonsense and scientific linguistic practices. When one tries to take up a scientific, "objective" attitude towards the linguistic behavior of others, and oneself, one is abandoning the intentional notions of agreement and disagreement altogether. This is what Quine does, and it leads inevitably to his indeterminacy thesis. I urge that we follow Putnam, not Quine, and take our participation in commonsense and sci-. entific linguistic practices seriously at the start of our philosophizing about meaning. There are important connections between these differences between Putnam and Quine, and Wittgenstein's discussions of following a rule, especially as they are interpreted by Kripke in Wittgenstein on Rules and Private Language (Cambridge, Mass.: Harvard University Press, 1982). It seems to me that what Kripke calls the skeptical paradox about meaning is the inevitable result of treating a speaker's linguistic behavior as an object of investigation, and asking for an "objective" basis for regarding that behavior as rule-governed. To avoid the paradox, we must come to see that the "objectifying" perspective on language is optional, and does not undermine our ordinary practice-based confidence that we are rule-followers. "Dreaming and 'Depth Grammar' ," Mind. Language and Reality, op. cit., 310. Ibid., 311. Ibid. Putnam also shows that the references of our terms depend on the nature of our social environments. This aspect of his view is not central to my present concern, which is to show that the standard metaphysical interpretation of Putnam' s causal picture of reference is not correct. On the other hand. the dependence of our references on social factors is part of the interdependence of reference and belief. "The Meaning of 'Meaning'," 237. "The Meaning of 'Meaning'," 240-41. Of course Putnam does not mean that water could have turned out to have no hidden structure. He is more precise in the following passage: When we say that it could have turned out that water had no hidden structure what we mean is that a liquid with no hidden structure (i.e. many bits of different liquids. with nothing in common except superficial characteristics) could have looked like water, tasted like water, and have filled the lakes, etc., that are actually full of water. (241)

25. Ibid., 241. 26. Ibid. 27. What I am calling "practice-based" judgments about when a term is correctly applied are different from theoretical beliefs about how the term is to be correctly applied. Putnam's examples show that almost all of a person's or a community's theoretical beliefs about how to apply a term may be false. This is possible because the actual pattern of application of a word within a given linguistic community may be misunderstood by many, or all. of the members of that community. Their theoretical understanding can in some cases be improved by an investigation into the nature of the things to which they actually apply the word. But my point here is that at the most basic level of our use of language, we cannot usefully separate our understanding of the reference of a word from our underlying practice-based convictions about when we have correctly applied it, even though we

29


28. 29. 30.

31.

acknowledge that those convictions can in some cases be underminded by further inquiry. (I am grateful to Scott Kimbrough for helping me to see the need for further clarification of this point.) These issues require further clarification and investigation. At present I aim only to sketch an alternative reading of Putnam, one which does not presuppose that we have a conception of the entities in the world completely independent of our participation in ongoing linguistic practices. I am focusing on the question of how we are to understand the relationship between our applications of a word-including those that Tyler Burge would call "archetypical"-and what Burge calls the "cognitive values" of our words. For an interesting discussion of some aspects of the relationship between archetypical applications and cognitive value, see Burge's paper "Intellectual Norms and the Foundations of Mind," Journal of Philosophy: 697-720. Ibid., 242. Chapter 15 of Mathematics, Matter and Method (Cambridge: Cambridge University Press, 1975). One might wonder how this part of Putnam's view fits with his earlier claim that a necessary and sufficient condition for some stuffs being 'Jade' is that it possess enough of the appropriate textual qualities. This sounds like the claim that 'Jade' is synonymous with a description. Despite first appearances, Putnam's claims about jade and his thought experiment involving pencils are compatible. The reason is that in the jade case, Putnam is assuming that there is in fact more than one kind of stuff which we correctly call 'jade'. In the pencil case, he is suggesting that we are wrong to assume that 'pencil' is an artifact term. Thus in the 'jade' case, Putnam is focusing on the conditions for applying a particular word, given that we assume we are right about the underlying structure of the things to which it applies. In the pencil case, he is warning us not to be complacent about our classification of words as artifact terms or as natural kinds. The pencil example looks more like the jade example if we assume that we are right that pencils are artifact terms. The interdependence of meaning and belief was first stressed by Quine in chapter two of Word and Object. Quine has a very different view of meaning from Putnam, and so there is not much in common between Quine's view of the interdependence and Putnam's view of it. For one thing, as we shall see later, unlike Putnam, Quine holds that reference is indeterminate. Davidson developed Quine's idea in his own way in a series of papers about meaning and truth. See his Inquiries into Truth and Interpretation (Oxford: Clarendon Press, 1984). Since Davidson also accepts that reference is inscrutable, his view is quite different from Putnam's too. In "The Analytic and the Synthetic," Putnam briefly notes, but does not explain, that on his view belief and reference are interdependent notions: ... I should like to stress the extent to which the meaning of an individual word is a function of its place in the network, and the impossibility of separating, in the actual use of the word, that part of the use which reflects the 'meaning' of the word and that part of the use which reflects deeply embedded collateral information. ("The Analytic and the Synthetic," 40-41)

As we have seen, Putnam's later developments of the picture sketched in "The Analytic and the Synthetic" reinforce and validate this enigmatic observation. 32. There may be other kinds of metaphysical realism, but I think this is the one which Putnam has in mind when he rejects metaphysical realism as part of his tum towards internal realism. For a slightly more detailed account ofthe absolute conception than I present here, see Bernard Williams, Descartes: the Project of Pure Inquiry (Harrnondsworth: Penguin, 1978),64-65, and Thomas Nagel, The View From Nowhere (Oxford: Oxford University Press, 1986), chapters II and V. 33. On page 386 of his paper 'The Significance of Quine's Indeterminacy Thesis," reprinted in Truth and Other Enigmas (Cambridge, Mass.: Harvard University Press, 1978), Michael Dummett observes that Quine appeals to conditional dispositions-Le., dispositions to assent or dissent from a sentence, given that one has already assented or dissented

30


34.

35.

36. 37. 38. 39. 40.

to another-to characterize the meaning of truth-functional connectives in objective behavioral terms. Thus the behavioral meaning of the truth functions is not exhausted by the stimulus meaning of sentences in which they occur. There is some question of whether Quine believes that assent and dissent can be given purely objective behavioral characterization. For example, in "Reply to Hintikka" in Words and Objections ed. D. Davidson and 1. Hintlika (Durdrecht: D. Reidel, 1969) he says: The linguist's decision as to what to treat as native signs of assent and dissent is on a par with the analytical hypotheses of translation that he adopts at later stages of his enterprise; they differ from those later ones only in coming first, needed as they are in defining stimulus meaning. This initial indeterminacy, then, carries over into the identification of the stimulus meanings. (312) This is in tension with the idea that the facts about stimulus meaning are objective. In a later paper, "Mind and Verbal Dispositions," in S. Guttenplan, ed. Mind and Language (Clarendon Press, 1975), Quine seems to reaffirm a commitment to the objectivity of assent and dissent, construed behaviorally: It has been objected that assent is no mere mindless parroting of an arbitrary syllable; utterance of the syllable counts as assent only if there is the appropriate mental act behind it. Very well, let us adopt the term suiface assent for the utterance or gesture itself. My behavioral approach does indeed permit me, then, only to appeal to surface assent. ... (91) Observation sentences are those which prompt assent only after an appropriate prompting stimulation, and whose stimulus meanings are (approximately) the same for all speakers. By truth functions of such sentences I mean those native sentences which are formed by joining simple observation sentences with the truth-functional expressions of the language. Such complex sentences will be objectively translatable because both the simple observation sentences and the truth-functional expressions are objectivity translatable, according to Quine. See chapter two of Word and Object, for the details. I say that the translations of these sentences are "fairly well" determined by their stimulus meanings because Quine later acknowledges indeterminacy even for these sentences. See Pursuit of Truth (Cambridge Mass.: Harvard University Press, 1990), § § 15-16, where Quine concedes that even the translation of observation sentences is not objective, but depends on "empathy." The tension in Quine's view is now very strong. For he holds both that "The requirement of intersubjectivity is what makes science objective" (5), and that there is no fact of the matter about how to translate even observation sentences. "Ontological Relativity," 31 See Word and Object, 50. Ibid., 33. "Ontological Relativity," 49. Some philosophers accept both Quine's requirement that there be an objective basis for using intentional notions and his scientific naturalism, but dispute his skeptical conclusion. They argue that there may be other scientific facts which, in conjunction with those Quine mentions, are sufficient to determine the translation of terms. For example, Hartry Field, in "Tarski's Theory of Truth," The Journal of Philosophy, vol. LXIX, no. 13 (1972), and "Conventionalism and Instrumentalism in Semantics," Nous 9 (1975), has argued that there may be a scientific basis for ascribing particular references to an investigator's words. On Field's view our words may have unique references, even if we require a scientific basis for attributing references to them. I don't find this response to Quine convincing. I doubt that there is a scientific basis for attributing particular references or meanings to a speaker's words. But I won't defend this claim here. Instead, I want to consider what our attitude should be to Quine's scientific naturalism, if it implies that agreement and disagreement are not objectively determined. Noam Chomsky has denied that facts about meaning and reference must be determined by physical facts in order for semantics to be scientifically legitimate. See "Quine's

31


41. 42. 43.

44.

45.

46.

47. 48. 49.

32

Empirical Assumptions," in Words and Objections. Chomsky claims in effect that semantics is a respectable "special science," and so it is not subject to the requirements Quine lays down for determinacy. It seems to me that Chomsky is confusing the legitimacy of rational inquiry into reference and meaning with the scientific legitimacy of reference and meaning. I do not think that all "facts"-or true statements-are scientific; some are part of our commonsense, everyday practices. Although from Putnam's perspective it is obvious that (a) ordinary statements about what a person refers to and means can be objective or "factual," it does not follow that (b) such statements can be part of a scientific description of the world. It seems to me that many philosophers who are convinced by Chomsky's reply to Quine are in effect sliding from (a) to (b). This slide is unwarranted in my view. From "00 Truth," in Leigh S. Caumon, ed., How Many Questions? (Indianapolis: Hackett, 11983),43. Ibid., 45. See, for example, Field's articles, "Tarski's Theory of Truth," The Journal of Philosophy, vol. LXIX, no. 13 (1972), and "Conventionalism and Instrumentalism in Semantics," Nous 9 (1975), and Michael Devitt's book Designation (New York: Columbia University Press, 1981). Even as late as "Realism and Reason," the paper in which Putnam first announces his tum towards internal realism, he claims to hold a scientific theory of reference and belief: ... a 'correspondence' between words and sets of things (fonnally, a satisfaction relation, in the sense of Tarski) can be viewed as part of an explanatory model of the speaker's collective behaviour .... Let me refer to realism in this sense-acceptance of this sort of scientific picture of the relation of speakers to their environment, and of the role of language-as internal realism. (123) Putnam sketches this explanatory model of speaker's collective behavior in "Reference and Understanding" in Meaning and the Moral Sciences. Perhaps more than any other of Putnam's papers on reference, this one seems incompatible with my interpretation. I can't go into details here, but I believe that what I say about Putnam's interest in scientific theories of reference is the first step towards reconciling "Reference and Understanding" with the interpretation I favor. See Saul Kripke, Naming and Necessity (Cambridge, Mass.: Harvard University Press, 1980), 93. On 68-70 Kripke explains why a successful theory of reference must not be circular. The non-circularity condition makes it extremely unlikely that a theory of reference could be stated using ordinary intentional notions. The reason is that these notions are part of what Quine calls "Brentano's circle", and therefore implicitly presuppose the notion of reference. I think it is equally unlikely that a theory of reference could be given in purely naturalistic tenns, though I will not try to defend this claim here. From "Language and Reality" in Mind, Language and Reality (Cambridge: Cambridge University Press, 1975),283. Ibid., 286. Putnam was probably too optimistic in his early writings about the prospects for developing a theory of reference, but this does not affect the point I am making here. I agree with Thomas Blackburn, who argues in "The Elusiveness of Reference" in P. French, T. Uehling, Jr., H. Wettstein, eds., Midwest Studies in Philosophy, Vol. XII (Minneapolis: University of Minnesota Press, 1988), that there is no way to develop a theory of reference from the sketches Putnam and others have given us. The reason is that the concepts of reference and belief are essentially interdependent. (Blackburn does not put it this way, but I believe that his arguments work only because of this interdependence.) Because they are interdependent, we can't give a non-circular statement of what reference consists in. The best we can do is to sketch a picture of reference which traces some of the interconnections between reference and other intentional notions, like belief.


50. In "The Meaning of 'Meaning'," often taken as the locus classicus of Putnam's Metaphysical Realism, he is quite clear that he takes his view to be a kind of realism precisely because it makes use of an inter-theoretical notion of extension. This emerges clearly when he notes the contrast between anti-realist positions and the position he endorses: ... for a strong anti-realist truth makes no sense except as an intratheoretic notion .... The antirealist can use truth intra-theoretically in the sense of a 'redundancy theory'; but he does not have the notions of truth and reference available extra-theoretically. But extension is tied to the notion of truth. The extension of a term is just what the term is true of. Rather than try to retain the notion of extension via an awkward operationalism, the anti-realist should reject the notion of extension as he does the notion of truth (in any extra-theoretic sense). Like Dewey, he can fall back on a notion of 'warranted assertibility' instead of truth .... (236) This passage, seen against the background of my reading of Putnam, shows how we are to understand his realism. We need the notion of inter-theoretic extension to make sense of commonsense and scientific agreements and disagreements, as we have seen above. The anti-realist cannot accept that notion. But then the anti-realist cannot make sense of the norms governing our commonsense and scientific practices. This is a reductio ad absurdum ofthe anti-realist's view, according to Putnam. Note also that Putnam associates the notion of 'warranted assertibility' and operationalism with this kind of anti-realist view. Since it is not part of Putnam's internal realism to abandon the idea of inter-theoretic extension, he ought still to maintain that we should not identify meaning with warranted assertibility. He is sometimes confused on this point, as I argue below, and this is no doubt one reason why he has been misunderstood. 51. Reason, Truth and History, x. 52. Ibid., 130. 53. J. Conant, ed., Realism with a Human Face (Cambridge, Mass.: Harvard University Press, 1990), vii. This is not the only place where Putnam makes this kind of claim. Similar claims appear in many of his writings on internal realism. For example, here is a passage from Representation and Reality: The suggestion I am making, in short, is that a statement is true of a situation just in case it would be correct to use the words of which the statement consists in that way in describing the situation .... We can explain what "correct to use the words of which the statement consists in that way" means by saying that it means nothing more nor less than that a sufficiently well placed speaker who used the words in that way would be fully warranted in counting the statement as true in that situation. (l15) 54. Preface to Realism with a Human Face, viii. 55. In fact, it seems to me that any situation in which this statement is true is one in which I am justified in believing that it is false (supposing that I am justified in believing anything at all). 56. Perhaps Putnam would deny that we really understand the statement "Last night my brain was transferred into a vat of nutrients, and my neural receptors are now stimulated in just the way they would have been stimulated had my brain not been removed." But it seems like a meaningful statement, it does not violate any of the conditions on meaning set out in section one, nor is it self-refuting. I see no reason to conclude that it does not make sense. Note that the situation described by (*) is different from the hypothetical situation in which my brain is always in a vat. In the first chapter of Reason, Truth and History, Putnam argues that the latter possibility can't be actual, on the grounds that my supposition that I am always a brain in a vat is self-refuting. 57. From Representation and Reality (Cambridge, Mass.: MIT Press, 1988), 115.

33

McDowell, John, Putnam on Mind and Meaning , Philosophical Topics, 20:1 (1992:Spring) p.35

PHILOSOPHICAL TOPICS VOL. 20 No.1, SPRING 1992

Putnam on Mind and Meaning

John McDowell University of Pittsburgh

1. To begin with in "The Meaning of 'Meaning' ,"1 and in a number of writ-

ings since then, Hilary Putnam has argued trenchantly, and I think convincingly, that in the case of at least certain sorts of words, the environment of those who use them enters into determining their extension. We cannot understand what constitutes the fact that a natural-kind word like "water," as used by ordinarily competent speakers of English, has the extension it does without appealing to the actual scientifically discoverable nature of a stuff that figures in their lives in a way that has an appropriate connection to the correct use of the word, and to facts of a broadly sociological kind about relations within the community of English speakers. Now it seems plausible that the extension of a word as a speaker uses it should be a function of its meaning; otherwise we lose some links that seem to be simply common sense-not part of some possibly contentious philosophical theory-between what words mean on speakers' lips, what those speakers say when they utter those words, and how things have to be for what they say to be true. 2 1f we keep those links, Putnam's thesis about extension carries over to meaning: that a speaker means what she does by "water" must be constituted at least in part by her physical and social environment. As Putnam memorably puts it: "Cut the pie any way you like, 'meanings' just ain't in the head!"3 I have rehearsed this basic thesis of Putnam's in a deliberately unspecific

35


way. The question I want to raise in this paper does not require going into possibly disputable details about how the physical and social environments serve to determine extension, or how the roles of the physical and social environments might be related. 4 Nor do I need to go into the question of how far similar theses can be made out to apply beyond the original case of words for natural kinds. 5 I am going to take it for granted that, however such details are to be spelled out, Putnam is right in this basic thesis: at least some meanings are at least in part environmentally constituted. My question is at a more abstract level. I want to ask what significance the basic thesis has for how we ought to conceive the nature of the mind. 2. One might take it to be another simply intuitive idea, not a bit of possibly contentious philosophical theory, that command of a word's meaning is a mental capacity, and exercise of such command is a mental act-an act of the intellect and therefore, surely, of the mind. In that case the moral of Putnam's basic thought for the nature of the mental might be, to put it in his terms, that the mind-the locus of our manipulations of meanings-is not in the head either. Meanings are in the mind, but, as the argument establishes, they cannot be in the head; therefore, we ought to conclude, the mind is not in the head. Rather than arguing, as Putnam does, that the assumption that extension is determined by meaning will not cohere with the assumption that knowledge of meanings is wholly a matter of how things are in a subject's mind, we should insist on making the two assumptions cohere and conceive the mind in whatever way that requires. I want to pursue this line, and urge a reading of the claim that the mind is not in the head that ought, I believe, to be congenial to Putnam, although as far as I can tell it goes missing from the space of possibilities as he considers things, which is organized by the idea that the two assumptions cannot be made out to be compatible. 3. Putnam's argument works against the theory that he sets up as its target, just because the theory is stipulated to include the claim that the mind is in the head. Another way of putting that claim is to say that states of mind, in some strict or proper sense, are what Putnam calls "psychological states in the narrow sense": that is, states whose attribution to a subject entails nothing about her environment. 6 The idea of "psychological states in the narrow sense" contrasts with the idea of "psychological states in the wide sense": these are attributed by intuitively psychological attributions that involve the attributor in commitments about the attributee's environment, as for instance "x is jealous of y" commits the attributor to the existence of y. The conception of meaning that Putnam attacks embodies the claim that knowledge of a meaning is exhausted by a certain psychological state, with "psychological state" stipulated to mean "psychological state in the narrow sense."

36


Now if we try to preserve the thought that knowledge of a meaning is a psychological state, consistently with Putnam's basic thesis that meanings are environmentally constituted, we have to suppose that knowledge of a meaning (at least of the kind that Putnam's thesis applies to) is a "psychological state in the wide sense." And if we try to make sense of that while maintaining the idea that the mental in a strict or proper sense is characterized by "narrow" psychological attributions, we have to suppose that knowledge of a meaning (of the relevant kind), qua mental, is, in itself, a "narrow" psychological state, which, however, can be characterized as knowledge of that meaning only by dint of taking into account the subject's placement in a physical and social environment. On this picture, knowledge of a meaning is, in itself, in the head; the moral of Putnam's basic thought is that we nee.d to be looking at relations between what is in the head and what is not, if it is to be available to us that knowledge of a meaning (at least if it is a meaning of the relevant kind) is what some state, in itself in the head, is. According to this picture, then, there is a sense in which the mind is in the head: that is where the relevant states and occurrences are. But this picture does yield a sense in which we might say that the mind is (at least partly) not in the head: the characterizations that display the relevant states and occurrences as ("wide") content-involving states and occurrences are characterizations in terms of meanings of sorts to which Putnam's argument applies, and hence characterizations that get a grip on the states and occurrences only on the basis of relations between the subject and the environment. At least some distinctively mental truths cannot come into view except in an inquiry that takes account of how the mind in question is related to its environment. The conclusion of this line of thought is that the concept of command of a meaning (at least of the kind that Putnam's argument applies to) is constitutively "duplex," as Colin McGinn puts it: it is the concept of something that is, in itself, in the head, but conceived in terms of its relations to what is outside the head. And this line of thought obviously extends from knowledge of the meaning of "water" (and whatever other meanings Putnam's argument, or something of similar effect, applies to) to, say, beliefs or occurrent thoughts about water (and similarly for whatever other meanings are relevant). It is widely supposed that Putnam's considerations compel a "duplex" conception of at least large tracts of our thought and talk about the mental. The idea is that part of the complete truth about the mind is the truth about something wholly in the head; another part of the complete truth about the mind is the truth about how the subject matter of the first part is related to things outside the head.7 4. This reading of the idea that the mind is not in the head is not what I meant when I suggested that the idea ought to be congenial to Putnam.

37


This reading preserves a role for what is in the head, in the constitution of knowledge of meanings, or more generally in the constitution of psychological states and occurrences such as beliefs or thoughts about water, to which Putnam's claim of environmental determination clearly extends. What is the attraction of this? I think the answer is that, on this "duplex" conception, at least one component of the constitutive truth about the psychological in the "wide" sense looks like an unquestionably suitable topic for a straightforwardly natural science, a science that would investigate how states and occurrences in the head are responsive to impacts from the environment, interact with one another, and figure in the generation of behavior. In Representation and Reality8 and elsewhere, Putnam argues that the role played by interpretation, in a proper account of the import of psychological characterizations in terms of ("wide") content, ensures that psychology in general cannot be within the scope of natural science. But, however convinced we might be by such arguments, there would still be some comfort for a scientistic orientation to the mental in the idea that, all the same, science can in principle be done, and indeed is already being done, about the intrinsic natures of the states and occurrences-in themselves in the headthat those "wide" characterizations get a grip on, in ways that, according to such arguments, are not amenable to scientific treatment. On this account, what makes the "duplex" reading of the thesis that the mind is not in the head attractive is that, by leaving part of the truth about the mind wholly in the head, it offers comfort to a possibly residual scientism about how our understanding of the mental works. 9 But at least since his conversion from scientific realism, Putnam's explicit attitude towards scientism has been one of staunch opposition. When I suggested that the thesis could be read in a way that ought to be congenial to Putnam, I had in mind a reading that would not make even this residual concession to scientism. I had in mind a reading that would place our talk about knowledge of meanings, thoughts about water, and so forth entirely out of the reach of a scientistie conception of the role played by our mental lives in our understanding of ourselves and others. 5. It will be helpful to distinguish a second possible reading of the thesis that the mind is not in the head from the one I mean. This reading is like the one I mean in that it focuses on the literal meaning of "in the head." We might begin explaining the point of denying that the mind is in the head by saying that the mind is not spatially located at all, except perhaps unspecifically, where its owner is. The mind is not somewhere in particular in the literal, spatial, interior of its owner; it is not to be equated with a materially constituted and space-occupying organ, such as the brain. But on the conception I am considering now, the mind is still conceived as an organ: it is just that it is not the brain but an immaterial organ. (A well-

38


placed embarrassment might induce one to add "so to speak.") What I mean by saying that the mind is conceived as an organ is that states of affairs and occurrences in a mind are, on this view no less than on the view that the mind is literally in the head, taken to have an intrinsic nature that is independent of how the mind's possessor is placed in the environment. It is just that this intrinsic nature is not conceived as capturable in the terms of any science that deals with matter, for instance, neurophysiological terms. This reading of the thesis that the mind is not in the head clearly cannot serve my purpose, because it is obvious that this conception of the mind, as an immaterial organ of psychological activity, does not open up a possibility of evading Putnam's argument, so that we could after all locate knowledge of meanings wholly in the mind. Characterizations of the mind, as it is in itself, are no less "narrow" on this picture than they are if conceived as characterizations of what is literally in the head. And Putnam's point is obviously not just that what is literally in the head cannot amount to knowledge of meanings, to the extent to which knowledge of meanings is environmentally constituted. Nothing "narrow," whether material or (supposing we believed in such things) immaterial, can amount to something that is environmentally constituted. We can put the point by saying that the phrase "in the head," in Putnam's formulation of his basic thesis, is already not restricted to a literal, spatial reading. When Putnam says that meanings are not in the head, that is a vivid way of saying that no "narrow" psychological attribution can amount to knowledge of a meaning of the relevant sort, whether it is a material or an immaterial organ of thought in virtue of whose internal arrangements such attributions are conceived as true. 6. I can now sketch the interpretation I mean for the thesis that the mind is not in the head. On this interpretation, the point of the thesis is not just to reject a more specific spatial location for someone's mind than that it is where its possessor is. It is to reject the whole idea that the mind can appropriately be conceived as an organ: if not a materially constituted organ, then an immaterially constituted organ. As I said, the cash value of this talk of organs is the idea that states and occurrences "in" the mind have an intrinsic nature that is independent of how the mind's possessor is placed in the environment. So the point of the different interpretation is to reject that idea altogether. Talk of minds is talk of subjects of mental life, insofar as they are subjects of mental life; and, on the interpretation I mean, it is only a prejudice, which we should discard, that mental life must be conceived as taking place in an organ, so that its states and occurrences are intrinsically independent of relations to what is outside the organism. Of course there is an organ, the brain, whose proper functioning is necessary to mental life. But that is not to say that the proper functioning of that organ is what mental life, in itself, is. And if we deny that, we need not be

39


suggesting instead that mental life is, in itself, the functioning of a mysteriously immaterial para-organ (an organ "so to speak"). Mental life is an aspect of our lives, and the idea that it takes place in the mind can, and should, be detached from the idea that there is a part of us, whether material or (supposing this made sense) immaterial, in which it takes place. Where mental life takes place need not be pinpointed any more precisely than by saying that it takes place where our lives take place. And then its states and occurrences can be no less intrinsically related to our environment than our lives are. 7. Putnam himself expresses skepticism about whether there is any point in reconstructing the intuitive or pre-theoretical conception of the mental, which counts "wide" states like jealousy as psychological, in the way that is prescribed by "methodological solipsism": that is, the thesis that psychological states in a strict and proper sense are "narrow. "10 That skepticism seems to recommend pushing his reflections about terms like "water" in the direction that I am suggesting. What is to be learned from those reflections is not, as Putnam himself argues, that it cannot be true both that "knowing the meaning of a term is just a matter of being in a certain psychological state" and that "the meaning of a term determines its extension"; so that if we retain the second of these assumptions, we must renounce the first. This presupposes that anyone who embraces the first assumption must be restricting psychological states to "narrow" states. Rather, the moral of Putnam's considerations is that the idea of a psychological state, as it figures in the first assumption, cannot be the idea of a "narrow" state. That is: we should not leave in place an idea of the mind that is shaped by the tenets of "methodological solipsism," and conclude that meanings are not in the mind, since they are not in the head. Rather, we should read the two assumptions in such a way that they can be true together and exploit such a reading to force us into explicit consideration of a different conception of the mind. At one point in "The Meaning of 'Meaning' ," Putnam concedes that "it may be trivially true that, say, knowing the meaning a/the word 'water' is a 'psychological state' ."11 The idea that this concession is trivial points to an accommodation of the basic thesis on the lines of the "duplex" conception of the mental. The concession is trivial, on this account, because it does not undermine the view that the two assumptions cannot be true together; given that psychological states in the strict and proper sense are "narrow," knowing a meaning (of the appropriate sort) would not be ''just a matter of being in a certain psychological state," any more than, on that view, any "wide" psychological state would be. What Putnam never seems to consider is the possibility of a position that holds that command of a meaning is wholly a matter of how it is with someone' s mind (the first assumption), and combines that with the determination of extension by meaning so as to force a radically

40


non-solipsistic conception of the mind to come to explicit expression. Instead, he assumes that anyone who wants to conceive knowledge of a meaning as wholly a matter of how it is with someone's mind must be already committed to a theoretical conception of the mind-a conception of the mind as in the head-which, in conjunction with Putnam's reflections about meaning, guarantees that the wish cannot be fulfilled. 12 There may be some temptation to deny that the idea that the mind is in the head is a bit of theory, on the ground of evidently untheoretical usages like "I did the calculation in my head, not on paper." But that idiom does not mesh with the sense that "in the head" bears in Putnam's argument. One might equally take in one',s stride, say, "It came into my head that I wanted a drink of water"; here the meaning of "water" is "in the head" in the sense of the idiom, and the possibility of talking like this obviously poses no threat to what Putnam means by saying that meanings are not in the head.13 The radically non-solipsistic conception of the mental that I am urging would dictate a way of talking about Twin-Earth cases that contrasts with Putnam's. In one of Putnam's cases, the correct extensions of "beech" and "elm" are reversed on Twin Earth, where Putnam has a Doppelganger who is as unable to tell the two kinds of tree apart as Putnam blushingly confesses he is. The words are nevertheless secured their different extensions, on the lips of Putnam and his Doppelganger, by the fact that each defers to a different set of experts. 14 Putnam says, about himself and his Doppelganger when each is in a psychological state, that he would express using one of those terms: "It is absurd to think his psychological state is one bit different from mine." On the conception I am urging, this is not absurd at all. Putnam's psychological state involves his mind's being directed towards, say, beeches (if beeches constitute the extension of the word that he is disposed to use in order to give expression to his psychological state); his Doppelganger's psychological state involves his mind's being directed towards elms. The psychological state of each as it were expands in accordance with the determination of the extensions of their terms, in a way that is compelled if we are to maintain both of the two assumptions. The possibility of talking like this would be merely trivial, in a sense like the one involved in Putnam's concession that we can count knowing a meaning as a psychological state, if the divergent psychological attributions ("thinking of elms" and "thinking of beeches") had to be seen as applying in virtue of some shared underlying psychological state, with the divergence resulting from different ways in which that shared underlying state is embedded in its environment. That is how the "duplex" conception would see things; on this view, Putnam and his Doppelganger do not differ in fundamental psychological properties. But we need not see things this way. It is certainly true that Putnam and his Doppelganger, in the case described, have something psychological in common. (We can make this vivid by noting that if

41


Putnam were transported to Twin Earth without knowing it, he would not be able to tell the difference.) But it is perfectly possible to hold that the psychological common property holds of each in virtue of his "wide" psychological state, rather than that the "wide" state is constituted by the common property, together with facts about how each is embedded in his environment. The common property need not be fundamental. Compare the psychological feature that is unsurprisingly shared between someone who sees that such-and-such is the case and someone to whom it merely looks as if such-and-such is the case. (Again, if one were switched without knowing it between possible worlds that differ in that way, one would not be able to tell the difference.) It is not compulsory to conceive seeing that such-and-such is the case as constituted by this common feature together with favorable facts about embedding in the environment. We can understand things the other way round: the common feature-its being to all intents as if one sees that such-and-such is the caseintelligibly supervenes on each of the divergent "wide" states. And it is better to understand things this way round. It is very common for philosophers to suppose that Twin-Earth comparisons compel the idea that "wide" attributions bear on states that are in themselves "narrow," with the "wide" attributions coming out differently by virtue of the different ways in which those supposedly fundamental psychological states are embedded in extrapsychological reality. But this idea is closely parallel to the Argument from Illusion, and that by itself should be enough to make us suspicious of it. 15 8. Putnam does not seem to consider the possibility that his reflections about meaning might be brought to bear against the idea that the mind is the organ of psychological activity. In fact much of his own thinking seems to presuppose just such a conception of the mind. In Representation and Reality (7), he describes Jerry Fodor's "mentalism" as "just the latest form taken by a more general tendency in the history of thought, the tendency to think of concepts as scientifically describable ('psychologically real') entities in the mind or brain." There is an equivalence implied here between "psychologically real" and "scientifically describable," which cries out to be questioned: it looks like simply an expression of scientism about what it might be for something to be psychologically real. (We do not need to surrender the term "psychological" to scientific psychology.) But as far as I can see Putnam leaves the equivalence unchallenged, even though a great deal of his point in that book is to attack the effects of scientism on how philosophers conceive the mental. The term "mentalism" has a perfectly good interpretation as a label for the view that the mental is a genuine range of reality. (We do not need to accept that the nature of reality is scientifically determined.) But Putnam, without demur, lets "mentalism" be commandeered for the view that the topic of mental dis-

42


course can appropriately be specified as "the mindlbrain." Talk of the mindlbrain embodies the assumption that the mind is appropriately conceived as an organ, together, of course, with the idea-which is in itself perfectly sensible-that if the mind is an organ, the brain is the only organ it can sensibly be supposed to be. The assumption that the mind is an organ is one that Putnam does not challenge. 16 An assumption to the same effect seems to underlie Putnam's argument, in Reason, Truth and History, 17 that one cannot suppose that mental states or occurrences are intrinsically referential-intrinsically directed at the worldwithout falling into a magical conception of reference. Putnam's governing assumption here is that a mental state or occurrence that is representational, sayan occurrence in which one is struck by the thought that one hears the sound of water dripping, must in itself consist in the presence in the mind of an item with an intrinsic nature characterizable independently of considering what it represents. (Such a state of affairs would be what an internal arrangement in an organ ofthought would have to amount to.) It clearly follows, from such a conception of that which is strictly speaking present in the mind, that such items cannot be intrinsically endowed with referential properties; to suppose that they might be would be to appeal to magic, just as Putnam argues. What never comes into view is this possibility: that being, say, struck by a thought is not, in itself, the presence in the mind of an item with a non-representational intrinsic nature. The argument is controlled by the assumption that occurrences in the mind are, in themselves, "narrow." Am I suggesting that being struck by a thought might not involve mental representation? It seems truistic that a thought that such-and-such is the case is a representation that such-and-such is the case. But this is not the notion of mental representation as it figures in Putnam's argument. In Putnam's argument, mental representations are representations in the sense in which, say, drawings or sentences are representations. A representation is an item whose intrinsic nature is characterizable independently of its representational properties: a symbol. The nerve of Putnam's argument is that symbols are not intrinsically endowed with their representational properties, and that claim seems beyond question. But from the fact that thinking, say, that one hears the sound of water dripping is representing that one hears the sound of water dripping, it does not follow that thinking that one hears the sound of water dripping must in itself consist in the presence in the mind of a symbol: something into which the significance that one hears the sound of water dripping can be read, as it can be read into the sign-design "I hear the sound of water dripping," although in both cases the symbol's bearing that significance is extraneous to its intrinsic nature. Putnam's solid point cannot dislodge the possibility that thinking that one hears the sound of water dripping is a mental representation, in the sense of a mental representing, that intrinsically represents what it represents.

43


What this means is that being struck by that thought, say, would not be the mental occurrence that it is if it were not that that one found oneself thinking. What the mental occurrence is in itself already involves that referential directedness at the world. The firm point in Putnam's argument is that this could not be so, except by magic, if the intrinsic nature of the mental occurrence were constituted by the presence in the mind of a representation, in Putnam's sense. So the possibility that goes missing in Putnam's argument could be described as the possibility of mental representing without representations. Putnam would disput~ something I have been suggesting, that it is just an assumption on his part that the contents of the mind when we think are representations in his sense. His claim is that this thesis is established by introspection. "Stop the stream of thought when or where we will, what we catch are words, images, sensations, feelings."ls (This is meant to be a list of kinds of items that are not intrinsically representational.) But to me it seems wildly inaccurate to suggest that when I am struck by the thought that I hear the sound of water dripping, the fact that my thought is, say, about water is not part of what I find in my stream of consciousness, but has to be read into what I find there. Putnam's phenomenological claim is not an unprejudiced introspective report. It is theory-driven; he tells us not what he finds in his stream of consciousness but what must be there, given the pre-conceived theory that the contents of representing consciousness are representations in his sense. I think an unprejudiced phenomenology would find it more accurate to say that the contents of consciousness, when we have occurrent thoughts, are thoughts themselves, on something like Frege's usage for "thought" (or "Gedanke"): senses potentially expressed by assertoric sentences, not vehicles for such senses. Similarly with imagery: if I close my eyes and visualize, say, my wife's face, it seems wildly wrong to suggest that the fact that what I am visualizing is my wife's face-a fact that relates my mental state to the extra-psychological environment-is extraneous to the contents of my consciousness, extraneous to what I find when I "stop the stream of thought." So far from supporting the apparatus of his argument, Putnam's phenomenological claim here is unconvincing enough to give us reason to raise questions about the theory that underlies the argument. 19 9. Putnam has often expressed suspicion of the idea that there is good philosophy to be done by grappling with questions like "How does language hook on to the world 1"20 It ought to be similar with questions like "How does thinking hook on to the world?" Such a question looks like a pressing one if we saddle ourselves with a conception of what thinking is, considered in itself, that deprives thinking of its characteristic bearing on the world-its being about this or that object in the world, and its being to the effect that this or that state of affairs obtains in the world. If we start from a

44


conception of thinking as in itself without referential bearing on the world, we shall seem to be confronted with a genuine and urgent task, that of reinstating into our picture the way thinking is directed at the world. But if we do not accept the assumption that what thinking is, considered in itself, is a mental manipulation of representations in Putnam's sense, no such task confronts us. The need to construct a theoretical "hook" to link thinking to the world does not arise, because if it is thinking that we have in view at allsay being struck by the thought that one hears the sound of water drippingthen what we have in view is already hooked on to the world; it is already in view as possessing referential directedness at reality.21 It would be a mistake to suppose that what I am doing here is what Putnam describes as "just postulating mysterious powers of mind"; as Putnam says, surely rightly, that "solves nothing."22 The proper target of that accusation is a way of thinking in which we try to combine conceiving the mind as an organ of thought, so that what an episode of thinking is in itself is a mental manipulation of a representation, with supposing that an episode of thinking has its determinate referential bearing on the world intrinsically. Putnam's cogent point is that this combination pushes us into a magical picture of the reference of the supposed mental symbols, and hence into a magical picture of the powers of the mind. But the conception I am urging needs no appeal to a magical theory of reference, precisely because it rejects the supposed mental symbols. My aim is not to postulate mysterious powers of mind; rather, my aim is to restore us to a conception of thinking as the exercise of powers possessed, not mysteriously by some part of a thinking being, a part whose internal arrangements are characterizable independently of how the thinking being is placed in its environment, but unmysteriously by a thinking being itself, an animal that lives its life in cognitive and practical relations to the world. "Just postulating mysterious powers of mind" would be an appropriate description for a misguided attempt to respond to a supposed problem that I aim to join Putnam in rejecting. It would equally be a mistake to suppose that what I have said about the phenomenology of thinking is merely a version of what Putnam calls "the attempt to understand thought by what is called 'phenomenological' investigation."23 Putnam's objection to this is that any such attempt must miss the point that understanding, or more generally the possession of a concept, is an ability rather than an occurrence. "The attempt to understand thought" is the attempt to respond to a philosophical puzzlement about how thought "hooks on to the world." But my aim is to bring out a way of conceiving thought in which there is no need to try to embark on such a project at all. It is true that understanding, or more generally the possession of a concept, is an ability rather than an occurrence. But it does not follow that there cannot be occurrences that are intrinsically directed at reality in the way that I have suggested is characteristic of occurrent thought. If the concept of

45


water is an ability that is exercised in thinking about water, we can conceive its exercises as, precisely, occurrences that are intrinsically episodes of thinking about water. 24 10. What is the attraction for Putnam of the idea that "the stream of thought" is populated by representations in his sense, rather than representings? Any answer must be speculative; an answer that seems to me to have some plausibility is that Putnam is himself swayed by the residual influence of a scientism like the one I mentioned in connection with the "duplex" conception of "wide" psychological attributions. Without the idea of intrinsic structurings in some inner medium, it is hard to see how we could picture a mapping of our psychological talk into a subject matter susceptible of scientific treatment. In particular, mental representings occupy a position in the causal order; and if we want to be able to integrate that fact into a naturalscientific conception of the causal order, it is very tempting to suppose that representings must owe their causal character to the causal character of structures in a medium that is ultimately susceptible of physical description. 25 Putnam's phenomenological claim reflects a plausible conception of the most that could be available to introspection, if we understand introspection as a capacity to scan or monitor such inner structures.26 What goes missing here is the thought that mentalistic talk can be intellectually respectable without any such mapping being needed. I do not suggest that this is an easy thought for us to get our minds around, subject as we are to intelligible pressures to scientize our conception of the causal order. But we ought to ensure that we are fully conscious of the effects of such pressures on our thinking, and we ought to be alive to the possibility that it is not compulsory to succumb to them. The suggestion that Putnam's thinking is partly shaped by a residual scientism will surely provoke from some people the response "So what? What's so bad about scientism?" In another context, I should feel obliged to say someth~ng in answer to that. Here, though, I shall not even begin to do so, since I am confident that that response will not be Putnam's own. Putnam ends "The Meaning of 'Meaning'" with this remark: "Traditional philosophy of language, like much traditional philosophy, leaves out other people and the world; a better philosophy and a better science of language must encompass both. "27 I am not sure how "traditional" the approach to language that Putnam attacks really is, but I do not want to make anything of that here. My point in this paper is that the "isolationist" conception of language that Putnam objects to is all of a piece with a similarly "isolationist" conception of the mind-at least of the mind as it is in itself. And Putnam's attack on the "isolationist" conception of language leaves the counterpart conception of the mind unquestioned. Taking on the whole package would

46


have yielded a deeper understanding of what underlies the "isolationist" conception of language. I think this broader project would have been better suited than Putnam's partial move is to his admirable aim of showing us what "a better philosophy" would be like. A general attack on "isolationism" promises a satisfyingly cohesive and radical reorientation, very much in the spirit of Putnam's own best thinking, of philosophy's approach to the relations between the individual subject and the world.

NOTES 1. Reprinted in Mind, Language and Reality (Cambridge: Cambridge University Press, 1975),215-71. 2. That the extension of a term is determined by its meaning is one of the two assumptions that Putnam plays off against each other in "The Meaning of 'Meaning'." (The other is that "knowing the meaning of a term is just a matter of being in a certain psychological state" [Mind, Language and Reality, 219].) What Putnam argues in the first instance is that the assumptions cannot be true together, and he registers the possibility that one might respond by discarding the assumption that meaning determines extension (e.g., at 266). But his own thinking (much more attractively) leaves that assumption in place. So he directs the argument against the other assumption. 3. Mind, Language and Reality, 227. 4. For some discussion of such details, see the Introduction to Philip Pettit and John McDowell, eds., Subject, Thought, and Context (Oxford: Clarendon Press, 1986),1-15. 5. Putnam considers this at Mind, Language and Reality, 242-45. 6. Mind, Language and Reality, 220. I have slightly altered Putnam's gloss on "the narrow sense" in line with some remarks of Jerry A. Fodor in "Methodological Solipsism Considered as a Research Strategy in Cognitive Psychology," reprinted in his Representations (Hassocks: Harvester Press, 1981),225-53. I think the alteration captures what Putnam intended. 7. See McGinn's "The Structure of Content," in Andrew Woodfield, ed. Thought and Object (Oxford: Clarendon Press, 1982), 207-58. Other considerations are thought to conspire with Putnam's to necessitate this picture, but in this paper I am restricting myself to the significance of Putnam's basic thesis. 8. Cambridge, Mass., and London: MIT Press, 1988. 9. I say "possibly residual" because of the attraction that this conception has for someone of a fundamentally scientistic cast of mind who accepts, perhaps on the basis of an argument like Putnam's about interpretation, that ("wide") content is not available to a scientific psychology. Of course there are people who have a less defensive scientism than that, because they are not persuaded by such arguments, or ignore them. 10. Mind, Language and Reality, 220-21. 11. Mind, Language and Reality, 220. 12. Given a Principle of Charity, this raises a question (which is made all the more pressing by Putnam's own lack of sympathy with "methodological solipsism") whether Putnam may have misinterpreted at least some of the philosophers against whom he directs his basic thesis. I am particularly doubtful about the case of Frege. But I do not want to go into questions about Putnam's reading of his targets here. 13. On the ordinary idiomatic use of "in the head," compare Wittgenstein, Philosophical Investigations (Oxford: Blackwell, 1953), §427.

47


14. Mind, Language and Reality, 22fr27. 15. There is some discussion of issues in this vicinity in my "Singular Thought and the Extent of Inner Space," in Pettit and McDowell, eds., op. cit., 137-68. 16. It is only in connection with mentalism on this interpretation that Putnam considers Gareth Evans's views in The Varieties of Reference (Oxford: Clarendon Press, 1982); see Reality and Representation, 129n. 4. Evans's thinking actually opens up the possibility of a satisfactory understanding of thought (a mental phenomenon, surely) and meaning as environmentally constituted: an understanding that ought to be welcome to Putnam. But Putnam restricts himself to finding it puzzling how Evans could conceive his thinking as a kind of mentalism, since Evans obviously does not equate thoughts with "representations inside the mindlbrain." 17. Cambridge: Cambridge University Press, 1981. 18. Reason, Truth and History, 17; see also 27 for a parallel appeal to introspective evidence. 19. For a "cry of disbelief' (69) against similar phenomenological falsifications, forced on philosophers by the theory that "an occurrent conscious thought bears its 'intention' or content in the same way as a bit of language bears its significance" (86), see M. R. Ayers, "Some Thoughts," Proceedings of the Aristotelian Society n. s.lxxiii (1972-73): 69-96. One of Ayers's targets is Wittgenstein; I suggest a rather different reading of Wittgenstein (although I would not dispute that there are passages that fit Ayers's reading) in "Intentionality and Interiority in Wittgenstein," in Klaus Puhl, ed., Meaning Scepticism (Berlin and New York: De Gruyter, 1991), 148-69. 20. His suspicions are expressed in several of the essays in his Realism with a Human Face (Cambridge, Mass.: Harvard University Press, 1990). 21. And the world that it is already hooked on to is not The World as contemplated by the metaphysical realism that Putnam has attacked. My thought that I hear the sound of water dripping has its point of contact with reality in the fact that I hear the sound of water dripping, or perhaps in the fact that I do not hear the sound of water dripping. I use my conceptual capacities (I just did) in pinpointing which possible facts these are; the world (which is all the facts, as Wittgenstein said in the Tractatus) is not here pictured as beyond the reach of concepts. 22. Reason, Truth and History, 2. 23. Reason, Truth and History, 20. 24. These remarks are directed against the close of chapter 1 of Reason, Truth and History, where Putnam suggests that the perfectly correct point that concepts are not mental occurrences, combined with the phenomenological claim about which I have already expressed doubts, demolishes the very idea that there can be mental episodes with an intrinsic referential bearing on the world. By claiming that concepts are "signs used in a certain way" (18), Putnam makes it look as if exercises of concepts would have to be occurrences (tokenings) of signs. He thereby forces on us a "narrow" conception of what exercises of concepts must be in themselves. This obliterates a perfectly workable conception according to which exercises of concepts are, for instance, acts of jUdgment, intrinsically possessed of referential bearing on the world. 25. See John Haugeland's suggestive discussion of "the paradox of mechanical reason," in Artifical Intelligence: The Very Idea (Cambridge, Mass., and London: MIT Press, 1985),

3fr41. 26. For an unusually explicit expression of such a view of introspection, see McGinn, "The Structure of Content," 253-54. 27. Mind, Language and Reality, 271.

48

Anderson, David L., What Is Realistic about Putnam's Internal Realism? , Philosophical Topics, 20:1 (1992:Spring) p.49

PHILOSOPHICAL TOPICS 20 No.1, SPRING 1992

VOL.

What Is Realistic about Putnam's

Internal Realism?1

David L. Anderson Illinois State University

In 1976, Hilary Putnam, longtime champion of metaphysical realism, startled the philosophical community by abandoning metaphysical realism and offering his own alternative which he has been elaborating and defending ever since. Putnam makes an interesting claim about his new position: He insists that the appropriate description of the view is "internal realism" (or sometimes "pragmatic realism") and that it is a view that a person motivated by "the realistic spirit" might justifiably hold. Very few contemporary realists are sympathetic to Putnam's suggestion that his view is "realistic." Admittedly, Putnam's position does boast a rich ontology. Electrons exist every bit as much as chairs and tables do, and electrons can even help to explain the superficial properties of macro-objects. Few realists, however, are willing to count this as a sufficient condition for being a "realist." After all, Putnam insists that ontological commitment is always internal to a conceptual scheme; there is no scheme-independent fact of the matter about the ultimate furniture of the universe. Putnam, then, is no more a realist than is Kant-and for many contemporary philosophers, that is to be no realist at all. Failure to recognize the "realistic" motivations for Putnam's rejection of metaphysical realism has led to a widely shared misunderstanding of Putnam's arguments against metaphysical realism. Realist critics of these arguments, convinced that they pose no serious threat to their views, typically

49


offer rebuttals that are either unapologetic appeals to realistic intuitions or demonstrations that the arguments in question are simply misdirected and thus irrelevant to the truth of metaphysical realism. Responses of this kind do not defeat Putnam's arguments; they fail to confront them. A principal shortcoming of these standard interpretations is their disregard for the role that these arguments play in the overall case that Putnam is building against metaphysical realism, a disregard which ultimately deflects their argumentative force. Simply put, Putnam's arguments are intended to show that metaphysical realism itself is not sufficiently realistic. If that claim can be substantiated then Putnam can go on to argue that his own view is, in relative terms, more realistic than metaphysical realism. 2 Such widespread misunderstanding of Putnam's arguments requires more than a simple corrective-it requires an explanation. I will attempt to provide that explanation in the first half of this paper (Sections I-III). Attention will be focused upon several popular interpretations of Putnam's favorite arguments. These interpretations are not wildly irresponsible; they are, in fact, quite natural and reasonably motivated. Understanding where and why they miss the intended force of Putnam's arguments will reveal something of interest about the nature of the realism-antirealism dispute itself. Once we have understood why Putnam's arguments receive the interpretations that they do, we have the background against which the arguments can be properly understood-as reductios. The second half of the paper (Sections IV-VI) will make clear why Putnam thinks that metaphysical realism is not sufficiently "realistic" and why he judges his own internal realism to be more in keeping with the "realistic spirit." This also provides an answer to a question that, on standard interpretations of Putnam's arguments, remains altogether mysterious: Why would Putnam-blessed with strong "realist" intuitions-have abandoned realism on the force of these arguments? My interest in all of this is not to defend internal realism. Putnam's arguments notwithstanding, I have not myself abandoned metaphysical realism. However, he has forever changed my relationship to realism. First, because I think that he is absolutely right that many popular contemporary versions of realism violate the "realistic spirit" in ways too little appreciated. A version of realism that does not compromise this tradition is a robust kind, out of favor with many contemporary physicalist-realists and not easily defended. And second, because he is also right that realism can be alienating. 3 We need a philosophical perspective that does not undervalue the "human perspective" not simply because our lives are more apt to flourish if we do, but because, as I have argued elsewhere,4 the failure to give proper place to the epistemic perspective of human beings inevitably leads to false semantic theories.

50


I. METAPHYSICAL REALISM Putnam has argued that metaphysical realism, while initially appealing to the realistic spirit in us, can be shown to be irreconcilable with that spirit. I define "metaphysical realism" as a conjunction of commitments---ontological, semantic, and alethic. To be a metaphysical realist on my account is to be committed to the following three metaphysical tenets of realism: (Ml) Correspondence Truth. Truth is a relation of correspondence between pieces of language and the world (i.e., ding-ansich-reality). A statement is correspondence-true iff it bears the (unique) relation "correspondence" to ding-an-sich-reality. (M2) Semantic Realism. Statements that express an existential commitment to concrete objects (middle-sized and theoretical) will be true or false in virtue of the intrinsic nature of mindindependent reality, and thus in virtue of conditions the obtaining of which may be, in principle, inaccessible to human beings. (M3) Ontological Realism. All (or most) of the objects (middlesized and theoretical) countenanced by twentieth-century science and common sense exist independently of any mind. s

While Putnam's definition of metaphysical realism is slightly different from this one,6 these are the tenets to which metaphysical realists typically have the fiercest loyalty, and Putnam offers reasons for rejecting all three. 7 When he abandoned realism, Putnam embraced a view he calls "internal realism," supplanting (M I) with what can broadly be considered a coherence theory of truth, deposing (M2) for an idealized verificationist semantics, and replacing (M3) with an "empirical" realism that is idealistic in the Kantian tradition. There are a variety of arguments that brought him to his conversion. I shall consider three of the most famous: Dummett's "language acquisition argument" (as I shall call it), Putnam's "model-theoretic argument," and Putnam's "brains in a vat argument." My aim here is not to provide a detailed reconstruction of these arguments-such reconstructions are available in abundance. Rather, I will provide a brief sketch of the basic intuition behind each of the arguments and a sampling of typical and predictable realist responses to them. The purpose of this survey is to show how remarkably easy it is for realists to resist these arguments, relying primarily on two distinctive strategies: The first, the "Burden of Proof' gambit (the subject of Section II) calls for realists to simply refuse to accept the burden of proof which the language acquisition argument and the model-theoretic argument are (purportedly) attempting to place upon the realist; the second, the "Shielding Metaphysical Realism From Attack" strategy (the subject of Section III) demonstrates how easily the force of both the model-theoretic

51


argument and the brains in a vat argument can be directed away from metaphysical realism, preserving it from harm. These two responses to Putnam's arguments are so widespread (coming from realist and antirealist commentators alike) that it is important to appreciate the very reasonable assumptions which ground them.

II. THE BURDEN OF PROOF One of the arguments responsible for Putnam's rejection of metaphysical realism is Michael Dummett's language acquisition argument. For the past twenty-five years Dummett has forcefully argued that realism cannot ultimately be reconciled with any plausible theory of understanding. How is it that we come to learn a language, to learn what it means to assert that p? On one reasonable account we learn to recognize the conditions under which competent speakers of a language are disposed to assert that p. That is, we acquire the skill of recognizing, from among the conditions that are epistemically accessible to us, those conditions which must obtain for p to be assertible. Now, i,f what it means to assert that p just is to assert that justification (or verification) conditions (possibly of an idealized sort) obtain, then we have at least the beginning of a theory of how a finite languagespeaker could learn the meaning of p. If, however, these conditions of assertibility are at most evidence of the truth of p, where the truth of p consists in something entirely non-epistemic-say, for example, "bearing the relation of correspondence to mind-independent reality"-then, it would seem, we are nowhere near a theory which could account for how finite languagespeakers could come to grasp realist truth-conditions and thus know what p means. If realist truth-conditions could be learned, it is reasonable to think that we would have some idea about how such learning takes place. Since we don't know how we could have the capacity, then we probably don't have it. This is an argument which, according to Putnam, played a prominent role in his conversion to internal realism. At a crucial moment, when Putnam was struggling with the obstacles that a substantive correspondence theory of truth must ultimately overcome, he spent time (in Jerusalem) discussing these issues with Dummett and became convinced that Dummett was, in the main, correct. In his 1976 Presidential Address to the APA, where he announced his conversion, Putnam says: The point is that Dummett and I agree that you can't treat understanding a sentence (in general) as knowing its truth conditions; because it then becomes unintelligible what that knowledge in tum consists in. We both agree that the theory of understanding has to be done in a verificationist way .... But now it looks as if ... I have given Dummett all he needs to demolish metaphysical realism-a picture I was wedded to!8

52


While there are some who find Dummett's argument compelling, many a realist finds this little more than an attempt to shift the burden of proof, unjustifiably, to the realist. If realism is true, says the realist, then an empirical theory of understanding of a Dummettian kind is just what one would not expect to be available. Just as in other philosophical disciplines--ethics, epistemology, logic, etc.-the strategy is to refuse to concede the skeptic any ground. For purposes of comparison, consider the skeptic about the external world and his stance on the skeptical conditional (S) If I am deceived by an evil demon, then this is not a hand.

The skeptic insists that modus ponens, the closure principle, and our inability to know that the antecedent of (5) is false demonstrates that we do not know that this is a hand. Is it incumbent upon the realist to give an argument to show that the antecedent is false, or otherwise to abandon the claim that she knows that this is a hand? Many realists about the external world have insisted that the answer is, "No!" In the spirit of G. E. Moore, the realist is apt to say that modus tollens, the closure principle, and the known falsity of the consequent of (5) demonstrates that we know that we are not being deceived by an evil demon. This is an instance of the most common kind of philosophical standoff. (As Putnam himself is fond of saying, "One person's modus ponens is another person's modus tollens.") The same strategy is available in the case of the Dummettian challenge over a theory of understanding. Coincidentally, the skeptical conditional (5) plays a role in this dispute as well. For realists, (5) expresses a coherent skeptical worry; for antirealists, since the antecedent is (by stipulation) not verifiable, the conditional is trivially false. Thus, the conditional around which the dispute over a theory of understanding revolves is as follows (note that (5) is mentioned in the consequent): (U) If there exists no true theory specifying how finite languagespeakers grasp realist truth-conditions, then human beings cannot grasp the meaning of (S) realistically interpreted.

The antirealist argues that the lack of a plausible theory of understanding makes untenable the realist's claim that she does in fact understand the meaning of (5) as a genuinely skeptical conditional. The realist, of course, insists that the undeniable fact that she understands the meaning of (S)-as a skeptical hypothesis and, thus, realistically interpreted-justifies the claim that there exists a true realist theory of meaning even if we are never able to articulate it to the verificationist's satisfaction. (This is the "modus tollens" strategy: Deny the consequent and infer the falsity of the antecedent.) Consider the following expression of this "modus tollens" strategy: The "demon hypothesis" is not just a noise that happens to evoke some "pictures in the head"; it is a grammatical sentence in a

53


language; it is one we can offer free translations of; it is subject to linguistic transfonnations; we can deduce other statements from it and also say what other statements imply it; we can say whether it is linguistically appropriate or inappropriate in a given context, and whether a discourse containing it is linguistically regular or deviant. The verificationists would retort: "It doesn't follow it has meaning". But they would just be wrong, for this is just what meaning is: being meaningful is being subject to certain kinds of recursive transfonnations, and to certain kinds of regularities; we may not know much more about the matter than that today, but we know enough to know that what the verificationists were propounding was not an analysis of meaning but a persuasive redefinition.

Where does the burden of proof ultimately lie on the issue of the coherence of semantic realism? Must the antirealist give some knockdown argument (not yet in evidence) to shake the realist's strong intuition that she can entertain the "demon hypothesis"? Or is the burden of proof on the realist to produce some plausible theory to support the claim that human beings have the capacity to grasp realist truth-conditions? The passage just quoted is of particular interest because Putnam himself is the author, circa 1971.9 At this point in time, Putnam knew of Dummett' s arguments, but he did not believe that these placed the burden of proof onto the semantic realist. We may not know a great deal, Putnam argues here, "but we know enough ... " The burden, he thought, lay with the antirealist. By 1976 he had changed his mind. Why? What convinced him that the burden of proof had shifted? Ultimately, we will attempt to answer that question. First, however, it is important to understand why Putnam's own arguments tend to receive a similar response-i.e., the burden of proof gambit-from present-day realists.

THE MODEL-THEORETIC ARGUMENT Complementing Dummett's language acquisition argument is Putnam's model-theoretic argument. 10 Putnam has contributed a whole family of arguments that advance what we might call the "problem of reference" objection to metaphysical realism. Each of these arguments is intended to show that literally nothing that twentieth-century philosophers believe in could fix the reference of terms of a human language to mind-independent objects. And if that is the case, we lack the capacity to either speak or think about mindindependent objects and the traditional realist picture collapses. Putnam has advanced various versions of this argument, 11 and there is considerable disagreement over the proper reconstruction of the details. It will suffice for this discussion, however, to offer the basic intuitive argument that requires little in the way of complicated machinery.

54


According to the correspondence theory of truth there is no logical relation between the truth of a statement and the justification of it. This generates a principled distinction between how the world is in itself (independent of the human perspective) and how the world appears (even in the long run) from the human perspective. If this distinction can be collapsed, then correspondence truth will be undermined and metaphysical realism with it. Putnam argues that a necessary condition for correspondence truth and the accompanying distinction is the existence of a determinate relation of reference between our words and mind-independent objects. Otherwise there simply could be nothing in virtue of which an epistemically ideal theory could come out false. The -core of his argument is quite simple: Assume that we have a theory, T, which meets all theoretical and operational constraints. T is then epistemically ideal. Now, if truth is correspondence to mindindependent reality, it is possible that T is false. If it is false, then reality will not be as T says it is. For example, assume that T is committed to the existence of tables, electrons, and planets, among other things. Reality, let us assume, includes no such entities, but is populated instead by F's, G's, and H's, entities radically different from anything countenanced by T. On the face of it, it seems that the situation we have just described underwrites the metaphysical realist picture. Putnam demurs, and asks: How can we be sure that T does not countenance F's, G's, and H's? We know that there will be an interpretation of the sentences of T according to which terms of T (terms like 'table,' 'electron,' and 'planet') refer to objects in the real world, for example, F's, G's, and H's. We also know that there will be an interpretation of this kind such that T comes out true. If there is an interpretation of T which not only meets all theoretical and operational constraints but also comes out true, then in virtue of what could it fail to be the proper interpretation of T? The realist will be quick to insist that an interpretation according to which 'table' refers to F's is a deviant interpretation: it is not the one intended by the speakers. Putnam has a ready reply. He demands that an account be given of what there could be (under heaven and earth) that could conceivably fix the so-called "intended" interpretation. It will not do, in the late twentieth century, to appeal to "magical" theories of reference or occult properties. The first move available to the realist is the "burden of proof' gambit which requires nothing in the way of philosophical argument. One simply refuses to admit that the burden of proof is on the realist. "Why," asks Carsten Hansen, "does Putnam think that it is incumbent upon realism to prove that our terms have determinate extensions?" 12 David Lewis states the matter more bluntly: Since Putnam's paradoxical thesis is patently false, we can be confident that there is indeed some further constraint, whether or not we can find out what it is. 13

55


The model-theoretic argument poses no serious threat, in Lewis's mind. Since he takes it to be an indisputable fact that our words have determinate reference, we are confronted only with a paradox that we may attempt to resolve or not as we wish. Realist objectors to the model-theoretic argument rarely leave the matter here, however. Most, including Lewis, are prepared to meet Putnam's demand by offering some account of what the reference-fixer might actually be. Putnam separates these theories into two broad categories: (a) those candidates that he considers to be scientifically respectable but incapable of actually fixing reference, and (b) those candidates that might have the metaphysical clout to fix reference but which require-precisely because of their metaphysical extravagance~ommitment to entities, properties, or powers of dubious scientific merit. The first category includes the most popular candidate for reference-fixer, "causal connections of the appropriate type." The second category includes solutions which, to Putnam, smack of "medieval essentialism"14 or "magical" powers. IS Two realist responses that apparently fall into the latter category are (i) David Lewis's suggestion that there exist "elite properties" (properties that cut reality at the joints) which provide an additional constraint on reference,16 and (ii) Alvin Plantinga's simply stated claim that we have the power to "grasp properties."17 Putnam gives little attention to these unabashed pronouncements that there exists potent metaphysical machinery which can fix reference. Instead he focuses upon the one candidate which (in his opinion) has, because of its modesty, the potential of being part of a plausible empirical theory with scientific respectability and genuine explanatory power. This is the view that causal connections provide the additional constraint that will enable reference to be fixed. Putnam himself used to assume that causal connections were the most likely answer to the reference-fixing problem. Given the model-theoretic argument, however, he became convinced that the causal theory lacks the necessary resources to meet the challenge. How, exactly, is a causal connection supposed to accomplish the task? Certainly we can add the sentence " 'Cat' refers to mind-independent cats by virtue of a causal connection of the appropriate type" to our theory, T. But this is just "more theory" according to Putnam. 'Causal connection' is just another linguistic symbol which itself will lack determinate reference without an acceptable reference-fixer. It simply begs the question to assume that your language has a determinate reference while you are in the process of showing how that reference is fixed. Or does it beg the question? The realist will insist that it is not the linguistic symbol 'causal connection' that fixes reference, it is causal connections themselves that do the work. To assume that sentences which the realist uses to express her theory of reference-fixing do not have determinate reference is itself to beg the question against the realist. Michael Devitt expresses it this way:

56


Putnam, in effect, accuses the metaphysical realist of begging the question in appealing to a theory to detennine reference for a theory. I have accused him of begging the question in claiming that the reference of 'causally related' is not detenninate. 18

Again the question arises, where does the burden of proof lie? Is producing a sentence about "causal connections" sufficient? Even if causal connections are considered to be possible reference-fixers, has anything really been explained? Aren't causal connections at best brute facts with little explanatory value? Isn't the burden of proof on the realist? No, argues Anthony Brueckner, We maintain that the use of language together with non-linguistic facts about the world (e.g., causal relations between the world and the use oflanguage) do fix the intended interpretation of the language. Unless Putnam can discredit the foregoing claim, he has given us no reason to suppose that every attempt at a linguistic specification of a theory's intended interpretation must fail. 19

Brueckner implies that Putnam has given no substantive argument to discredit the causal theory and so he has given the realist no reason to abandon his position. The purpose of this inquiry is not to arrive at a judgment on the merits of either Putnam's argument or this particular realistic response to it. Rather, it is to make explicit how the realist perceives her position and why she considers the "burden of proof' response a natural one. After all, it seemed natural to Putnam himself in 1971. If the realists' initial convictions are reasonable (as they obviously assume they are), then any argument that is interpreted as nothing more than a demand for a more sophisticated theory will hardly be received as a demonstration that metaphysical realism is incoherent. If one comes to the table assuming that the burden of proof is on the antirealist to give a knockdown refutation of realism, it is little wonder that the language acquisition argument and the model-theoretic argument are perceived to have so little force.

III. SHIELDING METAPHYSICAL REALISM FROM ATIACK The "burden of proof' gambit is basically a justification for ignoring an argument, denying that it provides an objection substantial enough to warrant rebuttal. Realist critics of Putnam are, however, rarely satisfied with a bald dismissal of his arguments. They are more than happy to provide a careful analysis of the details of each argument. Such an analysis frequently produces the following results: It is found that the arguments, even if sound, tum out not to entail the falsity of metaphysical realism. Since the arguments are

57


not strictly about metaphysical realism, why would they give Putnam or anyone else reason to abandon that view? Employing this second strategy, one simply protects metaphysical realism whatever the cost. David Lewis argues that even if Putnam is correct, even if reference is not fixed by anything and thus is indeterminate, realism still survives. The model-theoretic argument is simply about the wrong thing to be successful against metaphysical realism. Putnam has greatly overstated the consequences that would follow even if it were shown that reference is radically indeterminate. He explains it this way: My point is rather that even if the model-theoretic argument worked, it would not blow away the whole of the realist's picture of the world and its relation to theory. Something vital would be destroyed, but a lot would be left standing. There would still be a world, and it would not be a figment of our imagination. It would still have many parts, and these parts would fall into classes and relations-too many for comfort, perhaps, but too many is scarcely the same as none. There would still be interpretations, assignments of reference, intended and otherwise. Truth of a theory on a given interpretation would still make sense, and in a non-episternic way. Truth on all intended interpretations would still make sense. Despite Putnam's talk of the 'collapse' of an 'incoherent picture', he has given us no reason to reject any of these parts of the picture. The only trouble he offers is that there are too many intended interpretations, so that truth on the intended interpretations is too easily achieved. That is trouble, sure enough. But is it anti-realist trouble, except by tendentious definition? ... The metaphysics of realism survives unscathed. What does suffer, if Putnam has his way, is realist semantics and epistemology.20

Lewis's basic idea, I take it, is that if the heart of realism is a commitment to correspondence truth and to a mind-independent reality to which true propositions correspond, then acceptance of the model-theoretic argument does not force one to abandon realism. Our language will lack a unique, intended interpretation and so each sentence of our language will express a whole host of propositions, but this is not meaning skepticism in the sense that we simply don't know which unique proposition is being expressed; this is meaning pluralism in the sense that each sentence expresses as many propositions as there are interpretations of T which meet theoretical and operational constraints. This picture does not abandon correspondence truth and mind-independent reality; it literally presupposes it. The result, of course, will be a raging semantic pluralism that hardly leaves traditional realist semantics in place and wreaks considerable havoc with epistemology. Still, "The metaphysics of realism survives unscathed," according to Lewis, and since it is intended to be an argument against metaphysical realism, the

58


model-theoretic argument misses its mark. Admittedly, Lewis would have to pay a fairly high price to continue to love metaphysical realism in the face of this rather unseemly and promiscuous semantics. Yet love can make a person do remarkable things, and a true love of metaphysical realism can be sure to inspire tremendous feats of philosophical courage. The model-theoretic argument clearly lacks the necessary clout to force one to abandon metaphysical realism. We have arrived at an important juncture. Having seen that the "burden of proof' gambit is altogether reasonable from the realist's perspective and having seen that metaphysical realism can be preserved come what may, the following seems unavoidable: If the success of Putnam's arguments depends. upon either (a) shifting the "burden of proof" by simply demanding more theory from the realist, or upon (b) forcing the realist to abandon metaphysical realism on pain of contradiction, then his arguments must fail. Metaphysical realism can be maintained without logical contradiction even if reference is maddeningly pluralistic; the burden of proof cannot be shifted onto the realist simply because the antirealist would like a bit more theory. While this may seem to put Putnam at a distinct disadvantage vis-a-vis his realist opponents, such is not the case. Not only are these concessions not fatal to Putnam's arguments, but recognizing that Putnam is happy to grant these concessions is a prerequisite for understanding his arguments. Putnam has been there before. He knows all too well that these dialectical moves are available to the realist. (He has, after all, used them himself!) The case he has built against metaphysical realism is not merely a demand for more theory, nor is it the bald claim that metaphysical realism entails a logical contradiction. The main thrust of his argument lies elsewhere. To better appreciate the spirit of Putnam's arguments, consider Lewis's suggested strategy against a successful model-theoretic argument. Lewis says that one ought simply to marshall all available defenses and protect metaphysical realism from attack. Instead of allowing metaphysical realism to take the "hit" from the argument, sacrifice instead one's semantic and epistemological commitments. There is no logical constraint to prohibit this strategy, and all true realists will recognize their responsibility to preserve metaphysical realism come what may. First, it is important to notice the very different thresholds of philosophical suffering that individuals are will- . ing to endure before crying uncle and abandoning their basic presuppositions. What does the model-theoretic argument show if successful? Does it show (A) that we may have the inconvenience of "too many" intended interpretations, or (B) that the notion of "intended interpretation" collapses into incoherency on the metaphysical realist picture? Does it show (A) that metaphysical realism may reasonably be preserved as one's defining paradigm so long as certain semantic and epistemological corrections are made, or (B) that the altogether bizarre and counterintuitive consequences in semantics

59


and epistemology make it an effective reductio of metaphysical realism itself? We know that philosophers can stick to their guns come what may, so long as they are willing to live with the consequences. Lewis, as we know, is unlikely to be deterred by our incredulous stares. Lewis's strategy to protect metaphysical realism come what may provides a convenient foil to show off Putnam's arguments to their greatest advantage, revealing the role that these arguments ultimately play in Putnam's case against the classic realist picture. Putnam is well aware of the fact that the "burden of proof' gambit carries little conviction by itself. If a person has strong intuitions in favor of metaphysical realism and if that person detects no serious liabilities that cause unavoidable conflicts among other of her convictions, then Putnam's arguments will hold little conviction. Putnam assumes, however, that there are in fact deep and irreconcilable tensions within the traditional realist perspective--evidence that there is something disturbingly "anti-realistic" about the view. If this can be established then he may successfully shift the burden of proof by showing that metaphysical realism has absurd and intolerable consequences (even if it cannot be proven that it is logically inconsistent). Consider a parallel situation. In normal science there is often a lack of a plausible theory-with no immediate hope of a breakthrough-right where scientists would most like to have one. This need not jeopardize fruitful research, nor pose any threat to accepted theories. Historians of science have helped us to see that normal science continues, that the overall theoretical assumptions that have been successful in the past continue to be employed without question and without (additional) justification in spite of data that do not fit, and in spite of theories that never materialize. Bona fide crises do not arise merely for the lack of answers to theoretical questions; crises arise only when there is some tension within the accepted theory which is intolerable, some tension which is increasingly detrimental to the fruitful practice of normal science. I want to suggest that this is an apt model for the present case. It is true that the realist may not have as thorough and sophisticated a theory about realist semantics and correspondence truth as she might like. However, too little theory, alone, rarely precipitates a crisis; it often does no more than . motivate more "normal philosophy." Crises arise not when there is too little theory but when deep conflicts are detected within the theory, conflicts that seem irreconcilable without a radical adjustment in the defining paradigm. The force of Putnam's case against realism is not simply a demand for more theory from the realist. The arguments gain their persuasive force against a background of unanticipated anomalies that Putnam believes makes a "revolution" the only reasonable course. It is commonplace for realists

60


today to think that the causal theory will take care of the problem of reference and that traditional realist commitments will be preserved. Putnam thinks otherwise. There is a crisis within the realist paradigm, he argues, and there is simply no way to preserve the traditional picture intact. It may be possible to maintain a commitment to metaphysical realism without falling into logical contradiction, but one will be forced to abandon many strongly held convictions and, if Putnam is correct, protecting metaphysical realism is neither a rational nor a "realistic" thing to do. Most philosophers will agree, I think, that realism in the modem period was not simply a narrow, isolated commitment to metaphysical realism, i.e., (Ml)-(M3). Traditional realists have always taken for granted that one of the primary virtues of correspondence truth is that it seems to guarantee a particular commonsense view of semantics and epistemology. If it turns out that correspondence truth does no such thing, few realists, I think, would possess Lewis's steely nerve in the face of adversity and be willing to abandon these strongly held convictions. If the model-theoretic argument did obviously succeed, Putnam would have a compelling objection to metaphysical realismnot because metaphysical realism together with radical indeterminacy constitute a logical contradiction, but because they are incompatible with strongly held convictions-semantic and epistemological-that are an inseparable part of the traditional realist picture. Lewis's argument, then, is a perfect example of the kind of pressure that Putnam seeks to place on traditional realism. It is only necessary that he show that metaphysical realism has serious, possibly intolerable, consequences in some part of the traditional realist's philosophical picture. Even if we assume that this analysis is accurate, little has been done to actually further Putnam's case. No realist that I am aware of (including Lewis) is willing to grant that the model-theoretic argument is successful in the least. As we have already seen, the realist assumes that there is no good reason to doubt referential determinacy and that Putnam's argument is at best the unreasonable demand that the realist say something more about how causal relations fix reference and at worst the impossible demand that the realist say it in a way that doesn't presuppose that reference is determinate. All of this notwithstanding, Putnam believes that his arguments do reveal a crisis of irreconcilable tensions and anomalies wjthin the traditional realist picture. The realist, for her part, is having considerable difficulty understanding how the model-theoretic argument does anything but beg the question. Putnam anticipates this difficulty, however, and so offers another argument, the brains in a vat argument, as a practical demonstration of the unexpected consequences that correspondence truth has in the areas of semantics and epistemology.

61


THE BRAINS IN A VAT ARGUMENT The brains in a vat argument, if it is effective, will make manifest certain irreconcilable tensions that lie at the heart of the standard realist picture. It will help to show something that the model-theoretic argument alone could not: That even if we assume that "causal relations of the appropriate type" do fix reference, there are consequences that follow in the area of semantics and epistemology that are not reconcilable with strongly held convictions of most traditional realists. Realists have had great difficulty interpreting the brains in a vat argument as such an argument. First a brief summary of the argument. One crucial premise of the argument is that we postulate (what Putnam considers to be) a plausible theory of reference. Appealing to insights that he helped to popularize in his work on the new theory of reference, Putnam insists that there is a condition that must be met for successful reference-the causal requirement: One cannot refer to certain kinds of things, e.g. trees, if one has no causal interaction at all with them, or with things in terms of which they can be described. 21

Invoking this requirement, Putnam argues that a brain in a vat-whose experiences are the result of computer-generated electrical stimulation and who bear no causal relation to real trees~annot refer to real trees regardless of what mental images the brain produces and regardless of how many tokens of the symbol-string 't-r-e-e' it produces in its computer-generated environment. Thus, an envatted brain asserting "I am a brain in a vat" does not refer to the real vat in which it is suspended nor to the real brain that it is, since it does not bear a causal relation of the appropriate type to either one. What does it refer to? Well, possibly to a feature of the computer program, or to the electrical impulses that cause the brain's experiences, or to the phenomenal images themselves-"tree-images", "vat-images", and the like. Putnam's enticing conclusion is that we can know, therefore, that we are not BIV s. If I exist in a normal world consisting of real trees, real brains, and real vats, then the English sentence, "I am a brain in a vat," does express a claim about real brains and real vats, a claim which is, however,false. If, on the other hand, I actually am a brain in a vat, bearing no (appropriate) causal relation to real vats, then my utterance of "I am a brain in a vat" is ultimately a claim about electrical impulses or vat images. In that circumstance, the statement that I would express by such an utterance would be something like, "I am presently having experiences as of being a brain in a vat." But of course if this is what it means, it is straightforwardly false. With this constructive dilemma, we see that the skeptical threat is disarmed. I can be assured that the hypothesis expressed by "I am a brain in a vat" is false and thus I know that I am not a brain in a vat.

62


If this argument is to prove effective against the realist, it must reveal some consequence of metaphysical realism that the realist herself finds unacceptable. There has been a very common and predictable response to this argument, however, according to which it is no threat to realism whatsoever. While there are numerous articles that make exactly the same claim,22 I shall consider Anthony Brueckner's version because I find his overall approach to the argument especially instructive. Brueckner admits that there is some question about the force with which the argument is to be taken. The conclusion of Putnam's argument seems to be: That I am a brain in a vat is not a logically possible proposition. This conclusion does not follow from the premises, says Brueckner, and Putnam is wrong when he says In short, if we are brains in a vat, then 'We are brains in a vat' is false. So it is (necessarily) false. (15)

Brueckner takes Putnam's last sentence to express the claim that there exists a proposition (i.e., the proposition that is expressed when a person in the real world utters the sentence "I am a brain in a vat") such that that proposition is necessarily false. How could that be true? Brueckner asks. If it were, then there would be no possible world in which I am a brain in a vat. But the brain in a vat scenario offered by Putnam himself just is such a logically possible world. No, Brueckner insists, Putnam has not shown that the relevant proposition is necessarily false, but only that a sentence-a sentence that expresses different propositions in different contexts-will express a false proposition in all possible worlds in which the sentence is betokened. All I can claim is the metalinguistic knowledge that a certain sentence expresses a false proposition, rather than the objectlanguage knowledge that I am not a brain in a vat. Since the latter knowledge was required in order to refute the skeptical argument in the envisaged manner, the present anti-skeptical strategy fails. 23

Brueckner rightly detennines that one of the premises of the BIV argument presupposes metaphysical realism, just as Descartes's evil demon scenario does. He concludes, then (somewhat hesitantly), that the argument purports to show metaphysical realists that there is a "tricky" way to solve Descartes's skeptical dilemma. 24 The basic strategy of the BIV argument, if Brueckner is correct, is something like this: 1. Assume that metaphysical realism is true. 2. Describe a brains in vat scenario. 3. Show that no one can truly believe that she is a brain in vat. 4. Claim to have provided a clever way of undermining Cartesian skepticism.

63


5. Conclude, therefore, that even if metaphysical realism is true radical skepticism is not a real threat.

First it is important to recognize the parallel between this response to the BIV argument and Lewis's claim that the model-theoretic argument, even if successful, does not count against correspondence truth. Both responses refuse to countenance an interpretation of the argument that makes it a reductio. In Lewis's case the refusal is simply a bravura "in your face" stance which he (gleefully) takes against Putnam. In Brueckner's case there is some reason to suspect that he doesn't actually recognize that it is intended as a reductio. Throughout the paper, he admits his uncertainty about this interpretation and in a footnote to the first paragraph, he says: The argument of chapter I ["Brains in a Vat"] indeed depends upon causal-theoretic assumptions about reference which Putnam explicitly rejects in chapters 2 and 3. At no point, however, does he explicitly entertain the possibility that it is a reductio and thus he makes no attempt whatever to consider what force it might have, so interpreted. It is quite common, and not a wholly disreputable business, for analytic philosophers to consider a philosophical argument as an isolated text, as if it washed up on the beach somewhere with no clues as to its intended argumentative force. However, in this case, where the argument clearly presupposes metaphysical realism, and where the book, of which it is the opening chapter, is clearly committed to demonstrating the bankruptcy of metaphysical realism, it is difficult to miss the role that the argument is meant to play. Putnam makes it quite clear when he asks: Why is it surprising that the Brain in a Vat hypothesis turns out to be incoherent?25 Brueckner's interpretation is puzzling. How could it seem reasonable to interpret this argument as a straightforward rebuttal of Cartesian skepticism, presupposing the truth of metaphysical realism? The answer, I suspect, is that many commentators simply cannot figure out how the argument could plausibly have any force as a reductio against someone strongly committed to metaphysical realism. To substantiate this claim, consider Peter van Inwagen's interpretation of the argument. He rightly concludes that a proper interpretation of the argument must not ignore the fact that the BIV argument is advanced as an argument against metaphysical realism. He characterizes the argument this way: (1) If truth is radically nonepistemic, then it is intelligible to suppose that we are all wrong about almost everything. (2) It is not intelligible .10 suppose that we are all wrong about

almost everything.

.

hence, Truth is not radically nonepistemic. 26

64


While conceding that Putnam never makes this argument explicit, he says "But if he does not mean to be defending this argument, or some argument very much like it, ... then I do not see the point of that chapter." This is a good beginning, if only because the conclusion is consistent with the stated purpose of the book. Van Inwagen is, then, asking the right question: Why would Putnam (or anyone else) think that the BIV argument raises any difficulty for the metaphysical realist? Van Inwagen offers a reasonable answer to this question. The argument will have force against the metaphysical realist because realists have traditionally believed that (I) is true and now Putnam has given us reason to believe that (2) is true (because, given the causal requirement, even brains in vats will not be wrong about everything). Van Inwagen's response to the argument is predictable. He argues that even if one grants Putnam premise (2), for the sake of argument, one still need not concede the conclusion. Premise (1) is still vulnerable and, if one is really committed to nonepistemic truth, it will be sacrificed. Realists may have thought that (I) is true-and even with some justification. But as Putnam points out, realists also have tended to believe that "what goes on in our heads must determine what we mean and what our words refer to."27 If that belief proves unwarranted, then premise (1) will lose much of its intuitive force. The obvious 'strategy for realists, van Inwagen argues, is to give up (I) and hold on to correspondence truth. What could it hurt? Why shouldn't the "copyist," when he has seen Putnam's argument for the incoherency of supposing that we are brains in vats simply say, "Oh, now I see," and continue to be a copyist? Why shouldn't the proponent of the thesis that truth is radically nonepistemic react in that way, too? ... From the realist's point of view, Putnam's argument has nothing to do with the nature of truth. It has nothing to do with the nature of the relation that must hold between a proposition and the world in order for the latter to confer truth on the former. It has, rather, to do with the nature of the relation that must hold between a proposition and a subject in order for the latter to accept the former. The realist who accepts Putnam's argument will describe its import like this: Putnam has shown that certain of the necessary conditions for accepting the proposition that everyone is a brain in a vat--or for accepting any proposition about brains and vats---could not be satisfied by anyone if everyone were a brain in a vat. Of course (the realist might add), if everyone were a brain in a vat, this proposition would be true; it is just that no one would be able to accept it or grasp it. 28

Notice that we have now come full circle. The last sentence quoted above reflects Brueckner's reconstruction of the argument. Van Inwagen is well aware that this is not how Putnam wants the argument to go, but he sees no

65


way that Putnam could prevent the realist from taking this way out. After all, there is no strict logical relation between correspondence truth and any epistemological position. One can hold on to correspondence truth come what may. It is only necessary that the realist admit that a BIV could not think that it was a BIV and that our beliefs about the external world could not be glolr ally false in the way of Cartesian skepticism. No problem. What has any of this to do with realism? As a matter of fact, I think that Cartesian skepticism has a great deal to do with traditional realism. In the remainder of this paper I hope to show that the realist pays a much higher price for these concessions than van Inwagen might lead us to believe.

IV. THE REALISTIC SPIRIT Putnam insists that he has abandoned metaphysical realism for reasons that are genuinely "realistic" in character. He says: Recognizing such facts as these is part of what might be called "rejecting 'realism' in the name of the realistic spirit." It is my view that reviving and revitalizing the realistic spirit is the important task for a philosopher at this time. 29

Most realists are not convinced by this talk. They tend to think that there is nothing very "realistic" about Putnam's position and that his arguments will be compelling only to those who already have antirealist leanings. Michael Devitt holds this view. He sees nothing "realistic" about internal realism, which he considers to be a version of "radical idealism" and which he assumes will have appeal only to those with "anti-realist" intuitions. In a vivid and comprehensive way Putnam has captured most of the intuitions that motivate anti-realism. These intuitions have always had some appeal in philosophy and have recently become popular. 30

On what basis, then, does Putnam claim that his view is motivated by the realistic spirit? It hardly need be said that if the "realistic spirit" is nothing more nor less than a disposition to embrace the (three) metaphysical tenets of realism, then there is nothing realistic about Putnam's position. But is realism, traditionally understood, nothing more nor less than the conjunction of these three doctrines? Admittedly, there is some justification for identifying "realism" with the three metaphysical tenets of realism in recent literature because of the semantic focus of the realism-antirealism dispute on the contemporary scene. Nonetheless, realism has a long and rich history in the modem period, and fixation on the metaphysical tenets is a relatively

66


recent phenomenon. I shall argue, further, that even today most selfproclaimed realists are committed to realism more broadly construed. It is not uncommon for "realism" to be considered as much a philosophical attitude, or even a temperament, as it is a doctrine. The "realistic spirit" reflects a comprehensive intellectual perspective which is often typified by a commitment to certain familiar doctrines, but is not necessarily exhausted by those doctrines and could, conceivably, require a rejection of some of them. Crispin Wright has captured at least part of the perspective I have in mind in the following passage: The realist in us \yants to hold to a certain sort of very general view about our place in the world, a view that ... mixes modesty with presumption. On the one hand, it is supposed, modestly, that how matters stand in the world, what opinions about it are true, is settled independently of whatever germane beliefs are held by actual people. On the other, we presume to think that we are capable of arriving at the right concepts with which to capture at least a substantial part of the truth, and that our cognitive capacities can and do very often put us in position to know the truth, or at least to believe it with ample justification. The unique attraction of realism is the nice balance of feasibility and dignity that it offers to our quest for knowledge .... We want the mountain to be climbable, but we also want it to be a real mountain, not some sort of reification of aspects of ourselves. 3 )

This broad "realistic" perspective has found expression in a commitment to a variety of distinct philosophical doctrines reflecting the particular interests of different historical periods. Today, realism is assumed to be a theory about ontology and/or semantics and truth. Not very many years ago, however, the most common use of the term "realism" was to refer to a doctrine (or group of doctrines) that reflects the epistemological orientation of the spirit of realism. While there are any number of beliefs that "epistemological realists" have typically held, I will focus attention on what I shall call the four epistemological tenets of realism, beginning with the core doctrine: (E 1) Epistemological Realism. Belief in the mind-independent existence of all (or most) of the objects of common sense and of science is epistemically justified.

A wealth of different realistic epistemological theories have been advanced over the years, each offering a separate account of how it is that our beliefs about the external world are justified.32 As a result of its commitment to mind-independent objects, therefore, epistemological realism takes skepticism quite seriously. It is logically possible that the ultimate nature of reality is, in principle, inaccessible to the cognitive and sensible faculties of human beings and thus it is possible, as Putnam says, that

67


we might be 'brains in a vat' and so the theory that is 'ideal' from the point of view of operational utility, inner beauty and elegance, 'plausibility', simplicity, 'conservatism', etc., might be false. 33

To take skepticism seriously, in this sense, is to hold that skepticism is coherent. Thus, realists of this stripe have been committed to (E2) The Coherence of Skepticism. The skeptical conditional (S): If we are all brains in a vat, then this is not a hand. is both intelligible and true.

This doctrine reflects the modesty of which Wright speaks. The mountain to be climbed is no mole hill and success is no mere consequence of having justified beliefs. This is an expression of an attitude central to the realist spirit, viz., that the human pursuit of knowledge is a risky affair and that the strides that have been made in our knowledge of the world represent genuine victories won in the face of very real obstacles. Thus, skepticism is intelligible, but false; skepticism is taken seriously, but not too seriously. Not only is the mountain real but we really are climbing it. According to the realist, then, (S) is true, but both the antecedent and the consequent are false. (EI) and (E2), then, lead naturally to: (E3) Anti-Skepticism. Most of our beliefs about common sense and scientific objects are known to be true and the brains in a vat hypothesis is known to be false.

Finally, since epistemological realism presupposes that the beliefs which result from a careful exercise of our faculties, both cognitive and sensible, possess (to a greater or lesser degree) significant epistemic warrant, this is a perspective that has historically resisted any philosophically motivated revision of the deliverances of common sense or of scientific inquiry. Phenomenalism, for example, was immediately resisted by epistemological realists-even before the internal difficulties of the view were made manifest-simply because it constitutes too radical a revision of what is assumed to be the common sense (pre-philosophical) interpretation of our discourse. Thus, the fourth epistemological tenet: (E4) Anti-revisionism. Since we are justified in believing that our cognitive and sensible faculties give us reliable access to reality, there is prima facie justification to resist any radical, philosophically motivated revision of the deliverences of common sense and of scientific inquiry.

While anti-revisionism is rarely picked out as a distinct doctrine, it is a regulative principle that has shaped the epistemological attitude of a great many realists through the years. 34 Recapping the taxonomy of traditional realism, thus far presented, we have

68


SEVEN TENETS OF TRADITIONAL REALISM Three Metaphysical Tenets of Realism (M 1) Correspondence Truth (M2) Semantic Realism (M3) Ontological Realism Four Epistemological Tenets of Realism (E 1) Epistemological Realism (E2) The Coherence of Skepticism (E3) Anti-skepticism (E4) Anti-revisionism

Today, it seems that realism has been largely disassociated from epistemology, and many philosophers are untiring in their insistence that realism is not an epistemological doctrine. 35 The four epistemological tenets of realism, though still believed by most realist-minded philosophers, are no longer considered to be necessary for being a "realist." The metaphysical tenetsespecially (MI) and (M2)-are considered to be the ineliminable core of "realism," both necessary and sufficient for being a realist. Perhaps a reasonable case can be made for making (Ml) and (M2) the heart of "realism." Note, first, that one can be committed to (Ml) and (M2) without being either an ontological realist or an epistemological realist, but one cannot be either of the latter two without being committed to (M 1) and (M2).36 On this account, then, a skeptic, even a radical skeptic, can be a realist. And that is precisely as it should be. The skeptic and the Cartesian both share a common vision of the philosophical landscape and thus of the epistemological enterprise. They share a common understanding of our cognitive aspirations, of the meaning of our discourse, and of the goal of epistemology. They agree that our linguistic capacities allow us to make claims about mind-independent reality and that our claims to knowledge are (for the most part) claims about the intrinsic nature of reality. They agree that we can engage in speculative metaphysics and thus consider the possibility that reality might be radically different from what it appears to be. Both are equally animated by the "realistic spirit" in this sense, with the former simply a bit more pessimistic about the extent to which our beliefs are justified and (in some cases) a bit more pessimistic about the truth of our present theories. Surely this difference-a disagreement about the truth of (E1)-is not what separates realists and antirealists. The skeptic who embraces (Ml) and (M2) but rejects (El) is misleadingly described as an "antirealist." If realism-antirealism is a dichotomy, there is certainly justification for placing the traditional skeptic on the realist side of the divide. All this is well and good. Let us assume that this picture of the

69


philosophical enterprise does in fact capture something of the "spirit of realism." And, on this picture, the skeptic and the traditional realist do share a commitment to (Ml) and (M2). This, however, should give little comfort to the realist who insists that Putnam's arguments have nothing to do with realism. It is true that the realist and the skeptic hold (Ml) and (M2) in common but that is hardly the only thing they share, nor is it the most important. The conflict between epistemological realism and skepticism in the modem period does not begin with an agreement about theories of truth and meaning. Rather, it begins with a consensus about the coherence of skepticism, (E2), with presumptions about semantics and truth being largely implicit and little developed. In fact, we typically attribute a commitment to (Ml) and (M2) to a philosopher on the basis of her acceptance of (E2) because we assume that (E2) entails (Ml) and (M2). The philosophical vision which the ontological realist and the skeptic supposedly share is only coherent if Cartesian skepticism itself is coherent. But this is precisely why Putnam's arguments have force against the classic realist picture. Putnam argues that correspondence truth and semantic realism when constrained by a plausible view of language (Le., the causal requirement for reference) not only cannot guarantee the coherence of Cartesian skepticism but also are in fact incompatible with it. It may be true that radical skepticism presupposes correspondence truth and semantic realism; but if Putnam is correct, the reverse is not true. Correspondence truth and semantic realism, when joined with the causal requirement, are incompatible with radical skepticism. What does all of this mean? As we shall see in the following section, Putnam's brains in a vat argument reveals that introducing the causal requirement for reference into the traditional realist picture produces troubling tensions and anomalies which cannot be casually dismissed. Putnam challenges the assumption that a metaphysical realism which accepts the causal requirement (let's call it causal realism) preserves most of the things that realists typically care about.

V. WHAT IS "REALISTIC" ABOUT METAPHYSICAL REALISM? The brains in a vat (BIV) argument is intended to manifest the counterintuitive consequences of causal realism (i.e., metaphysical realism plus the causal requirement). While the BIV scenario presupposes metaphysical realism, the rather surprising implications for both semantics and epistemology ultimately give the argument the force of a reductio. This is not to suggest that metaphysical realism leads to logical contradiction. As we have seen, the realist can certainly avoid that. However, if Putnam is correct, merely avoid70


ing contradiction will not necessarily provide much comfort to the realist. Recall that there was a time when a person could continue to believe in the existence of the aether even in the face of the negative results of the Michelson-Morley experiment. The only price one had to pay was to admit that bodies "shrink" in size as their velocity with respect to the aether increases. While this concession saved the aether theory from outright contradiction, it seemed an intolerable consequence for most observers and today seems an effective reductio of the theory. Putnam's reductio of metaphysical realism is analogous. To be successful he need not show that causal realism is logically inconsistent, he need only show that it requires unanticipated and ultimately unacceptable revisions somewhere within the realist's philosophical perspective. It may be that the causal realist can muster the courage to hold on to correspondence truth. What Putnam will not concede is that this is a reasonable thing to do nor that the result is reflective of the realistic spirit. Most philosophers who assess Putnam's BIV argument begin by granting him the causal requirement. This is not necessarily because they believe that he is right that there is such a requirement, but because they believe that realism will not be threatened even ifhe is right. Having granted the causal requirement, they also grant that a brain in a vat who says, "That rhododendron is six feet high," is not making a false assertion about a physical, mindindependent shrubbery, but is instead making a true assertion about electrical impulses, a computer program, or "rhododendron" images. It follows, then, that a BIV who asserts (S 1) If we are all brains in a vat, then there exists no six-foothigh rhododendron

is not expressing the same kind of skeptical worry that Descartes expressed. A necessary condition for a skeptical conditional to be of a Cartesian sort is that the antecedent express some non-standard theory about ultimate reality. The antecedent of a BIV' s utterance of (S 1) expresses no such thing. At best it asserts a (philosophically uninteresting) falsehood about electrical impulses, a computer program, or "vat" experiences; at worst, it is incoherent. Since BIV s have no cognitive access to genuine Cartesian conditionals, they lack the capacity to do philosophy in anything like the way that we take ourselves to be doing it. It is most telling that realists who assess Putnam's BIV argument rarely voice any dissatisfaction with this revisionist interpretation of the BIV's discourse. We might have initially thought that a BIV would be capable of doing philosophy, just as we do it. Once it is conceded, however, that the cognitive range of the BIV's language is seriously limited because of the limits of its causal reach, it must be admitted that BIVs will be serious underachievers as philosophers. This is not to say, of course, that a BIV could not utter the following words:

71


I believe that we are brains in a vat. If so, then in normal discourse our term 'chair' may refer to electrical impulses or to "chair"-images. However, when I say "We are brains in a vat" I do not mean to say that electrical impulses are causing us to have experiences as of being brains in a vat, that, of course, is clearly false. Rather I intend my words to reach beyond the immediate causal web within which we are trapped and to make a claim about a real computer-independent vat. There is no reason to doubt that a BIV, having experiences phenomenologically indistinguishable from our own and having dispositions similar to our own, might make such a speech. If Putnam is correct, however, this speech is plagued by incoherency. The BIV simply cannot be saying what we would take ourselves to be saying if these were our words. Note the concession that the realist makes in denying that the BIV has the capacity to entertain a Cartesian conditional. On the classic (Cartesian) picture, the BN's utterances are interpreted quite differently. A BIV's everyday "rhododendron" statements would come out false, but the BIV's philosophical speech would come out both coherent and true. Part of the perspective shared by the traditional realist and the skeptic is that philosophical discourse of this kind not only makes sense but also reflects an amazing and ennobling human capacity. It is assumed that human beings would possess this capacity even if they were brains in a vat. It is this latter assumption, of course, which the realist must abandon if Putnam is correct. BIV s do not possess the capacity to entertain radically skeptical hypotheses. BIV s are not capable of doing philosophy in this way. The foregoing philosophical speech is not simply misguided, it is nonsense. We must conclude, therefore, that if we were BIVs we would lack these capacities as well. If realists find this outcome at all distressing, relatively few of them express that distress in print. There are several reasons, I think, why realists rarely consider it a major concession to grant that a BIV's cognitive repertoire may be seriously limited. In the first place, it is easy for us to shrug our shoulders and say: "Well, it is no wonder that there are odd consequences given the bizarre nature of the BIV scenario." But there is a second reason that is far more important. There is a strong temptation to assume that it is only brains in vats whose cognitive powers are seriously constrained by the causal requirement. It is frequently assumed that none of the foregoing argument gives us-those of us who are not BIVs-any reason to worry. We can certainly do philosophy just the way we suppose ourselves to do it because our epistemic situation is crucially different from the BIVs. BIVs cannot speculate about wildly different ultimate ontologies; their language goes awry when they contemplate statements like (S1). Our language, however, the language ofnon-BIVs living in a world of mind-independent objects, will be plagued by no such

72


difficulties. We retain the resources necessary to do philosophy, to speculate about radically non-standard theories about reality, and to worry (unnecessarily) that some such theory might be true. The difference, we are told, is that speakers in the actual world bear the proper causal relation to the kind of objects that make speculation about the ultimate nature of reality possible. Thus, for example, while the BIV cannot think that it is a brain in a vat, we can think that we are because our term 'vat' refers to the very kind of physical vat that would house our brains were we brains in a vat. Thus, causal realism limits the cognitive range of the BIV but not of us. The only concession we need make is that if we were brains in a vat, then we could not think that we were. Given the fact that we aren't BIVs, this is a concession of no consequence. But this is not a proper assessment of the situation. Causal realism of the kind under consideration is far more revisionist than this picture indicates. It does not leave speakers in the actual world with sufficient resources for philosophical reflection of the Cartesian type. Consider the skeptical conditional: (S 1) If we are all brains in a vat, then there exists no six-foothigh rhododendron.

There is a substantial constraint on the kind of reflection that is possible on the causal realist picture, even in the actual world. While I can reflect upon a counterfactual possible world in which I am a brain in a vat and can speculate about the truth value in that world of 'The rhododendron is six feet high,' there is one thing that I cannot do. I cannot (consistently) entertain the thesis that the actual world is a world in which I am a BIV. I cannot consistently think that my present reflection on the skeptical conditional is the reflection of a brain in a vat. The limitation arises because the two cases require different interpretations of the antecedent of the conditional. In the case where I assume I am in the actual world imagining how things are in some counterfactual possible world, (SI) will be in English, and the antecedent will assert that my physical brain is in a physical vat. So far so good. If, on the other hand, I am asked to consider the genuinely skeptical hypothesis that the actual world is a world in which I am a BIV and thus that my present reflection of (S I) is the reflection of a BIV--that I cannot do. To accomplish that I must take (SI) first to be a sentence in vat-English, and second to be the occasion of my entertaining the possibility that I am a brain in a vat. I cannot do both. If I do the first, then (S I) is a sentence that a BIV could possibly entertain, but it does not express the hypothesis that I am a brain in a vat-and, thus, I fail to do the second. If I do the second, I am presupposing that I am not a brain in a vat and thus I fail to do the first. To entertain the possibility that the actual world is a BIV world I must be able coherently to believe both that some sentence I entertain expresses the claim

73


that I am a BIV and that no sentence I could possibly entertain expresses the claim that I am a BIV. Attempts to entertain the possibility that the actual world is a BIV world will get no further than do attempts to entertain the possibility that my present cognitive state is the state of someone who is unconscious (and cognitively inactive). As a consequence of this, the causal realist must compromise two epistemological tenets of realism. Skepticism, at least of the traditional Cartesian sort, is not coherent, and thus (E2), "The Coherence of Skepticism," must be rejected. Further, because human beings are disposed to believe that skepticism is coherent, at least some of what they say will ultimately be incoherent. Causal realism must be revisionist, then, just as verificationism is revisionist--determining that many utterances of otherwise competent speakers have (unbeknownst to them) some fatal semantic defect. A causal realist must, therefore, reject (E4), "Anti-revisionism." One of Putnam's most profound insights is his recognition that causal realism shares many of the apparent weaknesses of verificationism while lacking many of its strengths. For many years Putnam resisted the temptations of verificationism because it was revisionist: it told people that they couldn't say many things that, as otherwise competent language-speakers, they were prone to say. Realism, it seemed, put no such constraints on our linguistic capacities, providing semantic resources that verificationism lacked. As Putnam explored the implications of causal realism, however, he made a startling discovery: causal realism does not provide the resources that it promises. Cartesian skepticism is no more coherent on the causal realist picture than it is with verificationism. While verificationism limits our cognitive access to that which is (at least in principle) verifiable, causal realism limits our cognitive access to that which is within our causal sphere. The limitations of causal realism, however, are far more insidious. Realism was originally embraced because it promised to give us the semantic resources to express all those thoughts that we recognized (from our own epistemic perspective) as being cognitively meaningful. Causal realism is not merely revisionist, it does not merely legislate that certain types of discourse are incoherent, but it makes us strangers to our own thoughts and utterances. The privileged access that we always assumed that we had to the content of our own thoughts is now a myth. John Heil, one of the few commentators on the BIV argument who addresses this issue, expresses the situation well: Let us suppose for the moment, however, that Putnam is right: meanings are, if they are anywhere at all, someplace other than in the head; content is externally determined. A startling corollary of such a view seems to be that agents have no special access to the content of their own thoughts. I am in no privileged position to appreciate what my thoughts are about simply in virtue of their being mine. 37

74


The meaning of our utterances, on the causal realist picture, is an external matter. The cognitive content of my thoughts and utterances is, in a very real way, not only beyond my control but also beyond my epistemic access. The ties between the epistemic perspective of the speaker and the meaning of his utterances is severed. The significance of my own cognitive activity is removed altogether from the realm of the epistemic. Putnam rejects this view of language and insists that it is not consistent with the spirit of realism because it alienates speakers from their own language. In the earlier book [Reason, Truth and History] I described current views of truth as 'alienated' views, views which cause one to lose one or. another part of one's self and the world; in these lectures I have tried to elaborate on this remark, and on the connection between a non-alienated view of truth and a nonalienated view of human flourishing. 38

It may reasonably be argued that Putnam himself is responsible for this view of language. He is the one who helped create this revolution in the philosophy of language by arguing that "meanings just ain't in the head," and by insisting upon the causal requirement. Putnam, of course, stands by these claims. The extension ofthe term 'gold' is fixed by (possibly unknown) facts about objects to which I bear a certain causal relation. Putnam's whole point, however, is that this very reasonable theory of reference, when joined to a verificationist semantics, has none of the consequences just described. It is only when the causal requirement is joined with metaphysical realism that the speaker is alienated from her own thoughts. Internal realism is, for Putnam, a more "realistic" philosophical perspective than is causal realism. On a similar note, Crispin Wright spoke of the "dignity that [realism] offers to our quest for knowledge." How does causal realism fare on this measure? The familiar rap against verificationism is that knowledge is too easily achieved. Since there is no room for radical skepticism, there is no risk of radical failure. The epistemological enterprise is thought, then, to be diminished in value, to lack the nobility that comes from facing a challenge where one risks complete failure. On the causal realist picture, however, the risk of radical failure seems to be missing as well. The reason that the brains in a vat are not engaged in the traditional epistemological enterprise which realists have found so ennobling is that the brains are caught in a causal web. Their words cannot escape the limits of that web and thus they cannot even think the kind of thoughts necessary for philosophical speculation of the traditional realist sort. They also cannotfail to be right in their beliefs about the world, assuming that those beliefs meet sufficient theoretical and operational constraints. Regardless of what goes on in their heads, their beliefs about "tables" and "chairs" just will be true beliefs about whatever reality is ultimately causing them to have table- and chair-experiences-whether that reality is a computer program or an evil demon.

75


We of course assume that our power to philosophize is. not threatened in this way. After all, we bear the proper causal relation to real tables and real brains so that we are capable of the ennobling philosophical thoughts of a traditional realist sort. But in what relevant way does our situation differ from that ofthe BIVs? We are also caught in a causal web. We cannot speak of or otherwise think about anything that lies outside that causal web. We may think that our epistemological search is genuinely threatened with the possibility of a radical failure. We may think that our search for know ledge has a kind of nobility because we are willing to acknowledge the limitations of our cognitive capacities and to consider the implications of the fact that justified beliefs, even in the long run, may not necessarily be true. But all of this is unjustified sentimentality. There is a very real sense in which we cannot/ail to hold mainly true beliefs in the way just described, any more than the BIV can. The meaning of our words will be causally tied to whatever reality is ultimately causing us to have the experiences that we do, and so we just will end up holding true beliefs about that reality. Thus, brains in vats will hold true beliefs about electrical impulses. And we, too, are likely to hold true beliefs. Certainly something of the "dignity" of the quest for knowledge is lost inasmuch as the wondrous powers of "causal relations of the appropriate type" will literally carry us up the mountain, regardless of how bad our epistemic situation might be. If there is more room for the "dignity" of a "real climb up the mountain" with causal realism than with verificationism, it is difficult to see how.

VI. WHAT IS "REALISTIC" ABOUT PUTNAM'S INTERNAL REALISM? The primary objectives of this paper have now been met. I have argued that Putnam's arguments do put pressure on any traditional realist who is committed to the causal requirement. Metaphysical realism can obviously be maintained come what may. Nonetheless, intuitions that many have considered to be all but constitutive of the "realistic spirit" must be forsaken. If this is accurate, then metaphysical realism together with a plausible theory of language results in a philosophical picture that is not fully "realistic" in the traditional sense. Admittedly, Putnam's internal realism is not fully "realistic" in the traditional sense, either. The central question then is this: All things considered, is causal realism or internal realism more in keeping with the realistic spirit? I do not presume to answer this question. Certainly any answer would be controversial. In the brief discussion to follow, I shall do no more than to argue that Putnam's internal realism keeps much of the spirit of each of the seven tenets of traditional realism. This will

76


not establish that internal realism is more "realistic" than causal realism; it will show, however, that internal realism has a reasonable claim to the attribution, "realistic." (MI) Correspondence Truth. Putnam has no problems with correspondence truth on a certain, commonsensical reading: 'The tree is in the quad' will come out true only if the tree really is in the quad; our statements will not be true unless the world is the way we say it is. His objection to correspondence truth traditionally conceived is that it naively assumes that it makes sense to talk about our words being firmly fixed to some uninterpreted, thing-in-itself reality. While Putnam does dismiss this account of truth, he does not dismiss our pre-philosophical convictions about the meaning of the word 'true.' Instead he argues that the actual role that the word 'true' plays in our discourse can be accounted for without appeal to the traditional realist picture. Let's return to Wright's metaphor of the mountain. If the realistic spirit requires that there be a mountain that is a genuine obstacle to be overcome, Putnam retains much of that. It is a real mountain! It is not merely a reification of our faculties because our faculties may, in the end, be incapable of carrying us all the way up the mountain. Achieving cultural consensus or consensus among any finite group of human beings is not sufficient for truth. Truth is a regulative ideal for Putnam. More than that, it is an ideal that is not necessarily achievable. The very real mountain that we are attempting to climb may be insurmountable, given the laws of physics and the limitations of human beings' sensible and cognitive faculties. Of course, a great deal of what we believe is true on Putnam's account. Our beliefs about tables and chairs, for example, are paradigmatic examples of beliefs that do meet our highest epistemic standards. However, while we have made considerable progress up the mountain, there is no guarantee that we will reach the summit. In the area of particle physics, for example, it is at least logically possible that at some deep level of investigation, there may simply be no theory which we are capable of grasping which will meet all the theoretical and operational constraints that we ourselves demand that a successful theory meet. To say that an epistemically ideal theory could not be false (which Putnam does say) is not to say that in every conceivable area of inquiry it is possible for human beings to achieve an ideal theory. Not only is 'truth' not a name for "the best that a culture can accomplish," it is not even a name for "the best that human beings can accomplish." Putnam's view of truth, while not the doctrine that traditional realists have typically embraced, is nonetheless a view that makes sense of a good part of our discourse about truth, leaving the spirit of much of that discourse essentially intact. 39 (M2) Semantic Realism. It is often assumed that there can be nothing "realistic" about verificationist semantics because the heart of semantic realism is its countenance of "mind-independent objects"-and that is precisely what is left out of the verificationist account. One of the primary reasons that

77


we are committed to "mind-independent" objects is that we believe that physical objects are not causally dependent upon any mind, and thus that the following is true: Dinosaurs existed before human beings did and would have existed even if human beings had never existed.

If it follows from verificationism that (a) 'Dinosaurs existed 50 million years ago' is nothing more nor less than a claim about the existence of certain bones to be found in museums, or that (b) it is a contradiction to say that dinosaurs would have existed even if humans had not (since there would be no humans around to do any verifying), then verificationism is not very realistic. Putnam's verificationism does not have these consequences. His is an idealized verificationism where the truth of a statement will be determined by its verifiability under ideal epistemic conditions and in the long run. For Putnam the dinosaur statement is true because if a human being had been present 50 million years ago then the dinosaur would have been verifiable (i.e., if we had been there, we would have seen it). Dinosaurs are ontologically "independent" of human beings in this regard, and this is independent enough for much of what we want to say. (M3) Ontological Realism. It is widely recognized that Putnam's internal realism includes a commitment (at the empirical level) to the objects of both common sense and of science. Phenomenalism, for example, which is a monistic ontology (i.e., all objects are merely permanent possibilities of experience) cannot do justice to statements like "Micro-physical entities ultimately explain why tables and chairs possess the macro-properties that they do." Internal realism, however, allows for a commitment to electrons as well as tables and chairs, together with all of the levels of explanation that are presupposed by an "objective" (as opposed to a merely operational) interpretation of scientific theories. Putnam's empirical realism includes a full rich ontology, the envy of any metaphysical realist. (EO Epistemological Realism. Putnam's complex ontology (as previously described) 'allows him to hold an epistemological position that shares much in common with the traditional doctrine of epistemological realism. We believe in the existence of the objects of common sense and of science and those beliefs are justified. Assuming that these beliefs are true (and on Putnam's view the electron beliefs are certainly not guaranteed to be true) it is meaningful (and not trivial) to say that they are known to be true. Much of the epistemological orientation of traditional realism is preserved. (E2) The Coherence of Skepticism. As has been argued previously, Putnam, like the causal realist, must reject Cartesian-style skepticism. This hardly means, however, that there is no coherent form that skepticism can take. We can be skeptical about ever arriving at an epistemically ideal theory (see the discussion of (Ml) above). More than that, we can worry

78


about the coherent possibility that none of our present experiences of external objects are veridical-not in the transcendental way presupposed either by Cartesian skepticism or by Putnam's own BIV scenario, but in an empirical way. It is perfectly meaningful for Putnam to say: "I may have been kidnapped last night by evil scientists and plugged into a computer which is presently giving me experiences of trees that I am incapable of discriminating from 'real' -tree experiences." This is a hypothesis that is not only coherent on Putnam's view but which raises genuine epistemological worries. Given the intelligibility of this hypothesis, the epistemological enterprise is not an altogether trivial one. (E3) Anti-skepticism. We have already seen how internal realism is anti-skeptical. We are justified in believing in the objects of common sense and of science. It is logically possible that I have been captured by evil scientists, but I am warranted in believing that that hypothesis is false. (E4) Anti-revisionism. Putnam's strongest case against causal realism may well rest upon his claim that internal realism is less revisionist than causal realism. Of all the epistemological tenets of realism, Putnam has been most reluctant to give up "Anti-revisionism." He has been loath to advocate a philosophical revision of discourse that humans take to be not only meaningful but valuable. He believes that Rorty, Foucault, and Derrida are unacceptably revisionist, that they are gripped by the following idea: The failure of our philosophical "foundations" is a failure of the whole culture, and accepting that we were wrong in wanting or thinking we could have a foundation requires us to be philosophical revisionists. By this I mean that, for Rorty or Foucault or Derrida, the failure of foundationalism makes a difference to how we are allowed to talk in ordinary life-a difference as to whether and when we are allowed to use words like "know," "objective," "fact," and "reason." ... I am not, in that sense, a philosophical revisionist. 40

While Putnam's position does not trivialize terms like 'know' and 'objective' he does admit that his view has revisionist implications. Discussing his rejection of the realism-idealism dichotomy, he confesses that it was a difficult thing to do. Nonetheless, the attempt to save all of the old intuitions is hopeless. My rejection of these dichotomies will trouble many, and it should. Without the constraint of trying to 'save the appearances', philosophy becomes a game in which anyone can-and, as a rule does-say just about anything. Unless we take our intuitions seriously, we cannot do hard philosophy at all. So I respect philosophers who insist that the traditional dichotomies are deeply intuitive, and who 'need a lot of convincing' before they will give them up.

79


On the other hand, the whole thrust of Putnam's work over the last fifteen years is to convince us that the traditional realist paradigm is unstable. Revision is a last resort, but revision is unavoidable. The only question is: Which revision will you choose, and why? But if philosophy which simply scorns our intuitions is not worth the candle, philosophy which tries to preserve all of them becomes a vain attempt to have the past over again . . . . The task of the philosophers, as I see it, is to see which of our intuitions we can responsibly retain and which we must jettison in a period of enormous and unprecedented intellectual, as well as material, change. If! reject the dichotomies I depicted, it is not, then, because I fail to recognize their intuitive appeal, or because that intuitive appeal counts for nothing in my eyes. It is rather because these dichotomies have become distorting lenses which prevent us from seeing real phenomena-the phenomena I have been describing-in their full extent and significance.41

Putnam is correct. The "alienated" view of language with which causal realism is saddled violates the spirit of realism with extreme prejudice. While verificationism also puts certain constraints on our cognitive activities, they are constraints that we can live with. At least the meaning of our discourse remains intimately connected to the epistemic features which make our use of that language intelligible. Internal realism does not preserve every conceivable realistic intuition. It does, however, preserve a great deal of what realists have traditionally cared about. Internal realism may not be more realistic than causal realism, that is difficult to say. It certainly is, however, realistic in ways that causal realism is not.

NOTES 1. Research for this paper was supported in part by an ISU University Research Grant, for

which I am grateful. I would like to thank the following people for helpful conunents: Ann Baker, Larry BonJour, Pat Franken, Patrick Murphy, Hilary Putnam, Mark Siderits, Bob Steinman, Mark Tinunons, Charles Travis, and especially Thorn Carlson. 2. Here is one passage where Putnam suggests, explicitly, that he may be more realistic than some metaphysical realists: "That truth is a property ... is the one insight of 'realism' that we should not jettison. But Hartry Field shows signs of being inclined to jettison this insight, although he calls himself a 'metaphysical realist' and says that I am a 'nonrealist.' Could it be that I am more of a realist-though not a 'metaphysical' one-than Field, after all?" Realism with a Human Face, James Conant, ed. (Cambridge, Mass.: Harvard University Press, 1990),32. 3. The Many Faces of Realism (Lasalle, Ill.: Open Court, 1987), 1.

80


4. "Semantic Dualism," (submitted for publication). 5. For those tempted by Michael Devitt's claim that realism is exclusively an ontological doctrine not a theory about truth or meaning see note 36 below for an extended discussion of the matter. 6. In Reason, Truth and History (Cambridge: Cambridge University Press, 1981),49, Putnam suggests that the metaphysical realist also believes "There is exactly one true and complete description of 'the way the world is' ." I do not include this requirement because few philosophers think that it is a necessary condition for being a realist, and it plays no crucial role in the arguments to follow. 7. There is some justification for claiming that Putnam may not necessarily reject (M3), ontological realism. It all hinges on how the qualifier "independently of any mind" is cashed out. Since I interpret "independently of any mind" as invoking a realist semantics, Putnam is not an ontological realist. 8. Meaning and the Moral Sciences (Boston: Routledge & Kegan Paul, 1978), 129. 9. Philosophy of Logic (London: George Allen & Unwin Ltd, 1971),65-66. 10. I have benefited greatly from discussions with Mark Siderits about the model-theoretic argument. 11. "Realism and Reason" in Meaning and the Moral Sciences, 123-38, "Models and Reality" in Realism and Reason: Philosophical Papers, vol. 3 (Cambridge: Cambridge University Press, 1983), 1-25, and "A Problem about Reference," chapter 2 of Reason, Truth and History, 22-48. 12. Carsten Hansen, "Putnam's Indetenninacy Argument: The Skolemization of Absolutely Everything," Philosophical Studies 51 (1987): 89. 13. David Lewis, "Putnam's Paradox," Australasian Journal of Philosophy 62 (1984): 225 (my emphasis). 14. Realism and Reason, xii. 15. Reason, Truth and History, 3,51. 16. Lewis, op. cit., 226-29. 17. "How to be an Antirealist," Proceedings and Addresses of the American Philosophical Association (September 1982): 61. 18. Realism and Truth (Princeton: Princeton University Press, 1984), 190. 19. "Putnam's Model-Theoretic Argument Against Metaphysical Realism," Analysis 44 (1984): 137. 20. Lewis, op. cit., 231-32. 21. The second disjunct in this quote (Reason, Truth and History, 16-17) suggests that a direct causal connection is not a necessary condition for referring to concrete objects. This is borne out later when he says "The idea that causal connection is necessary is refuted by the fact that 'extraterrestrial' certainly refers to extraterrestrials whether we have ever causally interacted with any extraterrestrials or not!" (ibid., 52). 22. Among those who consider an interpretation of the BIV argument that is similar to Brueckner's are: John D. Collier, "Could I Conceive Being a Brain in a Vat?" Australasian Journal of Philosophy 68 (1990): 413-19; Paul Coppock, "Putnam's Transcendental Argument," Pacific Philosophical Quarterly 68 (1987): 14--28; J. Harrison, "Professor Putnam on Brains in Vats," Canadian Journal of Philosophy 17 (1987): 427-36; Gary Iseminger, "Putnam's Miraculous Argument," Analysis 48 (1988): 190-95; Michael Kinghan, "The External World Sceptic Escapes Again," Philosophia 16 (1986): 161-66; Jane McIntyre, "Putnam's Brains," Analysis 44 (1984): 59-61; and James Stephens and Lilly-Marlene Russow, "Brains in Vats and the Intemalist Perspective," Australasian Journal of Philosophy 63 (1985): 205- I 2. 23. Brueckner, Ibid., 167. 24. Brueckner actually does believe that the argument rests on a "trick." He says: "Given the presuppositions of our anti-skeptical argument, it is difficult to avoid an uneasy feeling that

81


25. 26.

27. 28. 29. 30. 31.

32.

33. 34.

35.

36.

37.

82

there is some trick involved in the reasoning of the last paragraph. To see that there is a trick ... note that ... " "Brains in a Vat," Journal of Philosophy 83 (1986): 164. Putnam, Reason, Truth and History, 22. Peter van Inwagen, "On Always Being Wrong," Midwest Studies in Philosophy, Volume XII: Realism and Antirealism, Peter A. French, Theodore E. Uehling, Jr., and Howard K. Wettstein, eds. (Minneapolis: University of Minnesota Press, 1988), 95-112. Reason, Truth and History, 22. Ibid., 104 and 106. Realism with a Human Face, 42. Review of Reason, Truth and History in The Philosophical Review 93 (1984): 277. "Realism, Antirealism,lrrealism, Quasi-Realism," Midwest Studies in Philosophy, Volume 12: Realism and Antirealism, 25. If one is interested in a taxonomy of epistemological realisms, Roger Cornman, in Perception, Common Sense & Science (New Haven: Yale University Press, 1975), serves up a cornucopia. Meaning and the Moral Sciences, 125. Realism has not traditionally opposed revision in our ontological commitments when motivated by scientific, rather than philosophical, considerations. The realistic spirit inspired a rejection of beliefs in animism, heavenly bodies, and countless other entities assumed as uncontroversial for hundreds of years. This is something of an overstatement. In the philosophy of perception, for example, the terms "representational realism" and "direct realism" continue to refer to the same epistemological doctrines that they always have. Nonetheless, outside of an explicitly epistemological context, to call a contemporary philosopher a "realist" is usually to attribute to her a commitment to one of the metaphysical doctrines discussed below. Michael Devitt rejects the view that (M)) and (M2) are essential to realism and has gained some support for his claim that realism is strictly an ontological doctrine, according to which one can be committed to (M3) without any commitment to either (Ml) or (M2). On this view, one can be an ontological realist without holding a substantive view of truth and without interpreting any statement realistically. He insists that those, like Quine, who eschew semantics altogether and thus deny that the realism-antirealism dispute is even intelligible can themselves be ontological realists. I find this baffling. On my view, a necessary condition for being an ontological realist about kittens is to interpret (at least some) kitten statements realistically and to believe that so interpreted they are (correspondence) true. If you don't believe that plants exist, then you can't be a botanist; if you don't believe in correspondence truth and you don't think that statements express realist truth-conditions, then you can't be a realist. Putnam is correct; metaphysical realism does presuppose a substantive view of truth and meaning. While it is not possible to justify this claim here, I am convinced that Devitt fails in his attempt to find anything in virtue of which Quine will qualify as an ontological realist. In his definition of "ontological realism," Devitt interprets the phrase "independently of the mental" in such a way that it is not necessarily an invocation of correspondence truth or a realist semantics. Thus, Quine can be committed to (M3) without being committed to (Ml) or (M2). The problem with this move is that if you do not give (M3) a strong, semantic reading then Putnam himself will be happy to embrace (M3). Putnam, after all, believes that dinosaurs existed long before human beings ever arrived on the scene-and would have existed even if human beings hadn't. The kind of empirical realism that is left after substantive theories of meaning and truth are abandoned is too pale and uncontroversial a doctrine to deserve the title "ontological realism." "The Epistemic Route to Anti-realism," Australasian Journal of Philosophy 66 (1988): 164. Heil offers his own solution to this problem, attempting to show that the causal realist can have introspective access to the content of his own thoughts. The content of this second-order thought, however, will also be determined externally: "Presumably, the


38. 39.

40. 41.

content of this second-order, introspective thought is itself determined externally and nonepistemically. Its being about my thought that p, requires that certain circumstances external to it obtain ... it evidently follows that I can, after all, have an introspective grip on the contents of my thought that p" (171). While this is a response available to the causal realist, it is difficult to appreciate why it is thought to dispel the worry. A second-order thought, the content of which is itself externally determined and thus non-epistemic, is just one more thought from which I am epistemically alienated. For every thought that Heil produces, the same objection can be raised. The Many Faces of Realism, 1. Assuming that this is an accurate account of Putnam's view of truth (and, admittedly, it may not be), one is tempted to say the following: If, in some area of inquiry, there is no ideal theory accessible to human faculties, then there will simply be no truth in that domain. This is, of course, quite different from saying that truth exists but is beyond our reach. I leave it to the reader to decide how "realistic" this result is. [I am grateful for discussions with Mark Timmons who helped me to see this point.] Realism with a Human Face, 20. The Many Faces of Realism, 28-30.

83

Miller, Richard W., Realism without Positivism , Philosophical Topics, 20:1 (1992:Spring) p.85


Realism without Positivism

Richard W. Miller Cornell University

The spirit of Hilary Putnam's work pervades so much of philosophy that one can detect it even in heated criticisms of his current claims. Of course, his searching and open-minded style permits diverse interpretations of that spirit. Still, any understanding of his implicit advice about how to do philosophy would include some version of this motto: do philosophy on the assumption that positivism is really dead. No one has done more to end the projects that dominated the philosophy of science in the first half of this century-for example, the search for a logic of induction and the effort to define theoretical meanings in observational terms. And no one has been less inclined to use the debris to help construct new bandwagons straining toward old goals of general regulation on a priori grounds. Bayesian canons of probability-change and the "semantic" view of theori~s are just two of the many such projects from which Putnam is conspicuously absent. "Positivism" itself has always had many meanings. For the purpose of extracting an underlying message of all of Putnam's work, a certain broadly reductionist tendency in positivism stands out. Philosophical problems were supposed to be solved through the discovery of topic-neutral rules, governing all fields at all times, which every possible rational inquirer would accept on sufficient reflection, rules basing all characterizations of a given, problematic type on other (relatively) unproblematic ascriptions. Putnam was

85


hardly the first to criticize the canons that were proposed, allegedly connecting hypothesis with confinning (or disconfinning) data, explanans with explanandum, theoretical tenn with cognitive meaning, and supervening (e.g., mental) phenomena with their supervened (e.g., neural) base. But he was one of the very few whose criticisms cut so deep that they cast doubt on the whole reductionist project, including probabilistic, model-theoretic, and computational revivals. For the purposes of this paper, I will assume that the whole project should, indeed, be abandoned. My intention is to explore the implications of this premise for the dispute over "metaphysical realism." What is the relation between Putnam's rejection of positivism and the controversies over realism that Putnam has been stirring up? The rejection of all a priori specifications of causality, objecthood, or meaning in tenns of an unproblematic basic vocabulary and general, effective rules of construction obviously supports Putnam's denial that there is One True Inventory of the World. But it is not so obvious that the rejection of positivism lends support to the claims of his which have provoked the most outrage, his assertions that there are intrinsic connections between human cognition and the nature of mind-independent reality, connections that cannot be factored out by systematically distinguishing the part of cognition that is internal to mental activity from the part that is not. Indeed, it might be thought (and I have heard it said) that the anti-reductionism of the earlier Putnam sets strict limits to the validity of the later attacks on metaphysical realism: perhaps by abandoning the old positivist baggage, together with commitment to a single true inventory of the world, one can evade all further attacks on metaphysical realism. I think that the anti-positivist message of the earlier writings gives much more support than this to the later attacks on metaphysical realism. I will base my appraisal on a sketch of accounts of empirical justification and of meaning that are in the spirit of Putnam's enduring rejection of positivism. This sketch will be screamingly programmatic-both fragmentary and dogmatic within the fragments. But it will, I hope, appeal to considerations that are important to most who think that positivism is well and truly dead. Then, I will extract some morals about realism from these postpositivist accounts of justification and content. Here are the main ones, in cryptic summary. 1) Truth is ideal justifiability-but in a way that gives no support to idealism, however transcendental, since the ideal includes a starkly objective requirement of epistemic luck. 2) The contribution of mental activity to true belief can never be isolated by analyzing beliefs into a narrowly mental component, independent of whatever would exist without the mental activity of the believer or those on whom she relies, and a causal grounding that is wholly independent of mental activity. As a further consequence, it is impossible to separate the conventional and the factual aspects of true belief.

86


I will also present some reasons to think that these morals to be drawn from the death of positivism are all that is essential to the rejection of metaphysical realism, as Putnam intends it. Far from setting limits to this rejection, the abandonment of positivism forces one to go all the way.

AFTER POSITIVISM Empirical Justification. Even if the positivist project was misguided, and general specifications in- unproblematic terms are not adequate means of applying problematic notions, there are good reasons for developing general statements about empirical justification, causation, content, and so forth. For example, an account of empirical justification can serve as a useful means of criticizing arguments, even if some of its principles are topicspecific or some residual circularity limits its power. To start with just this project: a wholly postpositivist account of empirical justification would begin with the fact that the empirical justification of a truth-claim consists of showing that it is entailed by a causal explanation of what is observed that is better than any relevant rivals. Of course, this beginning quickly leads back to the old canons if causal ascription is analyzed in some broadly Humean style, or ratings of rival ascriptions are determined by an a priori canon of topic-neutral principles. But there is a way out (short of abandoning the attempt to say something useful about empirical justification). In the rational pursuit of the best explanation, people apply and further develop a common repertoire of topic-specific truisms, prima facie principles identifying needs for explanation and means of filling those needs. The way out of positivism is to take this fact as fundamental, rather than basing such activity on a canon of deeper, topic-neutral principles, governing causal ascription of all kinds in all circumstances. Someone with the experiences that are normal for humans who is not at all prepared to respond to the relevant kinds of phenomena in the ways described by the topic-specific truisms either is not rational or is not engaged in causal explanation. For example, suppose that someone does not take sudden change of course in a nonliving thing to be even prima facie evidence that the thing has been interfered with by something else. Her responses are not guided by any such first principle, even when she has no definite reason not to rely on it. Then, that person is not engaged-in any case, not rationally-in the explanation of motion. Suppose that someone has no initial inclination to take what makes a finegrained appearance succeed a corresponding coarse-grained one from the same direction as revealing actual features of a surface in that direction in greater detail. Then, short of special experiences different from our norm, that person is not rationally engaged in explaining shape'S and textures in

87


visual appearances. Suppose that someone is utterly unsurprised by sudden changes in the apparent color of a surface and has no inclination to suppose that they are due to a physical transformation of that surface or a change in what affects how a physically unchanged surface looks. Apparent color change is as unsurprising to him, as un suggestive of a definite novel cause, as cruler change (grue' s becoming bleen at the Goodmanesque moment) is to us. Such a person is not rationally engaged in explaining apparent colors. For a rational explainer of such appearances, apparent color is, primajacie, evidence of color actually possessed, not evidence of cruler actually possessed. It is part of rational engagement in causal explanation to be moderately stubborn in resisting change in one's basic causal principles. Inability to find the keys that one distinctly remembers putting in a drawer is not sufficient grounds for abandoning the principle that material objects do not spontaneously disappear. Still, the familiar truisms combined with surprising evidence do sometimes create a need for explanation acute enough to justify a change in the framework for causal ascription. Usually, this is a change in the current repertoire of possible causes. Here, the rational investigator makes no greater change than is required to satisfy the explanatory need. In Maxwell's metaphor, she will adopt a minimally novel theory in order to maintain "that system of communications with an established base of operations, which is the only security for any permanent extension of science.") In Maxwell's own generation, the propagation of stress through a universal mass-less ether and action-at-a-distance with no intermediate change were rival departures from the base; arguably, each strayed no further than explanatory need required, so that both sustained rational and informed explanations. In the next generation, both departures became needlessly remote-partly because of new evidence, such as Michelson and Morley's, largely because of Einstein's demonstrations that certain equivalences were needed to maintain fundamental principles of symmetry and invariance. Of course, this sketch is in desperate need of improvement. And some parts of this enterprise-above all, cashing in the metaphor of minimal distance from the base-lead to much blurriness in all current versions. 2 Moreover, the sketch neither provides nor promises a set of topic-neutral noncircular principles that would determine when any given hypothesis is justified by any given body of data. Even apart from the absence of a general rule for measuring departure from the base, the topic-specific truisms in the base are enumerated, not derived from a deeper principle. One can say that these are the prima jacie principles to which someone with the normal course of human experience must, in the main, conform if she is rationally engaged in causal inquiry. But that would hardly be a means of basing assessments of rationality on applications of other, less problematic terms. Still, familiar considerations, attractive to many who resist Putnam's warnings about metaphysical realism, suggest that the general truth about

88


empirical justification will be some working-up of this sketch. No one would deny the approximate accuracy of the sketch as a mere description of our rational inquiries into truth. Yet efforts to show that the sketch fails to capture the essence of those inquiries soon meet the familiar obstacles to the old positivist programs. For example, efforts to derive the topic-specific truisms from topic-neutral principles of causality are vulnerable to the same basic problems as the covering-law model of explanation. The alternative claim that rational principles are any whose use is as-yet untroubled encounters the same problem of empirical anomaly as the old positivist canons: every significant system of explanatory principles conflicts with the data at some times while it is rationally employed; the question is whether the overall history of data gathering and theorizing is best explained as due to the truth of the underlying principles combined with the difficulty of applying them. But this judgment does not seem to be governed by a topic-neutral canon. In particular, catalogs of general explanatory virtues are not adequate to regulate such judgments of anomalies. One or another variety of simplicity always plays a crucial role in such strategies of assessment. The simplicity that is prescribed had better not be causal simplicity, the avoidance of alleged facts in need of explanation that receive none; for no topic-neutral theory of simplicity provides a remotely adequate criterion of explanatory need. For example, a topic-neutral criterion of explanatory need would have to show why the gaps and anomalies in Darwin's account of speciation were not lethal while the gaps and anomalies in the creative-intelligence theory were. However, when noncausal demands for simplicity are definite enough to regulate judgments of how lethal empirical anomaly is in light of empirical success elsewhere, they always tum out to exclude perfectly rational beliefs. Moreover, it seems wishful thinking to believe empirically troubled hypotheses just on grounds of mere formal elegance or instrumental convenience. Finally, one could deny that inference to the best causal explanation is fundamental to the empirical justification of truth-claims. But this makes it a mystery how humans can believe in the truth of statements about mindindependent facts. If these beliefs are claims to have exercised relevant detection capacities, then the causal-explanatory account comes into play: detection relies on effects of some process whose efficacy depends on the detected phenomenon. But if successful detection is not at issue, what conceivable correspondence is asserted in our truth-claims? One is driven to old phenomenalist answers, and their familiar problems. 3 Meaning. Even if this sketch of empirical justification is on the right track, it creates a need to discuss another large topic with which the positivists heroically grappled, the question of what determines the content of our beliefs. For one thing, the explanation of conduct in tenus of motivating beliefs and desires is part of the base of causal explanation; in the postpositivist account, this rational practice will have distinctive principles, which are

89


not mere derivatives of a deeper, topic-neutral analysis of causality. Also, if the prejudices embodied in the basic truisms are to be rescued from the charge of dogmatism, there must be some appropriate response to the question of what determines content. For the crucial verdicts are disjunctions having to do with content: an unmotivated departure from the norm of response, against a background of normal experience, is supposed to establish that an inquirer either is being irrational or is not applying the concept in questionfor example, is not ascribing a cause of motion, a shape, or a physical color. I will try to show how a general yet postpositivist account of content relies on ascriptions of minimally competent cognitive activity without analyzing away such ascriptions in favor of non normative or nonintentional idioms. As a first step toward general truths about content which might be useful, consider the content of the beliefs of someone who is fully rational and who can remember whatever she has noticed should it be relevant to her inquiries. Since she is rational and capable of surveying the relevant evidence that she has acquired, she will only believe a proposition if it can be justified in light of the evidence that she has encountered (through the process of justification depicted in the previous sketch). But, of course, she may not actually believe what she is in a position to believe. What she believes depends on conduct and practical rationality. As part of her complete rationality, our superperson is rational in what she does and in what she is prepared to do in this or that circumstanceincluding such mental doings as moving to further conclusions and questioning what was accepted before. So her beliefs must be in propositions that figure in fully adequate rationales for this (actual and conditional) conducti.e., in arguments establishing the rationality of what she does in light of her basic goals in life, which are assumed to be rational. After all, a belief that would not provide a reason for any choice by the believer, should her circumstances make the relevant topic of interest to her, would be no belief at all; and our supersubject's choices are fully rational. The total belief set of a fully rational person is what she could rationally believe in light of her experience (assuming it is accurately accessible) and what rationalizes her conduct as well. On the one hand, what plays no role in rationalizing conduct is not a belief. On the other hand, questions of what is justified in light of her evidence might distinguish between different total belief-sets, each of them adequately rationalizing. But further discrimination would appeal to nothing relevant in her psychology, and the ascription of content to her beliefs is a characterization of her psychology.4 The normative element in these statements about content, i.e., the constraint of rationality, has not been reduced to topic-neutral rules connecting inputs with outputs. Giving up on positivism, with its dream of a canon of topic-neutral rules for rational belief, means abandoning the dream of this reductive treatment of rationality. So this first step toward an account of con-

90


tent fits Putnam's recent criticisms of functionalism. s On the other hand, it might seem to transgress his earlier warnings about the causal aspect of content. How can rational responding and acting connect a person with appr~ priate contents if "meanings aren't in the head"?-the objection is misguided, because rationality, experience, and conduct do dictate whether a given proposition is an object of a fully rational person's beliefs. At the same time, the objection points to the need for a further statement about content, describing what makes propositions available as possible objects of belief. A small instance of meaning's not being in the head will shed light on both morals. Suppose that our supersubject has had an experience like mine, one summer in the early 1970s, and has been awakened every Sunday morning by the very loud playing of "Honky Tonk Woman" with live drum accompaniment. On each occasion, she has seen a person of a certain shape and size in an apartment across the courtyard playing drums, stopping at the end of the music, walking across the room to a record player, and initiating blessed silence. She soon starts to fonn beliefs that there is one and only one person who loudly plays a drum accompaniment to "Honky Tonk Woman" across the courtyard every Sunday morning and that, whoever that one is, he has other characteristics as well. For example, upon noticing appropriate goings on, she will fonn the belief that the one and only one, etc., is having a dinner party (where this belief is to be analyzed as a general claim, in the style of Russell's "On Denoting"). Presumably, she will also start to fonn other beliefs, about a particular person, which she might inscribe in her diary using the tenn, "the Pest". What she might enter as "The Pest is having a dinner party" has a different content from any of her beliefs that there is one and only one, etc .... After all, if she had had no exposure to a noisy neighbor or to the person who actually is her noisy neighbor, she could still entertain the thought that there is just one person who plays "Honky Tonk Woman" across the courtyard every Sunday morning and he is having a dinner party. But she could not entertain the thought that the Pest is having a dinner party. Someone might take this case to show that constraints of rationality do not connect a person with a given proposition. "For Suppose," the objection might begin, "that, unbeknownst to the subject, a variety of different people had keys to the apartment and each had put on a Sunday morning performance. Then, when she wrote, 'The Pest is having a dinner party' , either she would not manage to have a genuine belief, or, in any case, none that is not one of the general claims about there being one and only one person who ... Yet her rationality and memory could be perfectly intact. So the prop~ sitions which she believes need not be those which make her doings rati~ nal, among those which she would be rational to believe." The final inference is incorrect. If there are multiple ghastly Sunday perfonners, then there is no nongeneral proposition expressed by the entry, "The Pest is having a dinner party." But among the propositions that there

91


are, the ones that our supersubject believes are the justified (general) ones that rationalize her doings. On the other hand, if there is just one ghastly Sunday performer (etc.) and she has heard or seen him on most of the occasions in which she thinks she has, then there is a nongeneral proposition, and she believes it because it figures among the justified rationales. The real issue is what the existence of that nongeneral proposition consists in. The existence of some sole ghastly Sunday performer is not sufficient, even supposing the subject does believe the various general propositions, on the basis of the experiences that I described. For the nongeneral proposition is about a particular ghastly Sunday performer; the existence of a different performer would make for a proposition with another content. At this point, it is tempting to suppose that content has been created by some kind of causation which is generally reference-fixing, so that an individual is made the referent of a belief by being part of the fact giving rise to the belief in the reference-fixing way. But an individual could cause a general belief, which would be a general belief for all that. At this point, some might be tempted to posit items in a neural language that intrinsically have the status of proper names. But this is neuromythology and, in any case, replaces a mere problem with the mystery of how a neural item could have such an intrinsic status. The solution is to take seriously the dependence of content on success in specific cognitive activities. In Evans' and Grice's phrase, the crucial activity is, figuratively, the keeping of dossiers.6 For there to be the proposition that our supersubject actually expressed in her diary entry, she must, as it were, maintain a dossier in which she keeps track of the further properties of the one and only one person who has repeatedly played "Hooky Took Woman" on Sunday morning. By this I mean, very roughly, that a) when she reaches conclusions about features of the one and only one ghastly Sunday performer, she is inclined to compare them with relatively enduring features that she has previously ascribed to the one and only one performer, to conclude that she has probably made a mistake in ascription if she recalls a contrary feature, and, otherwise, to treat the new feature as a basis for future comparisons; and b) she would abandon this way of inquiring about the world, if she thought that the accumulated beliefs did not, in the main, have a single source; and c) her believing that one person was not the agent in most of the Sunday morning playings of "Honky Tonk Woman" would motivate such abandonment.

Dossier keeping can generate content in the following way. If a dossier does have a single source of the kind that is the dossier keeper's condition for nonabandonment, then when he accepts a characterization as part of the dossier,

92


he believes a nongeneral proposition about that particular source. If there is no such source, there is no nongeneral proposition for him to believe in the course of his dossier keeping. Causation plays an essential role in this view of content. But it is relevant because of specific intentional activities of the subject, which help to make that content possible. No general process of causal molding would be enough. Suppose, for example, that the subject formed all of the general beliefs in question, as a result of interaction with the sole Sunday performer. Suppose that she did not organize her inquiries by means of the dossier. The general beliefs just register. Then she would have no relevant nongeneral beliefs. 7 This little parable of content and objectivity is special in many ways. Above all, our supersubject was specially self-sufficient in this referring. In most referring, one defers, in part, to episternic operations of others with whom one cooperates in the process of learning and deciding. (As Putnam has famously noted, the capacity of most modem city-dwellers to refer to beeches rather than elms does not rest on their capacity to distinguish beeches from elms.) Still, a vague moral of the parable of the Pest applies to content in general. Whether one refers to an individual, a property, or a relation, reference to that item depends on the success of episternic operations in which someone attempts to keep track of something in the course of learning. (The crucial successes may concern projects of others to whom the subject defers in the course of learning-botanists, say, in Putnam's exampIe.) Meanings are not in the head because success is not guaranteed; but specific cognitive activities determine what is success. Correspondingly, once one faces the question of how propositions come to be available for belief, it helps to strengthen the initial idealization. Suppose the subject is not just completely rational, but, furthermore, completely successful in all of her episternic operations. In particular, she successfully detects all presences and absences that she seeks to identify. Then, the detection capacities that she is currently prepared to exercise determine the individuals, properties, and relations that are subject to her ascriptions, her denials of ascriptions, and her more compound believings. What constructions from these elements actually figure among her beliefs will depend, as before, on what propositions, constructed from such objects of detection, are justified parts of rationales. Of course, no one is actually this ideal. We all have some beliefs lacking adequate justification, and act in some ways that lack adequate justification (often relying on abuse of beliefs to rationalize what we do). Even more clearly, each of us rnisattributes and inaccurately discerns. Still, the ideal of perfect rationality and detection plays an essential role in the possession of content. Our actual beliefs are determined by the capacities through which we approximate to this ideal. A being has a belief that p just

93


in case the proposition that p can be constructed through the exercise of her capacities for detection and inference (such as they are and under her circumstances), and the proposition provides rationales for behavior of hers that is rational. Otherwise, the belief is one she cannot grasp or one that does not play the rationalizing role that would make it her actual belief as opposed to a belief that she could form if properly disposed. Putting aside the rare and extreme cases in which a believer fails to refer (for example, belief in caloric theory or the Ossian hoax), capacities determine the beliefs of imperfect believers in the following way. Someone attributes something, denies its presence or otherwise uses it propositionally only if he is exercising a capacity to detect it, or deferring to the exercise of this capacity in others; the capacity may be imperfect, so long as it is a capacity for detecting the item, nonetheless. Someone has a belief in a proposition, constructed from such objects of his capacity for detection, if and only if the proposition figures in available arguments from his experience and to his conduct, and he is sufficiently rational that the rationality of the arguments (which need not be completely valid) is part of the explanation of what he does. 8 In this approach to questions of content, one accepts phenomena that are obvious enough outside of philosophy: someone can be able to detect something even though she is not a perfect detector of it; someone can have a rationale based on justified propositions even though rationale or justification are not fully adequate. Of course, the line between mere detection and nondetection, mere rationality and nonrationality, will be drawn differently from case to case, depending on the topic and on relevant interests. A four-yearold can tell who is a mother and can have beliefs in which she attributes motherhood, even beliefs that she has formed independently. But some imperfections in her capacity to detect motherhood are massive and dependent on quite important biological ignorance. (A four-month-old lacks enough capacity to discriminate, and can have no such beliefs.) On the other hand, when terms are technical, incapacities no more drastic than the fouryear-old's can guarantee that detecting is not good enough for believing. Despite the usual philosopher of science's smattering, I can't discriminate well enough to have any non borrowed belief about muons. Apart from its accommodation of semantic common sense, this reliance on the good-enough makes it possible to admit the grain of truth in verificationism, while asserting its falsehood over all. No one can refer to something unless he or those to whose activities he defers can detect it. Indeed, putting deference to one side (as in the parable of the Pest), reference is an exercise of the capacity to detect the referent. But relevant processes of detecting can, nonetheless, be defective in ways of which the practitioners are ignorant; so a believer's method of verification need not yield the truth conditions for her belief. For the same reason, the appropriate requirement of detection permits false belief, even false belief concerning a particular

94


matter of fact. When detection capacities are exercised, their imperfections can be at work in the exercise at hand, so that someone exercising an ability to tell who is a mother (or who is the Pest) gets it wrong about Agnes' being Jane's mother (or the Pest's giving a dinner party). As with the sketch of empirical justification, few would question the approximate descriptive accuracy ofthis portrayal. The live issue is whether these general statements about content can be derived from a deeper account which does not list topic-specific principles, yet dispenses with talk of rationality, adequate competence, and cognitive activities such as detection in favor of less problematic terms. The abandonment of-positivism dictates abandoning pursuit of this deeper account, as well. A topic-neutral account of content which avoids the problematic normative terms requires the positivist canon to define the ideal of rationality. Thus, if the positivist canon of rules determining rational belief is unavailable, so, too, is the deeper account. Indeed, pursuit of this account is even more far-fetched, since it also requires a topic-neutral rule for permissible departure from the ideal. Finally, the unavailability of the positivist canon makes it impossible to analyze away talk of detection, propositional combination, and other aspects of cognitive activity. Rationality must be entailed by any description of such a process in the believer. If someone's rationality plays no role in a process it does not constitute his detecting or predicating, for example. But, if a general condition for rational cognitive activity is to determine the content of mental states, it must not presuppose that goals and beliefs have already been ascribed. Such a condition must describe transitions of rationality-making kinds among states each of which is goal and belief independent. If we had access to this description, we could use it as our topic-neutral means of distinguishing rational from irrational responses to data. It would be the positivists' dream canon. And we must have access, at least implicitly, to this description of rationalizing transitions, if it is implicit in ascriptions of content. For we are capable of ascribing content. So dispensing with all ascriptions of intentional activity in a general account of the nature of content would require commitment to the positivists' quest. Projects and Meanings. Despite the emptiness of certain reductive projects, no old-fashioned reductive prejudice is needed to deny that the normative account in the previous section is the whole story of content. For one thing, more has to be said about the justification of ascriptions of content. The account just given could only be applied to determine whether someone has a particular belief in light of a variety of background beliefs about the subject. The relevant conduct must be taken to be guided by basic rationality. There must be some further assumptions giving sufficient access to the experience on which the believer relies, the intentions or interests constraining the rationality of her choices, and the auxiliary beliefs determining the rational-

95


ity of her inferences. These assumptions must be partly independent of any general presumption of rationality, since one can, after all, acquire grounds for concluding that a being who seemed to be rational (or, rational on a given topic) has turned out to be basically nonrational (or, nonrational on that particular topic). Moreover, this content-relevant access to another's experience, interests, or background beliefs is not always provided by the great maxim that what sounds like the use of a language one knows probably is its use, as one would use it. After all, one was able to learn one's own language, and one can come to understand people whose language is initially strange or whose use of a familiar language differs in a way that affects content. In sum, we are justified in our ascriptions of content through reliance on a fund of principles which are not internal to the normative account of content, but, rather, are our means of applying it. The basis for justification here is ineliminably piecemeal and topicspecific. For example, warranted beliefs about what a subject notices must playa crucial role. Her beliefs need not remotely fit what is present in her environment if what is present is not at all salient to her. Yet our judgments of salience could hardly be justified without any reliance on the principle that shifting one's eyes in the direction of something is evidence of interest in it, as is reaching out and touching it. Beginning with the elementary access that these piecemeal principles afford, and deploying the general connections between content, rationality, and detection, one can often construct a rich account of someone' s beliefs-perhaps, in Grice's phrase, moving "from the banal to the bizarre."9 On the other hand, topic-neutral canons for interpreting belief-states either lead to absurdity or revert to the positivists' dream canon. A rule that someone is to be taken to ascribe a property just in case she is in a state that she would not be in if the property were absent is only true of infallible people; it leaves no room for false ascription. The rule that someone is ascribing a property if and only if her state would be causally dependent on the property, were she ideally situated and competent, is more sensible. But how can this standard of ideal causal molding be applied? Must one know that a distinctive kind of competence and luck-say, that involved in color ascription-provides the conditions in which the subject's current state would infallibly indicate the property? Then considerable background knowledge of the subject's intentions and noticings is already assumedknowledge of the sort that was to be derived on the basis of topic-neutral principles. Suppose, on the other hand, that one relies on a topic-neutral standard of competence and luck in the warrant for one's judgment that someone's belief-state would have to reflect the presence of a certain property if his competence and luck were ideal. Then, there would have to be a topicneutral, effectively applicable description of rational response to evidence, an essential aspect of this competence; so the positivists' topic-neutral canon could not be an illusion.

96


A general rule of linguistic charity has a similar fate if it is meant to do the work of a motley of piecemeal evidential principles. It leads to absurdity if it is treated as a demand to maximize truths. "Caloric" is interpreted as molecular kinetic energy, and so forth. Otherwise, the charitable assumption becomes the expectation that inference and deliberation conform to some topic-neutral canon for rational believing and choosing. Even if the piecemeal approach to semantic justification is accepted, there is at least one other nonreductive reason for regarding the discussions of content so far as radically incomplete. These discussions have often made the satisfaction of specific norms relevant to the semantic description of a subject. For example (to adapt a famous example of Putnam's), whether someone is talking or thinking about gold might depend on whether she is exercising a capacity to detect gold, i.e., whether she is engaged in a project with appropriate requirements for success, meeting them well enough. What makes a norm a standard governing someone's activity, so that the norm is relevant to the characterization of her beliefs? Probably there is no single unified, nontrivial answer. Nonetheless, there are quite nonreductive reasons for expecting broadly applicable accounts of what makes norms relevant, accounts that are noncircular enough to provide rationales for an indefinite variety of ascriptions of norms in light of specific circumstances. On the one hand, the fact that a norm is relevant to what someone is doing on an occasion generally depends on her life prior to this occasion. For what she does now could be a mistake in pursuit of the norms to which she seeks to conform, norms that must, then, be hers at present because of her life before now. On the other hand, at least where the ascription of content is concerned, the ways in which a norm is made relevant by facts in the historical background must involve a relatively few general principles, supplying reasons why a norm is relevant. For on the basis of extremely limited knowledge of the histories of others, we are wonderfully good at identifying the contents of their beliefs, and, hence, the terms on which they would be guilty of confusion, hasty inference, lapse of attention, misidentification, or some other failure. The best-worked terrain in the identification of content-relevant norms is the theory of communicative intentions, pioneered by Grice and enormously enriched by Lewis's account of conventions. At-least in the normal usage of "language", if someone is making an assertion that p in a language, then she intends that others believe that p as a result of their using the conventions of the language to discern her intention that they should so believe. But in virtue of what does she adhere (and expect her listeners to recognize that she adheres) to this set of conventional norms, establishing the proper connections between signs and beliefs? Typically, the norms are made relevant by the speaker's and intended audience's participation in common projects of advancing learning through cooperation. Generally and roughly, adherence to a convention is participation in a 97


practice, burdensome if not generally observed, which each participant in a common project can justify as the best means of advancing the project should others be willing to do the same. In a common project of advancing learning through cooperation, the fact that homophonic ally similar people can be expected spontaneously to associate certain beliefs with certain utterances is a vital resource. Because the evocations are spontaneous, there is no need constantly to interrupt the process of learning about the world to find new means of conveying one's relevant knowledge (in stipulations which must, in any case, rely on some given, spontaneous responses). So every participant has reason to rely on these given response tendencies, when they are part of common lore, using them to create beliefs on the basis of trust and expecting others to do the same. It will be a violation of a norm implicit in everyone's commitment to cooperate in learning if anyone departs from expectations which he is, rationally, counted on to use, without some appropriate signaling of the departure. Still, this common lore is only the raw material for linguistic norms binding the cooperators. For learning would rarely be advanced by a practice in which each seeks to conform to all associations of expressions with beliefs that have been routine up until the present moment, counting on all other homophonic participants in learning to do the same. For one thing, by not taking on the whole baggage of spontaneously evoked belief, one avoids pointless complexities, for example, constant qualifications to adjust to atypical cases (a six-footed chair) and needlessly elaborate tests for competence in communication (dismissing a potentially useful infonnant as incompetent in communicating about elephants if he does not share in the common lore that all elephants have African or Asian ancestors). Moreover, by relying only on a proper subset of the routine evocations, each of us maintains a basis for communication that is rationally open to surprising revisions of common lore. As time goes on or as the number of cooperators increases, someone who meets all conditions for trust that are readily applicable may have good reason to believe that the current consensus is wrong. If too much of the current consensus is taken to be conveyed by current means of utterance, he will be unable to contribute to the advance of learning by means of the old words, without clumsy hedges or confusing new coinages. In sum, the rational limit on the scope of a term is the one that narrows the scope in ways that make it an effective discriminator in the tasks of learning which dominate its use, without narrowing it any further. This is the norm to which someone using a tenn, participating in a project of homophonic cooperation, will seek to conform, expecting others to do the same. The norm will be relevant to her use of the tenn because of two kinds of phenomena, readily accessible in spite of our ignorance of most details of others' lives: features that every cooperator in the common project can be expected to associate with the tenn, and goals of the common project in

98


which the cooperators are engaged, goals in light of which cooperators can regard discriminations as useful or burdensome. For example, the communicative account of linguistic norms explains what it is in virtue of which users of "gold" are committed to the norm that emerges from Putnam's analysis of nontechnical natural-kind terms in "The Meaning of 'Meaning'." A plain natural-kind term such as "gold" will evoke certain features on the part of most competent language-users, for example, glittery yellow, heavy, and malleable. The reference of a plain natural-kind term should be controlled by these common associations, but not too directly. These terms are used in the project of making nature useful. If actual possession of the properties were required, the term would be an ineffective means of finding items that lack the commonly associated features but can be worked up so that they possess them. But the mere capacity to acquire those features is not of interest, either, if acquisition adds to nature rather than developing what is naturally there. Gold in low-grade ore is gold, but lead is not gold even though it can be gilded. The limit on content that advances our common projects must take into account both the reason for breadth and the reason for narrowness. Suppose that the samples we have encountered which display the commonly associated features typically have those features because of certain properties which are also found in other stuff. Because of these underlying properties, the other stuff would manifest those commonly associated features if we could put it in the same situation or arrangement. The grouping together of all such stuff is the classification that will best serve the purpose of making nature useful through communicating facts as to what is there. Of course, a specific, accurate description of the underlying properties provides ideal help. Yet someone ignorant ofthe best-established detailed description, or disbelieving it for good reasons, may also contribute to the common project. Best to have a norm that "gold" is taken to refer to whatever has the underlying nature (if any) producing glittery yellow and so forth among typical instances in the cooperators' environment. I 0 In this way, the Grice-Lewis apparatus can be used to connect contentdetermining norms with a background of interests and evocations within a homophonic community. In addition, the full scope of human cooperation in learning reaches beyond any single such community, potentially embracing all humankind. Our interest in nature commits us to learn from users of other languages, with their mostly different environments, and to maintain a historical record of findings by our ancestors, who, also, responded to a different environment from our own. These global, transhistorical interests ground the translations in which the communicative intentions of others are put in our own terms. In such farflung cooperative learning, it would be pointlessly complex if current "gold" users committed themselves to a distinctive terminology for

99


findings derived from each of those other communities, with a separate expression corresponding to each environment whose typical yellow, glittery, heavy, malleable stuff were of interest to the inhabitants. So long as those others, in their applications of a term, are concerned with what typically underlies their local samples of yellow, etc., stuff and the underlying structure is actually the same as what underlies typical samples in the current "gold"- using community, it advances learning to treat their uses of the term as having the same content as our uses of "gold". Moreover, the same general technique for using communications from others is the one those others use in dealing with alien communications about natural kinds and the one they intend others to employ in deriving information from their findings. After all, those others are also participants in the global, transhistorical activity of advancing learning through communication. So it is not merely convenient, but accurate as well to identify Archimedes' use of "khrusos" as referring to gold: reference, here, derives in the standard way from shared goals of communicators and audiences. I hope that this final sketch has indicated how an appropriate grounding of norms might occur in one vast territory, the assertion of beliefs in language. 11 I think that similar considerations establish the content of languageindependent thoughts, which are lines of communication with one's self in the future. This would help to support Putnam's view of thinking as ineliminably conventionaL Whether talking, writing, or merely thinking, one describes the world in a certain way in virtue of the fact that certain human needs are well served by exploiting given human acts and tendencies in the interest of coordination. If the practical rationale or the coordinative associations did not exist, one would not be engaged in that descriptive activity. This is clearest when description depends on a prior, authoritative stipulation. Describing lengths on the scale of the atom in Angstrom units requires participation in cooperative activities advanced by deference to given decisions of the relevant scientific congresses. But the interests need not be so social and the means for their pursuit need not be a stipulation. Describing surfaces by their color requires color-dependent interests in using nature together with a tendency for differently colored surfaces, under moderate illumination, to evoke different color sensations in a normal observer. Even when characteristics are not approximately definable, as colors are, in terms of human responses which they contingently evoke, their employment indispensably depends on human interests and the exploitation of contingent associations as means of coordination. Thus, our engagement in geometric description depends on our having an interest in moving things, for consumption, amusement, or other use, a variety of given tendencies to be bored by continuities, surprised by discontinuities, and an inclination to associate

100


strategies of manipulation with spatial images. Thinking about spatial characteristics relies on these tendencies, to coordinate past experience with future conduct in the interest of those locomotive goals. So it is on a continuum with the measurer's reliance on the Angstrom-unit convention. The strategic approach to content clarifies and vindicates the postpositivist perspectives that I described before. For example, this approach makes it clearer that the intimate connection between co-reference of natural-kind terms and actual identity of underlying causal structure depends on the interests governing specific cognitive activities, rather than reflecting some general causal recipe determining reference in each case. Above all, the strategic approach helps to save the postpositivist account of justification from verdicts of dogmatism. The prejudices embodied in the topic-specific truisms are supposed to be fundamental to rational inquiry because the failure to accept those starting points would constitute either nonengagement in rational inquiry or unconcern with the concepts governed by those prima facie truisms. Such a restriction could hardly be based on some mind-independent aspect of nature, since nature has no commitment to aid our inquiries. But it can be based on the strategic considerations that set limits to content. In each case, a more relaxed norm would be out of place because our projects of cooperation in learning would be retarded if someone's adherence to the norm were not a condition for taking his communications seriously, i.e., tentatively adding them to one's own fund of information. Learning is advanced if unmotivated departures from the norms connecting color-sensation with physical color, or heightened clarity and distinctness with veridicality, are not taken seriously as sources of information about color or shape. However, the usual rationales of convenience and open-mindedness make it undesirable to insist on much more robust conditions for taking communications seriously. So conceptual necessities that regulate rational inquiry into nature can be distinguished from mere entrenched convictions, without supposing that we have a priori access to mind-independent necessities. These various views of justification, content, and content-determining strategies have all been in the spirit of Putnam, at least in the following sense: once their approximate descriptive accuracy is acknowledged, Putnam's anti-positivist arguments make it impossible to dismiss them as partial, superficial reflections of deeper, general principles such as canons of scientific method, causal recipes for reference-fixing, and functionalist algorithms for mental functioning. I will assume that doing semantics and epistemology in the spirit of Putnam means developing these views, rather than dispensing with them. What are the further implications of this approach for the assessment of realism?

101


IDEAL WARRANT AND EPISTEMIC LUCK Of all Putnam's thrusts at "metaphysical realism", none has occasioned more outrage than his identification of truth with ideal justifiability. As he put it in an early statement of this view, "[T]ruth is an idealization of rational acceptability. We speak. as if there were such things as epistemically ideal conditions, and we call a statement 'true' if it would be justified under such conditions."12 What such statements entail and whether they are idealist depends on the underlying conception of justification itself. On relatively formalist conceptions of justification, the identification of truth with ideal warrant would be idealist. Though some of Putnam's early anti-metaphysical-realist statements encouraged this charge of idealism, his own attacks on positivism showed that those formalist conceptions of justification were distorted ideals of rational acceptability. Epistemology in the spirit of Putnam leads to a different understanding of the ideal warrant that is truth, an understanding that fits his emphatic assertions of a small "r" realism that is neither metaphysical realism nor transcendental idealism. 13 The positivists took rational inquiry to consist of responses to accumulated experience in which certain goals were pursued: the accurate recording of experiences, the formation of accurate expectations concerning future experiences, the development of economical means of inferring some experiences from others, and (perhaps) the facilitating of desire-satisfaction. The pursuit of these goals would consist of the cultivation of certain virtues in one's system of experience and belief. "Coherence, comprehensiveness, functional simplicity and instrumental efficacy" is a list that captures the major virtues the pursuit of which was supposed to convert mere cumulative experiencing into rational inquiry. These general virtues, exercised in working over the record of experience, were supposed to give any more specific principle whatever rational force it had; hence, my label "formalist". This conception of justification brings with it a distinctive notion of ideal justification. The ideal epistemic circumstance is one in which someone surveys all data that ever have been or will be given in experience, responding to it with hypotheses displaying the general virtues. This ideal follows from the positivist account of the goals pursued in rational inquiry. If accurate recording, accurate expectations, economical inference, and the facilitating of satisfaction are the goals of rational inquiry, then a hypothesis is not liable to any of the defects a rational inquirer seeks to avoid if all the data are in and the hypothesis fits them with the greatest attainable combination of accuracy, simplicity, and convenience. The identification of truth with this idealization of rational acceptability strikes many people as outrageously idealist. They protest that our hypotheses are rarely about the form or the content of our experiences and

102


insist that nonmentalistic hypotheses could be false no matter what their virtues in organizing the totality of experiences. These protests were understandable in response to some of Putnam's earlier statements about truth and ideal warrant. For example, my list of virtues was taken from Reason, Truth and History (1981), in a passage where Putnam identifies metaphysical realism with the view that science strives for a correspondence of thought and world transcending such virtues. 14 On the other hand, epistemology in the spirit of Putnam, which starts with the rejection of positivism as well and truly dead, leads to a different conception of empirical justification, which I sketched earlier in this essay. And this conception also. implies a distinctive epistemic ideal. In the postpositivist conception, a rational inquirer seeks the best explanations of the data, applying a framework of causal principles that is a minimal adjustment of prima facie truisms to the experiences that she and others have had and have sought. An ideal response to available experience will embody certain relatively formal virtues: the principles must be applied correctly and applied extensively enough to reveal any conflicts or pressing explanatory needs that would result from correct applications of principles to data. Rational inquirers acknowledge this ideal by seeking to avoid the corresponding defects of confusion and inattention. However, even an ideal way of responding may not provide an ideal basis for belief because of defects in the data-base itself. Perhaps an event that might have occurred would have revealed an inadequacy in the current repertoire of causal principles or would have dictated a revision in inferences employing those principles. Rational inquirers acknowledge such defects of incompleteness by accepting hypotheses tentatively and putting the hypotheses in further jeopardy when they are specially concerned with their truth. They try to reduce their dependence on epistemic luck, mere happenstance in the data employed in belief-formation. If the goal of inquiry were the economical summary and prediction of experiences and (perhaps) the maximization of actual satisfactions, then the ideal corresponding to this process of self-critical inquiry would be the old one. If all the data that will ever happen to occur are in and all are rationally processed, there is no dependence on luck, since no expectations will be disappointed. But the goal is the discernment of causes of experience. And the failures of positivist analyses of causation prevent the reduction of this goal to the development of means of summary and prediction. Even if all the data are in, full copfidence in the favored hypothesis will still require an independent assumption of epistemic luck. Perhaps the ultimate belief would be a rational falsehood due to epistemic misfortune, i.e., the non-occurrence of a circumstance that might have occurred and would have given rise to experiences revealing the falsehood of the ultimate hypothesis. The actual rational inquirer's fear of dependence on happenstance would not cease to be rational at an ideallirnit at which everything that will happen has happened.

103


In sum, rationality in discerning causes, unlike rationality in summarizing memories and fonning expectations, is liable to error when all experiences that will ever occur have occurred and have been ideally processed; the ideal circumstance, incompatible with error, is one in which no experience whose non-occurrence is accidental would have revealed the falsehood of a belief that ideally responds to the actual data, revealing its falsehood through the usual process of interpretation via causal principles and accommodation of old principles to new experience. From now on, this good fortune is what I will mean by "epistemic luck." One way to evoke this need for epistemic luck in coping with all data is to think of the great scientific revolutions of the past. Just as Aristotelian natural tendencies, gravitational attraction, and classical thermodynamic equilibrium were previously bases for false causal ascription, yet rationally responded to available data, the final causal repertoire might yield falsehood because it depends on an accidental limitation in total data. In describing epistemic luck, I have used the concept of truth itself, describing bad luck as missing experiences whose absence gives rise to rational falsehood. I do not think this ultimate circularity is avoidable. It would not do to say that the ideal circumstance is one in which some experience which might have occurred would have made belief revision a dictate of rationality. There is always such evidence, since skewed extra evidence can overturn true beliefs as surely as evidence overturns false beliefs in luckier processes. In response to this possibility of rational revision that does not lead toward truth, it makes no sense to speak of rational belief formation in light of some alleged totality in which all the evidence that might have occurred is actual. For one thing, the occurrence of some evidence often rules out the occurrence of other evidence that might have been revealing. The self-same lump cannot be both burned and dissolved at the same instant. Suppose, then, that truth is what would be justified if all the data were in, the adjustment of basic causal principles to accumulated experiences were formally ideal, and there is no experience that might have arisen which would have revealed the falsehood of the actually justified belief. This way of connecting truth with ideal warrant is not idealist, since the requirement of epistemic luck acknowledges nature's freedom not to fulfill our epistemic needs. This connection is not a means of defining away or eliminating the concept of truth-an untroubling lack at this point, since the rejection of positivism is, quite generally, an abandonment of reductioni~t demands. In this view, "truth that does not consist in ideal justification" is empty verbiage, distorting the real content of "truth"; but without reliance on the concept of truth, the nature of ideal justification is unspecifiable. Finally, though it avoids the wishful thinking about human capacities of the prior connection of truth with ideal warrant, i.e., the prior doctrine that total evidence and total reason must be revealing, this new connection

104


humanizes truth. It makes truth the satisfaction of goals that actual humans pursue, if they are responsible. Ordinary epistemic responsibility requires anxiety about epistemic bad luck, and efforts to remove it. I think that the new connection is the one Putnam intends to make when he associates truth with ideal warrant. For example, toward the end of Representation and Reality, Putnam notes the dependence of ideals of acceptability on "facts that it is within the capacity of speakers to determine, if they have the good fortune to be in the right sorts of circumstances." Refusing to define "the right sorts of circumstances", he continues, " ... I am not offering a reductive account of truth ...The suggestion is simply that truth and rational acceptability are interdependent notions ... truth depends on rational acceptability. But ... the dependence goes both ways: whether an epistemic situation is any good or not typically depends on whether many different statements are true."IS The claim that truth is ideal justifiability, on the construal that I have proposed, receives considerable support from the strategic account of content. "Truth" is an ordinary term of appraisal, whose use is dominated by the project of exchanging and storing information in the course of learning. The application of the term routinely conveys a certain appraisal of the proposition in question: belief in it does not essentially depend on some defect in the way the belief is formed. If someone has a false belief, we expect there to be some way that he or the authority he trusts went wrong: epistemic bad luck or a lack of rationality is responsible for the belief. If someone has a true belief, we regard any defect as inessential: with rationality and epistemic luck, the belief would have been the same. Thus, we routinely associate "truth" with acceptability at the ideal limit that I previously described. Of course, routine associations are a part of meaning only if reliance on them serves the dominant communicative purposes. In the case of "truth", reliance on our association with ideal justifiability does advance the relevant, informational projects. For, on the one hand, cooperative learning is advanced if assertion is regulated by criticism based on this ideal, in which discernment of a defect in belief formation leads to retraction (and withdrawal of the label "true"), unless the defect is held to be an anomaly, which disappears in ideal circumstances. On the other hand, if a belief does not essentially depend on departure from the ideal, it would,not advance the goal of learning about the world to withhold the most basic general term of positive appraisal from the proposition believed. For the failure to merit the appraisal would be a defect that no one could possibly seek to avoid. The ideal-warrant criterion is the broadest that makes the distinctions of interest, like the definition of "chair" as an object with a back and legs built to be capable of sustaining a human or humanoid's bottom. Further requirements are too restrictive for inclusion in meaning given the relevant interests, as the requirement that a chair have just four legs would be too narrow (or the

105


requirement that a plain natural kind correspond to current technical science's description of the structure underlying the stereotypical features).

BEYOND THE INTERNAL Declarations that truth is ideal justifiability are not Putnam's only way of rejecting metaphysical realism. Indeed, they are one aspect of a broad denial of separability, of the possibility of separating out human activities and interests without residue from the rest of cognition. " ... [W]e are forced to recognize with William James that the question as to how much of our web of belief reflects the world 'in itself and how much is our 'conceptual contribution' makes no more sense than the question: 'Does a man walk more essentially with his left leg or his right?",j6 "If one must use metaphorical language [to express Putnam's view of dichotomies between the subjective and the objective], then let the metaphor be this: the mind and the world jointly make up the mind and the world together.'tJ7 Here is another way in which postpositivist semantics and epistemology require the rejection of metaphysical realism. They sustain the following non separability thesis, which seems an important part of Putnam's claims of intermingling, not captured by my arguments so far: true belief, even true belief in a wholly external state of affairs, can never be analyzed into two parts, a wholly internal state and a' wholly external circumstance, cause, or basis. By something "wholly external", I mean something whose existence is entirely independent of the existence of mental activity on the part of the believer or those on whom she relies. By something "wholly internal", I mean something whose existence depends on mental activity and on nothing independent of it. Our beliefs are often concerned with the wholly external. This externality of subject matter is flamboyantly clear in the case of beliefs in states of affairs prior to the existence of any mental activity, such as our belief that the earth had a methane atmosphere before life evolved. On the other hand, it is not so obvious that anything is wholly internal (for example, Burge has his doubts). But I will not suppose that this is an empty category. The issue is whether the two categories capture every component of true belief. My claim is that rational activity, which is intrinsic to belief, permits no such dissection. The simplest separation thesis would be this: a true belief concerning a wholly external state of affairs consists of the belief, which is wholly internal, and a wholly external circumstance, making the belief true. But insistence on such internality for beliefs themselves is just what Putnam undermined in "The Meaning of 'Meaning'." Archimedes had beliefs about where gold is that would not have existed were it not for facts about ancient

106


Greek rocks that would have been the case in the absence of mental activity. Even when people do not rely on interpersonal exchanges of information-for example, when Supersubject tracks the Pest-belief is not wholly internal. So it would not do to gloss the externalist conclusion concerning belief about gold as a matter of mere convenience in correlating different people's internal states. After all, Supersubject herself would acknowledge that her having beliefs about the Pest depends on her actual success in dossier keeping. In general, the content of a belief depends on what capacities for detection the believer is exercising. And what capacities the believer has depends on something not wholly internal, the nature of her successes and failures in learning. A more complicated alternative, analogous to Fodor's distinction between narrow and wide content, is an attempt to locate the cut between internal and external within the phenomenon of belief. From this perspective, the whole fact that S has a true belief that p consists of a phenomenon wholly internal to S' s mental life (which does not itself have sufficient content to constitute S's belief that p), together with a wholly external circumstance that has an appropriate causal potential for constituting the true belief, when combined with the internal state. But what would be an appropriate wholly internal target for such a wholly external causal power? Even when a state is robustly mental and its occurrence is unequivocally dependent on an external fact, state and circumstances in combination can still fail to constitute a true belief. The familiar feeling of heat and dryness in my skin, in the circumstance in which the feeling is unequivocally dependent on sunburn, does not constitute belief that I suffer from sunburn. A mental phenomenon relevant to believing must be part of a rational response to experience available for rational deliberations of the one who responds and believes. If the positivists' dream canon existed, these requirements of rationality might simply impose a formal pattern in relations between wholly internal phenomena. The occurrence of an internal state appropriately related to other internal states would be an instance of generically rational activity; the influence of causal powers independent of mental activity would give rise to a true belief about a specific state of affairs. But the positivists' dream was empty. There is no such thing as rational inquiry in the abstract, ungrounded in inquiry into more specific topics. Rational inquiry depends on prima facie reliance on at least some of the topic-specific principles at the base of inquiry. And the possession of rationality, in this nonformal construal, is not wholly internal. For engagement with a topic requires relevant detection capacities. 18 For example, suppose a subject, S, has the true belief that a certain apple is red. Suppose, in addition, that S's belief does not involve deference to the detection capacities of others. The component of S' s true belief that

107


is not wholly external must include an instance of S' s sufficiently rational inferring and deliberation, or else S' s beliefs would not be in question. But the rationality of this coping must involve S' s inclination to fonn beliefs with specific topics on grounds of specific kinds-for such is the nature of the required rationality. In the case at hand, S, as an independent ascriber of physical redness, would be inclined to attribute physical colors on the basis of corresponding color appearances. To have this rationalizing inclination, S must be capable of attributing physical colors and, hence, must have at least an imperfect capacity to detect physical colors. But even an imperfect capacity to detect physical colors entails truly identifying them when one tries, on a variety of familiar occasions, should no special obstacle intervene. The truth of such an attribution of capacity depends on facts that would exist in the absence of mental activity. So the rationality of the nonexternal state guarantees that it is not a wholly internal state. I have supposed that S ascribes redness without deferring to the detection capacities of others. S is not, for example, red-green color-blind and reliant on the unaffticted in such matters. However, deference does nothing to reduce true belief to a combination of wholly internal target and wholly external cause. The mental activity of others to which the believer defers must be grounded on some person or group's nondeferential rational inquiry, and this requires at least imperfect detection capacities on their part. In a wholly red-green color-blind community, entirely isolated both historically and geographically, no one can attribute physical redness-even if all have a misguided idea that others use two words to mark off the crucial color distinction, rather as some of us confusedly think that experts distinguish gnats from midges. My example concerns a secondary quality, a property that has an approximate definition in terms of tendencies to give rise to certain kinds of sensations in nonnal humans. However, as connections with immediate experience become more flexible, indirect, and theoretical, the dissection of true belief into a wholly internal and a wholly external part becomes even more dubious. Because the relevant mental activity requires knowledge that a nonnally equipped rational inquirer might lack, it is even more obviously concerned with specific topics and dependent on specific detection capacities. Aristotle's beliefs concerning the presence of water were quite reliably prompted by the local prevalence of HOH molecules. But he did not believe that the Aegean mostly consisted of HOH molecules. The external cause of his beliefs lacked an appropriate mental target because he could not identify hydrogen and oxygen atoms and their chemically stable combination. Nonetheless, it might seem that the separability of true belief into wholly external and wholly internal parts is required by the possibility of sufficiently massive falsehood. Science fictions such as Putnam has explored might be the start of this objection. Consider people who are mere brains in a vat,

108


receiving programmed impulses that mimic our own neural stimuli. The brainpeople differ from us in external circumstances, which doom them to pervasively false beliefs concerning their circumstances (or so the objection begins). Our true believing is due to the wholly external circumstances in which we differ from them together with the states which we share with them. But these states must be wholly internal. After all, if both envatted brainpeople and we are in these states, their possession must be wholly independent of external facts. For the brainpeople are utterly mistaken about these facts. The objection misdescribes the sort of doom that produces false belief, as opposed to nonbelief.. If the envatted living brains in the science fiction have false beliefs about external facts, they must have capacities for detection that are blocked by their misfortune, capacities which would be effectively exercised in other circumstances. Interpreting the brains as false believers means seeing them as analogous to an astronomer who is unaware that the front lens of her telescope has been painted over. But then, the states that the brainfolks share with us believers are not wholly internal. They actually have detection capacities, which would enable them to detect colors when obstacles are removed-as a sleeping athlete, unlike a carrot, has the capacity to run a five-minute mile. Being someone with a blocked capacity to describe one's environment is not a wholly internal state. So vatic beliefs are only the same as ours if their possession is not a wholly internal state. Of course, the physiology and physics of vathood do not demand description in terms of capacity and interference. Some brain-vat stories encourage such description. Imagine that an evil scientist kidnaps normal humans in deep sleep, extracts their brains, and then .... But some versions make description in terms of blocked normal capacity quite unappealing. Suppose the brains, though physiologically identical with active human brains, have evolved through bizarre mutations from sponge corals in a mad scientist's vat. Or perhaps the brain-vat story is the one that Putnam explores in most detail in Reason, Truth and History (taking care to note its special absurdity even by the normal standards of science fiction): there simply always has been and always will be a universe consisting of automatic machinery tending a vat full of brains (see 6). Now it seems inappropriate to describe the brains as having blocked human capacities, capacities to detect colors, trees, chairs, and so forth. For just this reason, it seems wrong to describe them as having false beliefs. Perhaps, however, they detect qualities of brain impulses and have mostly true beliefs about them. Probably, there is no fact of the matter about which descriptions of vatic cases are correct-not because there are no semantic facts in general, but because the characterization of a strange tendency as an imperfect instance of our own detection capacities is made true by communicative rationales involving potential cooperation with strangers such as I described

109


in a previous section. In all these stories, cooperation between the envatted and vat-free is ruled out by the terms of the fiction. Still, if one does describe a vatic situation as one of massive falsehood, one must describe the brainfolks' state as one of blocked detection capacity, a description that is not wholly internal. 19 In addition to the worry about possibilities of massive falsehood, there is another concerning the apparent constitution of belief. The rejection of separability might seem to make a mystery of belief through strange denials of the completeness of sensible inventories of the constituents of beliefs. "What beliefs a person has," the objection begins, "is determined, at most, by her experiences, images (in all sensory modalities), feelings, behavior, environment, and dispositional relations among them. Perhaps we individuate her beliefs in ways that refer to phenomena at both sides of the line separating behavior from environment, the boundary between the internal and the external. Still, these are ways of sorting out bundles of phenomena consisting entirely of facts wholly located on one or another side of the boundary." (I have assumed that the objector includes behavior as part of mental activity. Otherwise, behavior itself might be divided up into itsintrospectible concomitants, on the one hand, and, on the other side of the boundary, correlated body movements.) The first step in this objection, in which the inventory is presented, is the basic mistake. Occurrences and dispositions, mental and behavioral, only determine someone's beliefs (when combined with her environment) if they are due to her sufficiently rational coping and inquiring. They must be part of her mental life in this way to characterize her believing. As Putnam has long insisted, occurrences and dispositions in an Earthling due to the manipulations of a Martian with a wonderful telescope and wonderful apparatus for creating feelings, images, and bodily movements on Earth would not constitute beliefs of the Earthling. But once the subject's rationality is required as a cause, belief is determined by a fact that is neither wholly internal nor wholly external. In sum, the demand for separability without residue neglects the fundamental role of rational activity in belief. (Because of this fundamental role, there is something misleading about the characterization of beliefs as "states." What makes believing like engagement in an activity is at least as important as the absence of voluntary control which prevents unqualified inclusion in that category.) Believing requires rational responding and coping, or else the subject's contribution will be too brutish or passive for belief. But rational activity must have some specific subject matters. Rational engagement with specific subject matters requires the possession of some detection-capacities. The possession of a detection capacity is not a wholly internal fact. Although it is misguided, the worry that true belief is a mystery, with-

110


out separability, reflects an important philosophical inclination, with roots in Hume's great work. According to this Humean inclination, all facts are nonrelational, in the final analysis. They are constructible, by logical operations, from facts that properties are possessed by individuals at particular points in space and time, none of which facts has intrinsic consequences for facts of property possession by other individuals or at other points in space and time. As Lewis puts it, in a principle of "Humean supervenience" that he endorses, " ... [A]ll there is to the world is a vast mosaic of local matters of particular fact, just one little thing and then another.... For short: we have an arrangement of qualities. And that is all.,,2o This metaphysical principle of separability is violated in the ascription of rational inquiry, since, on the most complete analysis, such ascription attributes nonlocal capacities to the subject, capacities whose possession has consequences for the world apart from the subject. If reductive analyses of causation-whether Humean regularity analyses or their counterfactual or probabilistic successors-are defective, then causal ascriptions violate the metaphysical principle of separability, as well. So, as Putnam emphasizes, reductive construals of causation and metaphysical realism do stand and fall together. 21 Finally, far from being scientific, the metaphysical principle of separability depends on the rejection of current physics, where property ascriptions are relational in the final analysis. The potential possessed by a point in a field has intrinsic consequences for the possible trajectories of charged particles. On any viable interpretation of the facts recorded in a quantum state description, their possession has intrinsic consequences for other possible eventualities. It is understandable that Hume might have taken himself to have remade philosophy in the image of the latest, Newtonian science. An educated layperson in his generation might well have taken insistence on anything more than the most general correlations among nonrelational facts to be just an archaic barrier to scientific progress. But in the next generation, a creative professional physicist, Kant, would insist, more scientifically, on the unavoidably relational character of Newtonian force ascriptions. In sum, insistence on the separability of the internal from the external in true belief applies a misreading of science to the analysis of the norms guiding science, often (ironically) in the name of eliminating mere folklore in favor of science. I will conclude this exploration of semantic non separability by extending it to a certain non separability thesis that has become Putnam's favored statement of his rejection of metaphysical realism. " ... [W]hile there is an aspect of conventionality and an aspect of fact in everything we say that is true, we fall into hopeless philosophical error if we commit a 'fallacy of division' and conclude that there must be a part of the truth that is 'the conventional part' and a part that is 'the factual part'." 22 My arguments so far about nonseparability and about ideal warrant

111


soon lead to this further warning about conventionality, when combined with the strategic approach to content. One moral of those arguments is that some minimally rational believing will be part of any component of true beliefthat is not wholly external. But belief always has some content. If the strategic approach is right, engagement with the appropriate content will consist of the pursuit of appropriate interests in learning and reliance on certain given response tendencies to coordinate relevant projects of learning. In this sense, the not-wholly-external part of true belief must have a conventional aspect. At the same time, the topic-specific interest in learning has turned out to require relevant detection capacities. So the not-wholly-external aspect is not wholly internal. It is factual, as well. As for the external fact making a belief true: it consists of the truth of the proposition believed, which is the same as the rational acceptability of the proposition in ideal circumstances. The failure of positivist projects tells us that the nature of those ideal circumstances will, ineliminably, depend on the belief in question, which depends on the believer's interests and given responses. So the external fact, too, has a conventional aspect.

CONCLUSION Putnam's attacks on metaphysical realism are sometimes treated as brilliant distractions, a series of tricky arguments that ought to be unraveled and dismissed so that they do not impede the serious business of ending the disarray brought on by the death of positivism. I confess to having shared this view, at times, despite the keenest appreciation of his work as a whole. Though I have not touched on all of the facets of Putnam's rejection of metaphysical realism or most of his arguments for such rejection, I hope that this essay contributes to a very different appraisal. The project of constructing a thoroughgoing replacement for positivism, motivated by Putnam's earlier work, leads to the rejection of metaphysical realism. Of course, rival projects can be pursued, and are, with great resourcefulness. One can amend, refine, and supplement classical positivism. Perhaps Bayes' theorem, the apparatus of counterfactuals, or structural descriptions derived from computation theory will be means of discovering the canons of justification, explanation, and belief attribution for which self-described positivists searched in vain. But is such resurrection really the way forward, after Putnam's criticisms, and Quine's and Goodman's and Kuhn's? For those of us who think: not, Putnam's attacks on metaphysical realism will be a vital guide in the new beginning.

112


NOTES 1. See "On the Dynamical Evidence of the Molecular Constitution of Bodies" in E. Garber et al., eds. Maxwell on Molecules and Gases (Cambridge, Mass.: MIT Press, 1986),219. 2. Most of my own version is presented in Fact and Method (Princeton: Princeton University Press, 1987), chapters 2, 4, and 10. 3. Even if the pursuit of a canon of topic-nuetrual rules produces philosophical distortions, it might seem the only fruitful project for psychologists who study how we learn. But this would be mere positivist prejudice, as well. The study of "domain-specific principles" by Spelke, Baillargeon, Gelman, and others, is at the forefront of current work on the basis of human learning. See, for example, Elizabeth Spelke, "Principles of Object Perception" and Rachel Gelman, "First Principles Organize Attention to and Learning of Relevant Data" in Cognitive Science 14 (1990); Renee Baillargean, E. Spelke, and S. Wasserman, "Object Permanence in Five-Month-Old Infants," Cognition 20 (1985). 4. Perhaps sufficiently detailed mental doings, actual and conditional, would make the"constraint of practical rationality sufficient by itself. But perhaps not. It is not clear how busy the internal life of a rational believer must be. Moreover, beyond vague and elementary needs, based on aversion to hunger, pain, and the like, the goals that a rational person has depend on the beliefs that would be rational on her part. So belief justification cannot be purged from practical rationality. In any case, we attribute contents to individual beliefs in ignorance of the total belief-set. And here both constraints have useful roles to play. 5. See especially Representation and Reality (Cambridge, Mass.: MIT Press, 1988), chapter 5. 6. See Gareth Evans, "The Causal Theory of Names" (1973) in his Collected Papers (Oxford: Oxford University Press, 1985), 16; Evans, The Varieties of Reference (Oxford: Oxford University Press, 1982),272; H. P. Grice, "Vacuous Names" in D. Davidson and 1. Hintikka, eds. Words and Objections (Dordrecht: Reidel, 1969), 14lf. 7. My story of Supersubject and the Pest was essentially the same as the story of Mary's aggressive cello playing that Stalnaker uses, in Inquiry (Cambridge, Mass.: MIT Press, 1984), to expose the inadequacy of the view that beliefs and desires are simply dispositional states of a potentially rational agent (17f). In drawing morals from a similar fable, my main disagreement will be the denial that reference is fixed by a "naturalistic" causal relation, which can be described without reliance on intentional or epistemic notions (compare, for example, ibid., 25). Also, my description of content-deterrning mental activity will imply a more articulated account of propositional structure than Stalnaker's. 8. In unsuccessful reference, it is not the case that the believer has a capacity to detect whatever he attributes. Thus, physicists who attributed caloric lacked a capacity to detect caloric. No one can detect whatever he attributes. Thus, physicists who attributed caloric lacked a capacity to detect caloric. No one can detect what does not exist. Still, the availability of the proposition in question as a subject for belief will depend on detection capacities combined with rationality. Physicists believed in caloric in virtue of their ability to detect heat transfer together with their rationales for explaining phenomena of heat transfer as due to the vicissitudes of an incompressible fluid. 9. Much of Grice's essay, "Method in Philosophical Psychology (From the Banal to the Bizarre)" (1975) might be used to show how relatively few crit'erial principles can license a rich variety of empirically justified psychological ascriptions without defining psychological notions in a nonintentional, nonnormative vocabulary. See H. P. Grice, The Conception of Value (Oxford: Oxford University Press, 1991). 10. For simplicity's sake, I have sketched the communicative rationales with a bearing on ascriptions of "gold" as if they concerned "gold" alone. But really, the discriminations pursued using the whole contrasting set of plain natural-kind terms determine the projects relevant to the use of each to refer to a natural kind. This is why someone ignorant of the preciousness of gold could still competently use "gold" to refer to a natural kind. Similarly,

113


11. 12. 13.

14. 15. 16. 17. 18.

19.

20. 2i. 22.

114

"pure gold" and "Au" do not have the same content, even though in this particular case we do not expect pay-offs from open-mindedness about the identification, even though this particular identification approaches common lore. For there are benefits in general to openmindedness about the technical specification of plain natural-kind terms. I fill in this sketch, and apply it to moral discourse, in Moral Differences (Princeton: Princeton University Press, 1992). "Realism and Reason" (1976) in Meaning and the Moral Sciences (London, 1982), 125. This realism is especially emphatic in such recent writings as The Many Faces of Realism (Lasalle: Open Court, 1987), lecture I, and Representation and Reality (Cambridge University Press, 1988), chapter 7. See Reason, Truth and History (Cambridge University Press, 1981), 134. Representation and Reality, 115. The Many Faces of Realism, 77. Reason, Truth and HIstory, xi. Because the proposition that p is true, there is no need, here, to unravel the difficult question of how belief is achieved when reference fails (as when people fall into the trap of believing that Ossian wrote ancient epics or that heat transfer is the diffusion of caloric). But such believers must exercise some capacities for detection, or they would not even have false beliefs. Caloric theorists were misguided about heat. Ossianists were misguided about epics. Hence, they were capable of detecting heat, or epics. (Compare cats, who are quite incapable of being taken in by literary hoaxes.) In Reason, Truth and History. Putnam was even less tolerant than I have been toward the ascription of massive false belief to brainfolk in the universe that always consists of the brains-vat-automaton situation. He took this ascription to be false, not just the consequence of an unappealing descriptive option, since the brains would lack any causal connection with trees, chairs, and the other subjects of the alleged false beliefs. But I think my response to semantic characterizations of vatic stories is in the spirit of his insistence that semantic facts are not inherent in physical and physiological connections. Indeed, in Reason, Truth and History there is a step toward the admission of blocked capacities as sufficient for reference in a suggestion that future causal connection with a natural kind might be a basis for certain means of reference to it at present (16). At any rate, either Putnam's response or mine rules out the use of vatic stories as a means of establishing separability of actually true beliefs into wholly internal and wholly external components. Either we do not share beliefs with the envatted (as Putnam would say), or the hypothesis that beliefs are shared depends on a characterization that is not wholly internal (as I would say). Philosophical Papers, vol. 2 (New York: Oxford University Press, 1986), ix-x. See, for example, "Is the Causal Structure of the Physical Itself Something Physical?" (1984) in Realism with a Human Face. See Realism with a Human Face, Preface, x, where Putnam notes that the importance this claim has assumed for him marks a shift in emphasis beginning with The Many Faces of Realism. In the latter, see, for example, 40, 77.

Conant, James, The Search for Logically Alien Thought: Descartes, Kant, Frege, and the "Tractatus" , Philosophical Topics, 20:1 (1992:Spring) p.115

PHILOSOPHICAL TOPICS 20 No.1, FALL 1991

VOL.

The Search for Logically Alien Thought: Descartes, Kant, Frege, and the Tractatus 1 James Conant University of Pittsburgh

[l]n order to draw a limit to thought we should have to be able to think both sides of this limit (we should therefore have to be able to think what cannot be thought). The limit can, therefore, only be drawn in language and what lies on the other side of the limit will be simply nonsense. -Ludwig Wittgenstein 2 The only proper way to break an egg is from the inside. -Parva Gallina Rubra 3

This essay is about three things: Wittgenstein's ideas concerning the question of the possibility of illogical thought, the sources of those ideas (especially in Kant and Frege), and Putnam's recent interest in both of these matters. Along the way, this paper briefly sketches the broad outlines of two almost parallel traditions of thought about the laws of logic: one rather long and complicated tradition called the History of Modern Philosophy, and one rather short and complicated one called Hilary Putnam. Here is a thumbnail version of how these two traditions align: Descartes thought the laws of logic were only contingently necessary; not so recent Putnam agreed. St. Thomas Aquinas believed that they were necessarily necessary; relatively recent Putnam agreed (this is only confusing if you think Aquinas should not

115


be a step ahead of Descartes). Kant thought they were simply necessary. Frege wanted to agree-but his manner of doing so raised the worry that there was no way in which to express his agreement that made sense. Wittgenstein agreed with the worry. He concluded that sense had not (yet) been made of the question to which our two traditions sought an answer; very recent Putnam agreed.

HISTORICAL PREAMBLE: A DIFFERENT KIND OF CARTESIANISM What is the status of the laws of logic, the most basic laws of thought? Wherein does their necessity lie? In what sense does the negation of a basic law of logic represent an impossibility? The Scholastics were forced to think hard about these questions since they believed in the existence of an omnipotent God for whom all things are possible. If God is omnipotent does that mean that He has the power to abrogate the laws of logic? The Scholastics, on the whole, were quite reluctant to draw this conclusion. But does that then mean that God is not allpowerful, that there is a limit to his power, that there is something he cannot do? That is a conclusion that the Scholastics were, on the whole, at least equally as reluctant to draw. Posed here in a theological guise is a version of a question that has continued to haunt philosophy up until the present: do the laws of logic impose a limit which we run up against in our thinking? If so, what kind of a limit is this? Do their negations represent something that we cannot do or that cannot be? If so, what sort of "cannot" is this? Here is Aquinas's attempt to reconcile the omnipotence of the Divine Being with the inexorability of the basic principles of Reason: All confess that God is omnipotent; but it seems difficult to explain in what His omnipotence precisely consists. For there may be a doubt as to the precise meaning of the word "all" when we say that God can do all things. If, however, we consider the matter aright, since power is said in reference to possible things, this phrase, God can do all things, is rightly understood to mean that God can do all things that are possible; and for this reason He is said to be omnipotent. Now ... a thing is said to be possible in two ways. First, in relation to some power ... If, however, we were to say that God is omnipotent because He can do all things that are possible to His power, there would be a vicious circle in explaining the nature of His power. For this would be saying nothing else but that God is omnipotent because He can do all that He is able to do. It remains, therefore, that God is called omnipotent because He can do all things that are possible absolutely; which is the

116


second way of saying a thing is possible. For a thing is said to be possible or impossible absolutely, according to the relation in which the very terms stand to one another: possible, if the predicate is not incompatible with the subject, as that Socrates sits; and absolutely impossible when the predicate is altogether incompatible with the subject, as, for instance, that a man is an ass. . .. Therefore, everything that does not imply a contradiction in terms is numbered among those possibles in respect of which God is called omnipotent; whereas whatever implies contradiction does not come within the scope of divine omnipotence, because it .cannot have the aspect of possibility. Hence it is more appropriate to say that such things cannot be done, than that God cannot do them. Nor is this contrary to the word of the angel, saying: No word shall be impossible with God (Luke i.37). For whatever implies a contradiction cannot be a word, because no intellect can possibly conceive such a thing.4

Aquinas is caught here between the Charybdis of asserting a mere tautology (God can do everything within His power) and the Scylla of implicitly ascribing a substantive limit to God's power (by declaring God can do all those sorts of things which fall under a certain general description X, and hence apparently implicitly declaring: He cannot do those things which do not fall under X). One way out-a way out which, as we shall see, is gradually refined in the course of these two traditions of thought about logicwould be for this description (of those things which God cannot do) to tum out not to be a genuine description at all. Aquinas, indeed, tries to argue that those things which fall under the (apparent) description things which God cannot do are not, properly speaking, things which can be done at all. These are things which "cannot have the aspect of possibility." Of these, Aquinas says, "it is more appropriate to say that such things cannot be done, than that God cannot do them." But the worry arises: hasn't Aquinas just offered us a redescription of what kind of a thing a logically impossible sort of a thing is? It would seem that we still have here to deal with a certain (albeit remarkable) kind of a thing. If so, the question remains: what sort of a thing is this and is it something not even God can do? Even if we concede to Aquinas that perhaps, strictly speaking, we should not speak of it ~s if it were a doable kind of a thing, nevertheless, there certainly still appears to be an "it" here that our words are straining after and which has formed the subject of our thought throughout the preceding paragraph. Aquinas appears to be on his strongest ground when he tries to make out that the "it" which falls under these descriptions-"that which is logically impossible," "that which even God cannot do"-is not a kind of a thing at all. What we have here instead is an attempt to conceive of a kind of a thing which "no intellect [i.e., not just a human intellect] can possibly conceive;"

117


it is an attempt to speak. a word "which cannot be a word." In order to set up this way of dissolving the appearance of an "it" (which not even God can do), Aquinas invokes Aristotle's distinction between those things which are impossible in relation to some power and those things which are impossible absolutely. It is not clear, however, that this distinction really helps. It threatens to recreate the appearance that we have to do here with two different kinds of things, belonging to two distinct orders of impossibility: the merely impossible and the absolutely impossible. Just as it is natural to picture that which is possible for a finite being (such as man) as contained within the space of that which is possible for God, it can seem natural to take Aristotle's distinction as marking an analogous boundary, only at a higher level. One pictures the distinction in terms of two degrees of impossibility: things belonging to the second degree (the absolutely impossible) are situated on the far side of the outer limit which encompasses things belonging to the first degree (the merely impossible). So now it seems that although God never chafes against anything which lies within the circumference of this exterior circle, nonetheless, Great as He is, that is as far as He can go---even He must remain within this circle. This picture of a circle (circumscribing the limits of that which is absolutely possible) lying within a wider space (the space of the absolutely impossible) inevitably leaves us with the feeling that we have, after all, succeeded in describing a genuine limit to His power. The existence of this outer space of absolute impossibility seems to settle the question in precisely the contrary direction from the one in which Aquinas had hoped to lead us. The apparently innocent step of picturing the space of absolute possibility as bounded by a limit seems to have led us to the opposite conclusion about God's ornnipotence. 5 What sort of a thing lies beyond the limit of God's power? Answer: the sort which is absolutely impossible. And now it becomes irresistible to add: even for Him. Descartes concluded that Aquinas, along with most of the rest of medieval theology, had wandered into blasphemy.6 "If men really understood the sense of their words,"7 they would never speak. as they do. For their mode of speech clearly implies a limit to God's power. The only way to avoid such blasphemy is to refrain from ascribing any limits to what the Divinity is able to bring about: I turn to the difficulty of conceiving how God would have been acting freely and indifferently if he had made it false ... in general that contradictories could not be true together. It is easy to dispel this difficulty by considering that the power of God cannot have any limits, and that our mind is finite and so created as to be able to conceive as possible the things which God has wished to be in fact possible, but not be able to conceive as possible things which God could have made possible, but which

118


he has nevertheless wished to make impossible. The first consideration shows us that God cannot have been detennined to make it true that contradictories cannot be true together, and therefore that he could have done the opposite. The second consideration assures us that even if this be true, we should not try to comprehend it, since our nature is incapable of doing so.s

Descartes positively asserts here that God could have made contradictories true together. 9 He further asserts that this means that God can bring about thiVgs which our minds are incapable of comprehending. If only that which is comprehensible to minds such as ours were possible for God-if fundamental truths (sucl). as that contradictories cannot be true together) were external and prior to God's will-then He would not be omnipotent. For His will would not be free with respect to such truths, but rather subject to their determination. But this would be to deny the infinitude and incomprehensibility of God's power. lO The only way to avoid such an unworthy blasphemy is to acknowledge that such truths do depend upon the will of God and that it lies within His power to bring about the negations of such truthS.11 The sense in which they are nonetheless necessary or eternal lies in the fact that God has decreed them to be true: hence they are necessary for us. But, from a Divine point of view, they are only contingently necessary. For we must allow that there is some sense in which God could have done otherwise: [E]ven if God has willed that some truths should be necessary, this does not mean that He has willed them necessarily; for it is one thing to will that they be necessary, and quite another to will this necessarily, or to be necessitated to will itP

If God had not been free to choose such laws as he did, if He were by necessity constrained to will the truth of the laws of logic, then there would be a necessity that binds even Him. God would be inexorably subject to those laws, just as we are subject to His decrees. There would be afatum that binds even the Divinity, making a mockery of his alleged omnipotence. So we must say that God freely willed the laws of logic to be true. Descartes is very careful, however, to insist that, although these laws do not bind God, this does not make them any less binding for us. The hubris lies in our thinking that because we cannot comprehend how the negation~ of such laws could be true-for example, how it could be true that "He could have made contradictories true together"-we are therefore in a position to conclude that it cannot be done, even by Him. It is hubris to think that the limits of our powers of comprehension enable us to specify something He cannot do: In general we can assert that God can do everything that is within our grasp but not that He cannot do what is beyond our grasp. It would be rash to think that our imagination reaches as far as his power,13

119


[S]ince God is a cause whose power surpasses the bounds of human understanding, and since the necessity of these truths does not exceed our knowledge, these truths are something less than, and subject to, the incomprehensible power of God. 14

Descartes, nonetheless, wants to be able to say: we think rightly when we think in accordance with these laws. We perceive correctly when we clearly and distinctly perceive the truths of logic to be in some sense 'necessary': they are necessary in our world. But Descartes will not follow Aquinas and say that their negations are absolutely impossible. They are not, as it were, necessarily necessary: God could have created a very different sort of a world. Of course, since our powers of conception are constrained by the principles of logic, Descartes must say that we cannot make any sense of the possibility of such a world-nonetheless, we should admit the mere possibility of its existence: [T]here is no need to ask how God could have brought it about from eternity that it was not true that twice four make eight, and so on; for I admit this is unintelligible to us. Yet on the other hand I do understand ... that it would have been easy for God to ordain certain things such that we men cannot understand the possibility of there being otherwise than they are. 15

Descartes concedes that any attempt on our part to comprehend such a world must meet with failure. This raises the worry: doesn't Descartes's position ultimately collapse into Aquinas's? What are we to make of his assertion that we should believe in the possibility of such a world even though he himself freely admits that we cannot hope to comprehend it? How does one undertake to believe in something one cannot understand? Descartes himself feels at least some of the force of this problem. In an attempt to get around it, he helps himself to a fine distinction-a distinction between our being able to conceive of such a world and our being able to conceive that such an inconceivable world could be. The possibility of such a world is not something we can comprehend, but it is something we can apprehend. 16 Descartes's own way of expressing this slippery distinction is to say that the ultimate contingency of these truths (which we take to be necessary) is not something we can embrace in our thought, but we can touch it in our thought: I know that God is the author of everything and that these [eternal] truths are something and consequently that He is their author. I say that I know this, not that I conceive it or grasp it; because it is possible to know that God is infinite and allpowerful although our mind, being finite, cannot grasp or conceive Him. In the same way we can touch a mountain with our hands but we cannot put our arms around it as we could put them around a tree or something else not too large for them. To grasp something is to embrace it in one's thought; to know something, it is sufficient to touch it with one's thoughtY

120


We have here what I will call "the Cartesian Predicament": We want to frame a thought (about that which cannot be thought) but we run up against the problem that the thought we want to frame lies in its very nature beyond our grasp.IS We need a way to pick up this thought by the comer without fully taking it into our hands. We need a way to think right up close to the edge of the limit of thought, close enough to get a glimpse of the other side. Descartes's distinction between what we can embrace in thought and what we can only touch in thought is an attempt to characterize what is involved in trying to think both sides of the limit. In drawing this distinction, Descartes concedes that in order for us to be able to properly grasp an illogical thought, our minds would have to be constituted otherwise. We, with our finite powers of conception, simply cannot grasp what it would be like for the fundamental principles of our thought to be false. Nonetheless, we can make contact in our thought with the mere possibility that they might be. The Evil Demon Hypothesis is the way the author of the Meditations touches upon such a possibility in his thought. He apprehends what he cannot comprehend: namely, that even his most clear and distinct perceptions of truth might have been implanted in him by a Creator who wished to deceive him. Although the most radical doubts voiced in the First Meditation (the Evil Demon Hypothesis and the meditator's doubt about his own sanity) are in the end ultimately to be overcome, it is important for Descartes that these doubts represent minimally intelligible possibilitiespossibilities we can sidle up to in our thought, even if we cannot wrap our minds all the way around them. To insist upon the absolute impossibility of an Evil Deceiver would be blasphemy; it would be another way of insisting upon a limit to God's power. 19 The reason we should assert that God does not deceive us is not because we are in a position to claim that it is absolutely beyond His power to do so, but rather, because (if we have an adequate idea of God) we can clearly and distinctly perceive that He is infinitely benevolent and hence would choose not to do so.2oln His benevolence, He arranged it so that the principles which bind our thought enable us to think in accordance with the truth. He created our minds so that our clear and distinct ideas would correspond to the necessities of this world, the one that He created as our habitat. The principles of thought, implanted in us by our Creator, are so ordered that they are in harmony with the fundamental principles to which the natural world accordsY This brings us to a crucial tenet of the Cartesian conception of logic: a logical contradiction is something which is naturally repugnant to our reason. Just as God has failed to give us the power to genuinely withhold assent from what we clearly and distinctly perceive, so He has failed to give us the power to affirm that which is utterly repugnant to the natural light of reason.22 The necessity of the laws of logic is to be accounted for by the fact that our minds are so constituted that we cannot help but think in accordance with

121


them. The basic principles of human thought articulate, as it were, the mechanics of the human mind-the optics of the natural light of reason. Their appearance of necessity is simply due to a general fact about our mental constitution: namely, that our Creator endowed us with these (rather than some other) fundamental principles of thought. That we find logical contradictions repugnant is a contingent fact about the structure of our thought. Descartes is perfectly aware of this implication of his doctrines: I do not think we should ever say of anything that it cannot be brought about by God. For since every basis of truth . . . depends on his omnipotence, I would not dare to say that God cannot make a mountain without a valley, or bring it about that I and 2 are not 3. I merely say that He has given me such a mind that I cannot conceive a mountain without a valley, or a sum of 1 and 2 which is not 3; such things involve a contradiction in my conception [myemphasis].23

He could have made contradictories true together. 24 We cannot comprehend this, although we can know it. We can acknowledge that God can do this without being able to fathom it. The attempt to fathom such a possibility would involve us in an effort to think a kind of thought which is logically alien to us. That we cannot do this is due to an ultimately contingent fact about our minds; it is due to how God made them. We cannot think in this other way because of the sort of mind He has given us. That which is logically alien to our minds does not therefore represent an absolute impossibility (in Aquinas's sense), but only something which is incomprehensible to us and hence seems, to our finite intellects, to be absolutely impossible. This suggests the following Cartesian diagnosis: Aquinas underestimated the power of God by overestimating the power of human reason-he mistook the limits of human comprehension for the limits of absolute possibility. Given that Descartes usually figures in a story about the history of philosophy as the archetypical Rationalist, there is a certain irony in the fact that, with respect to the philosophy of logic, Cartesianism would appear to represent the position that even the most basic principles of reason are only contingently necessary truths. This is a position most of the classical empiricists would have recoiled from in horror. Consequently, although the label 'Cartesian' is often used to name the opponent of the 'Empiricist', in the philosophy of logic, 'Cartesianism' can properly be taken to stand for the view that the laws of logic are only contingently necessary-they are the laws according to which we cannot help but think.25 Considered in this light, certain forms of radical empiricism can be viewed as species of Cartesianism. Margaret Wilson sums up the historical significance of this region of Descartes's thought by casting him as the forerunner both of Kant's account of necessity (in terms of the "structure and workings of our own minds") and of the most stridently naturalistic current in contemporary philosophy.26 She

122


cites (not-so-recent) Hilary Putnam as an example of a contemporary Cartesian about logical 'necessity'. In support, she quotes a passage in which Putnam allows himself to describe the shift from Euclidean to Riemannian cosmology as a case in which "something literally inconceivable turned out to be true."27 Putnam's example here (of something inconceivable which turned out to be true) is the statement "one cannot return to the place from which one started by travelling in a straight line in space in a constant direction." Putnam goes on to describe the moral which he drew from this development in cosmology: I was driven to the conclusion that there was such a thing as the overthrow Of a proposition that was once a priori (or that once had the status of what we call an a priori truth). If it could be rational to give up claims as self-evident as the geometrical proposition just mentioned, then, it seemed to me that there was no basis for maintaining that there are any absolutely a priori truths, any truths that a rational man isforbidden to even doubt. 28

After quoting this passage, Wilson comments on the "Cartesian elements" she finds in evidence here: "the generalized suspicion of 'inconceivability' as a basis for claims about what cannot be, and a consequent attenuation (at least) of the concept of 'necessary truth."'29 Wilson's narrative about the place of Cartesianism in the history of modem thought about logical necessity, from Descartes through Kant to Putnam, prepares the way for three further ironies which will preoccupy us in the pages to follow: firstly, Kant's views about logical necessity, on Putnam's reading of them, will not tum out to be a way-station between Descartes and not so recent Putnam; secondly, Kant turns out to be the father of a stridently anti-Cartesian tradition which runs through Frege to Wittgenstein; thirdly, very recent Putnam's views on logical necessity will tum out to be (roughly) Kantian (and hence stridently anti-Cartesian).

A VERY RECENT PUTNAM There is certainly something to the thought that certain classic papers of Putnam and Quine30 offer perhaps the closest thing to be found in twentiethcentury philosophy to an attempt to rehabilitate Descartes's claim that it would be hubris for us to assert of an omnipotent God that He would be inexorably bound by the laws of logic-those laws which happen to bind our finite minds. In a move which is characteristic of much of contemporary naturalistic thought (both in and out of the academy), science is substituted for God. Cartesianism in the philosophy of logic, freed of its theological trappings, becomes the view that it would be hubris for us to assert of the

123


ongoing activity of scientific inquiry that it will be forever bound by the laws of classical logic-those principles which happen to be most fundamental to our present conceptual scheme. The contrast is now no longer, as in Descartes, between the finite powers of man and the omnipotence of God, but rather between the finite limits of present scientific thought and the infinite possibilities latent in the future of science as such. According to this contemporary accusation of hubris, the laws of logic are merely part (however basic a part) of our best current scientific theory of the world. We should, with proper empiricist humility, hold them to be at least in principle revisable in the course of some major theoretical reconstruction that future scientific research may require of us. If Descartes is led by a sense of theological piety to insist that God can do anything-no matter how inconceivable it may be to us-the contemporary ultra-empiricist is led by an equally fervent sense of naturalistic piety to insist that the science of the future might require a revision of any of our present axioms of thought-no matter how unacceptable such a revision might seem by our present lights. The exploration of the contours of possibility belongs to the business of the physicists. In this regard, we philosophers must issue them a blank checkit would compromise our standing as underlaborers to put a ceiling on how much they can spend. To paraphrase Descartes on God: we must not conclude that there is a positive limit to the power of science on the basis of the limits of our own (present) powers of conception. All of its hostility to theology notwithstanding, this contemporary form of piety is, in a sense, no less religious (in its unconditional deference to a higher authority) than Descartes's-it has simply exchanged one Godhead for another. But, unlike Descartes, precisely because it is overtly hostile to theology, it is able to easily blind itself to the fact that it is a form of piety. In a paper entitled "There is at Least One A Priori Truth,"3l a relatively recent Putnam sheds his piety and argues that there are, after all, a priori truths in exactly the sense that less recent Putnam and (any vintage of) Quine had famously been concerned to deny that there could be. At least one truth is unrevisable, Putnam now declares, in the sense that it would never be rational to give it up. Putnam's candidate for such an a priori truth is the minimal principle of contradiction, the principle that not every statement is both true andfalse. Putnam's strategy is to try to argue that there are no circumstances under which it would be rational to give up this principle, and therefore that it provides us with an example of at least one "absolutely, unconditionally, truly, actually a priori truth."32 Putnam wishes to quarrel with the claim that a fundamental logical law is merely contingently necessary. This leads him occasionally to assert an opposing claim, to declare that a fundamental logical law must be necessarily necessary. Indeed, one can hear an echo of Aquinas's distinction between the merely and the absolutely impossible in passages such as the following:

124


The statement ... "This sheet of paper is red and this sheet of paper is not red" ... simply asserts what cannot possibly be the case. And the reason that "when I open the box you will see that the sheet of paper is red and the sheet of paper is not red" does not count as a prediction, is that we know-know a priorithat it can't possibly tum out to be the case [my emphases).33

Putnam adduces in the course of the paper a number of arguments, which I will not rehearse here, that purport to show that the principle of minimal contradiction plays a role in our reasoning which is "prior to anything that might be offered as an explanation of its truth"34 and hence also prior to anything which might count against its truth. Putnam summarizes the conclusion of his paper as follows: The idea is that the laws of logic 35 are so central to our thinking that they define what a rational argument is. This may not show that we could never change our mind about the laws of logic, i.e. that no causal process could lead us to vocalize or believe different statements; but it does show that we could not be brought to change our minds by a rational argument . .... [The laws of logic] are presupposed by so much of the activity of argument itself that it is no wonder that we cannot envisage their being overthrown ... by rational argument. 36

Has Putnam here exchanged one fonn of piety for another: a piety about natural science for a piety about logic? This is the problem that exercises very recent Putnam: how to avoid one of these fonns of piety without falling into the other. So much, for the moment, for relatively recent Putnam. I will be primarily concerned here with very recent Putnam. In particular, I want to try to follow up and flesh out some intriguing claims made in a very recent paper: historical claims about how to understand a tradition of thought about logic (one which runs from Kant through Frege to the Tractatus) and philosophical claims about what is involved in attempting to think the negation of a logical truth. The locus of these claims is a paper titled "Rethinking Mathematical Necessity."37 (I will only discuss those aspects of the paper which bear on the topic of logical necessity.) Putnam turns his attention here once again to the ancient and honorable question: what is the status of the laws oflogic-analytic or synthetic, a priori or a posteriori? As one has come to expect of Putnam, he approaches the question afresh, defending a conception of logical necessity which he claims to (now) find in later Wittgenstein. Putnam says at the outset of the paper that he sees contemporary philosophy as faced with two equally unsatisfying alternatives-alternatives he associates with the names of Camap and Quine respectively: a linguistic conventionalism, on the one hand, according to which the laws of logic are analytic truths, and a naturalized epistemology, on the other,

125


according to which they are synthetic a posteriori and hence not dissimilar in kind from ordinary empirical truths (only-so the mixed metaphor goesfar more deeply entrenched in our web of belief). After canvassing these standing responses to the question, Putnam turns his attention toward what he calls "a very different line of thinking-one which goes back to Kant and Frege." He continues: This line is one I believe Carnap hoped to detranscendentalize; and in Carnap's hands it turned into linguistic conventionalism. My strategy in this essay will be to suggest that there is a different way of stripping away the transcendental baggage, while retaining what I hope is the insight in Kant's and perhaps Frege's view, a way which has features in common with the philosophy of the later Wittgenstein rather than with Carnap.

The invocation of Kant and Frege here might lead one to presume that very recent Putnam is simply concerned to uphold the conclusions put forward by relatively recent Putnam in "There is at Least One A Priori Truth." For, as we shall see, there is much in Kant and Frege that rhymes with those conclusions-that the laws of logic are absolutely central to our thought, that they define what rational argument is, that they are prior to anything which might be offered as an explanation of their truth. Putnam, however, as the above passage indicates, is now after a view he finds in later Wittgenstein: his interest in Kant and Frege is as stepping-stones to that view. Putnam's concern in the paper is in part to trace the roots of the later Wittgenstein's views on the nature of 'grammatical propositions' through a tradition of thought about logic which begins with Kant and runs through Frege and early Wittgenstein. This is how Putnam tells the story: Kant's Lectures on Logic contain one of the earliest-perhaps the earliest-polemic against what we now call 'psychologism' ... [W]hat interests me here ... is closely related to [that polemic] ... What interests me ... is to be found in the Critique of Pure Reason itself, as well as in the Lectures on Logic, and that is the repeated insistence that illogical thought is not, properly speaking, thought at all ... It is this that brought home to me the deep difference between an ontological conception of logic, a conception of logic as descriptive of some domain of actual and possible entities, and Kant's (and, I believe, Frege's). Logic is not a description of what holds true in "metaphysically possible worlds," to use Kripke's phrase. It is a doctrine of the form of coherent thought. Even if I think of what turns out to be a 'metaphysically impossible world', my thought would not be a thought at all unless it conformed to logic. Indeed, logic has no metaphysical presuppositions at all. For to say that thought, in the normative sense of judgment

126


which is capable a/truth, necessarily conforms to logic is not to say something which a metaphysics has to explain. To explain anything presupposes logic; for Kant, logic is simply prior to all rational activity.

While I would not claim that Frege endorses this view of Kant's, it seems to me that his writing reflects a tension between the pull of the Kantian view and the pull of the view that the laws of logic are simply the most general and most justified views we have. If I am right in this, then the frequently heard statement that for Frege the laws of logic are ... [the] "most general laws of nature" is not the whole story. It is true that as statements laws of logic are simply quantifications over "all objects"-and all concepts as well-in Begriffsschrift. There is no "metalanguage" in Frege, in which we could say that the laws of logic are "logically true"; one can only assert them in one language, the language. But at times it seems that their status, for Frege as for Kant, is very different from the status of empirical laws. (It was, I think, his dissatisfaction with Frege's waffling on this issue that led the early Wittgenstein to his own version of the Kantian view.) It was this line of thinking that helped me to understand how one might think that logical laws are sinnlos without being a Carnapian conventionalist. Laws of logic are without content, in the Kant-and-possibly-Frege view, insofar as they do not describe the way things are or even the way they (metaphysically) could be. The ground of their truth is that they are the formal presuppositions of thought (or better, judgment). Carnap's conventionalism ... was an explanation of the origin of logical necessity in human stipulation; but the whole point of the Kantian line is that logical necessity neither requires nor can intelligibly possess any "explanation."

The preceding quotation has a lot packed into it. We are being offered roughly the following capsule history of a tradition of philosophical thought about logic: I) Kant held that illogical thought is not, properly speaking, thought at all. 2) Frege inherited this view from Kant. 3) Frege held another view of logic as well-one according to which the laws of logic are the most general laws of nature. 4) These two views oflogic are in tension with one another. S) The early Wittgenstein's view (that the propositions oflogic are sinnlos) should be read as attempting to resolve this fundamental instability in Frege's philosophy.

Putnam then goes on, later in the paper, to argue that the crucial idea here, with which he himself is in sympathy, is that logical truths do not have

127


negations that we are able to understand. It is not that these propositions represent a content that we grasp and then reject as false; rather, we are simply unable to make sense of these propositions in a way which allows the question of their truth or falsity to arise in the first place. As he puts it at one point: "the negation of a theorem of logic violates the conditions for being a thinkable thought or judgment. "38 Putnam argues that it is out of this idea that the later Wittgenstein's view of logical propositions develops, and so Wittgenstein's later view is best understood against the background of this tradition of thought. Relatively recent Putnam asserted the negation of what not so recent Putnam maintained. In particular, he was concerned to argue that at least one logical law (the minimal principle of contradiction) represented an absolutely unrevisable a priori truth. Very recent Putnam (following what he takes to be Wittgenstein's lead) now wishes to claim that the question whether such a principle can be revised or not is one which we are unable to make any clear sense of.39 In the course of outlining his new position, he offers a suggestive and provocative rough sketch of how to tell the history of an important chapter in the development of contemporary philosophical thought. It is in part through his provision of that sketch that Putnam attempts to indicate what his present view is. My aim in the remainder of this paper will be to try to fill in some of the details of this rough sketch-in part in the hope that it will bring into sharper relief the view Putnam is presently after, but mostly because the story that emerges is one which I find myself wanting to tell. I will argue at the end of the paper that this story sheds a helpful light on why the text of the Tractatus assumes the form that it does--one of having the reader climb up a ladder which he is then asked to throwaway.

THE KANTIAN CONCEPTION OF LOGIC Kant's conceptions of reason and freedom-and his conception of the intimacy of these topics-t) is simply that bringing about cat t would raise e's chance at t of occurring (at t'). I think it is clear that bringing about event c may raise the chance of e's subsequent occurrence even though c is part of no robust total cause of e. This will be the case just so long as pee/c) > pee), even if for any independent cause k of e there exists some causal antecedent j of k such that p(e/c&k):t:. p(e/c&j&k). A significant part of our interest in causes is indeed based on the conception of causation presently under consideration, since we are so often interested in finding means appropriate to bringing about desired ends. A slight sophistication of this interest even provides a practical justification for requiring robustness of a cause whenever we can get it. Deciding what to do often involves reasoning in situations of uncertainty where the appropriate decision rule is to choose the action associated with maximum expected utility. This requires knowledge of the probability of various possible outcomes associated with a given action, which is viewed as a cause of whichever outcome in fact obtains (provided it raises the probability of that outcome). Suppose that while event e indeed raises the probability of event c, the probability pee/c) is not invariant under conditionalization on antecedents of

187

Healey, Richard, Chasing Quantum Causes: How Wild Is the Goose? , Philosophical Topics, 20:1 (1992:Spring) p.181

c, and nor is c part of any epistemically accessible total cause T such that p(err) is so invariant. Then in deciding whether to (try to) bring about c in order to maximize expected utility, one could never be in a position correctly to evaluate the chance that c would in fact be followed bye. Suppose, on the other hand, that if one were able to succeed in bringing c about, then it would be part of a robust total potential cause T of e, with known chance p(err) of causing e. Then one could know that p( err) represents the chance of e's occurrence consequent upon bringing about c and use this knowledge in an expected utility calculation to determine whether or not to (try to) bring about c. This provides a practical justification for seeking out epistemically accessible robust total causes whenever possible. But it falls short of warranting belief that every cause is in fact part of such a total cause. Consequently, even this sophistication of the potential means/ends conception of causation fails to provide a convincing justification for a requirement that an event c is a cause of an event e only if c is part of some epistemically accessible robust total cause of e. What then is the status of robustness conditions? Is robustness a necessary condition on any stochastic causal relation? The foregoing survey of attempts to justify the necessity of various robustness conditions would seem to support the following moderately skeptical conclusion. While one can formulate plausible robustness conditions based on R and IR, there is no clear answer as to whether any of these really expresses a necessary condition on any stochastic causal relation. There are conceptions of causation which would tolerate the failure of any and all of these conditions. But there are other conceptions of causation which would be incompatible with the failure of one or more of these conditions. Moreover, I see no clear and convincing reasons for taking anyone of these conceptions of causation as more basic than the others. The concept of causation is just not clear enough to permit a decision as to whether or not a stochastic causal relation must be robust.

III If appeals to robustness do not help to determine the causal relations that obtain between events es' eL , and eR in Aspect-type experiments, perhaps appeal to some other causal principle may be more successful. A natural candidate is Reichenbach's common cause principle which, loosely stated, says that any probabilistic correlation between two events is either the result of a direct causal connection between these events or the result of a common cause in their past. Since eL and eR are probabilistically correlated, es would appear to be an excellent candidate for their common cause. However, some have taken a more precise statement of Reichenbach's

188


principle to place constraints on the probabilistic relations among any common cause and its joint effects.14 One such constraint implies that if es is indeed the common cause of eLand eR, then the outcomes of these events are rendered probabilistically uncorrelated by conditionalizing on e s: i.e., p(e L&eies) p(eL!eS) • p(eR!eS)' where eo e R mark either of the two possible outcomes at L, R respectively. This threatens to rule out es as common cause of eL and eR, since according to quantum mechanics the only relevant feature of es is that it prepares the two "particle" system in the (zero angular momentum) quantum singlet state, but in that state the outcomes of eLand eR are by no means uncorrelated. Moreover, the threat cannot be averted merely by taking it that the quantum mechanical correlations would disappear after further conditional,· ization on some "hidden" feature of es' Not only is it inconsistent with standard interpretations of quantum mechanics to suppose that there is any such feature. But also, to assume that further conditionalization on some "hidden" state of the two particle system produced at es would render the outcomes of eL and eR uncorrelated is tantamount to assuming a condition commonly known in the literature as Jarrett Completeness or Outcome Independence. Notoriously, together with other plausible principles (including a locality principle such as Jarrett Locality or Parameter Independence), this condition is sufficiently powerful to imply Bell inequalities which are violated for certain combinations of detector settings in the quantum singlet state. Should we conclude that since p(e L&eies) i= p(eL/e s) . p(eR/es)' es cannot be a common cause of eL and e R? This would be the correct conclusion to draw if the following factorizability condition indeed expresses a necessary condition on any common cause C of two effects d and e which are not directly causally connected.

=

(F)

A common cause c of events d and e is factorizable just in case p(d&elc) = p(d/c).p(e/c)

It is quite clear that (F) does not express a necessary condition on all common causes of pairs of causally unconnected events. If d and e share two common causes c1 and c Z' then it may be that p(d&e/c1) i= p(dlc1) . p(e!c1) even though p(d&e/c1&cZ) =p(dlc1&cZ).p(e/c1&cZ)' (F) can express a plausible necessary condition only on a total common cause of a pair of causally connected events. I shall not pause to explicate the relevant notion of totality here, but will simply assume for the sake of argument that if in an Aspecttype experiment, es is a common cause of eLand eR, then it is a total common cause of these events. If (F) does indeed express a necessary condition on any total common cause of a pair of correlated but causally unconnected events, then es is not a common cause of eL and e R (assuming these events are not directly causally connected). But how can (F) be justified as a necessary condition even on a total common cause?

189


In a recent paper, Chang and Cartwright have given an answer to this question appropriate to just this case. IS They claim that factorizability is essentially a propagation condition and they support this claim by sketching an argument deriving an instance of factorizability applicable to an Aspecttype experiment from several premises appropriate to a conception of causation as essentially involving continuous propagation of a causal process from cause to subsequent effect. The premises of this argument are as follows. (i) Contiguity Condition: every cause and its effect must be connected by a causal process that is continuous in space and time. (ii) Relativistic Finite-Speed Condition: all processes must propagate at a speed less than or equal to that of light. (iii) Markov Condition: the stages of a causal process must have no memory, so that complete information on temporally intermediate stages makes earlier stages causally irrelevant. (iv) Cause-Correlation Link: all correlations between spatially separated events must be completely explainable by either a direct (spatio-temporally continuous) causal connection between them, or by common events in the past causal history of each. The argument seeks to establish that there is a factorizable common cause of e L and eR-a preceding cause that renders the correlated outcomes of these events probabilistically independent. Suppose that eL and e R are spacelike separated. Then the contiguity and finite-speed conditions imply that they are not causally connected. It therefore follows from the causecorrelation link that there are common events in the past causal history of each event that completely explain the probabilistic correlations between their outcomes. But e s is the only relevant event in the past causal history of both events: indeed, if the "particles" are photons created at es and traveling in opposite directions, then it is the only event in the past causal history of both "particles". By the contiguity condition, any causal influence of es must propagate continuously to each of eL and eR. By the finite speed condition, such propagation must be confined to the forward light-cone of es• Moreover, on each spacelike hyperplane through that light-cone, there must be a stage SL of the process leading to eo and a stage SR of the process leading to eR (see figure 1). Consider one such hyperplane, so chosen that every point on it which lies within the backward light cone of e Lis spacelike separated from eR, and every point on it which lies within the backward light cone of e R is spacelike separated from e c Assuming, in accordance with the finite-speed condition, that SL is confined to the backward light-cone of eL, and similarly for SR and eR, the finite-speed condition further entails that SL is causally irrelevant to e R, and that SR is causally irrelevant to eL. The Markov, contiguity,

190


and finite-speed conditions then entail that eo eR are probabilistically independent, conditional on SL and SR. 16 I shall return to this argument later, since I think it relies on yet another implicit premise (of separability) which my account of Aspect-type experiments rejects. But, modulo this premise, this argument of Cartwright and Chang offers the clearest justification I have seen for taking factorizability as a necessary condition on a putative common cause of eL and e R in an Aspect-type experiment. It is important to stress, however, that this putative common cause is not es' but rather SL and SRo The argument does not yield the conclusion that es is itself a factorizable common cause of eL and eRo Therefore its soundness would not rule out an explanation in terms of a nonJactorizable common cause es' whose effects propagate continuously and subluminally to eL and eRo But note that Cartwright and Chang present their argument not to justify, but rather to undermine, its conclusion by exposing the weakness of the premises on which it rests! They argue that the weak link is the contiguity condition: that taking quantum mechanics seriously involves abandoning the view that causes always operate via continuous processes, and that this renders the Markov condition inapplicable if not actually false. The conclusion they draw is that there is simply no reason to require common causes in quantum mechanics to be factorizable, and so no reason to deny that es is indeed the common cause of eL and eR in an Aspect-type experiment. But one who takes the contiguity condition as essential to causation has a natural response to this line of argument. On this conception, once the contiguity condition is dropped, there is no longer any reason to suppose that the

.'

........./ ..

.......... ..........

SR If-II~:---t---:o'~~

........

",

................

..............-

.....,..

........

... '.....

.x:···...-

..,/ .........

.............. /

........... ,

.'

'.

"

. ...........

....

FIGURE 1

191


stochastic relation between e s and eL or e Ris causal. For what would then be the grounds for supposing that it is? Certainly es temporally precedes eL or eR, and is regularly (probabilistically) associated with the possible outcomes of each of these events. But that cannot be enough to establish causation. After all, if ~ occurs invariantly before eR, then we would have the same pattern of temporal precedence and regular association between the possible outcomes of these events. But surely this would not establish eL as a cause of eR, precisely because of the absence of any causal process capable of linking these events. No more then does regular association and temporal precedence suffice to establish es as a cause of ~ in the absence of any continuous causal process linking the two events. To counter this response, a supporter of Cartwright and Chang needs to do more than argue that there is no reason to deny that, say, es is a cause of eR; she needs to further provide some positive reason to suppose that it is a cause. One such reason is based on a means/ends conception of causation. Producing es would raise the chance of either possible outcome of eRo In the circumstances, if es does not occur, nor will eR: the probability of either possible outcome will then be zero. But if es does occur, then (given the presence of other independent causes) so will eR: the probability of either outcome will then be 112. On the other hand, producing eL has no effect either on the chance of getting some outcome at R, or on the chance of either possible outcome of eRo And even though the possible outcomes of e L are probabilistic ally correlated with the possible outcomes of eR, there seems no way of manipulating the outcome of eL so as to affect the chance that eRwill have one outcome rather than the other. This gives one a reason, based on a means/ends conception of causation, to hold that es is a cause of eR, but that eL is not. But note that a counterfactual version of the means/ends conception may restore parity between es and e Lin this situation. And such a version is independently plausible. For suppose that it is impossible to manipulate a putative cause c of an event e, where that impossibility is more than just practical or technological impossibility. Does it follow that c is not in fact a cause of e? Surely not: it is not just for practical or technological reasons that we are unable to manipulate those events during the first few minutes after the "big bang" we believe to have caused the formation of most of the stable particles of which matter is currently composed. Such considerations support a weakened counterfactual version of a means/ends conception of causation, according to which it is sufficient for an event c to be a cause of an event e that, were it possible to manipulate c, this would affect the chance of eo But on this weakened conception, it is hard to avoid the conclusion that in the situation considered, not only is es a cause of eR, but so also is eLo A proponent of a conception of causation as essentially involving propagation in accordance with the contiguity condition will likely take this last conclusion as a

192


reductio of the weakened means/ends conception of causation on which it is based. He will argue that eL is not a cause of e R unless it is connected to eR by a continuous causal process. But if there is no continuous process connecting es to e R, then for the same reason e s is not a cause of e R either! I see no way to resolve this dispute. The parties to the dispute simply proceed from different conceptions of causation. Once more, our concept of causation is just not sufficiently univocal to permit a decision as to which conception of causation is correct. Suppose that Cartwright and Chang are right that the contiguity condition fails in Aspect-type experiments. Then on one conception es is indeed the common cause of eL and eR in an Aspect-type experiment, even though the factorizability condition (F) fails. But on a rival conception, es is not a common cause of eL and eR just because of the failure of the contiguity condition. One can have a complete physical account of the correlations manifested in an Aspect-type experiment and still not know how to characterize the causal relations between these key events, because of the "open texture" of the concept of causation. And this shows that causation is not a physical relation.

IV While Cartwright and Chang argue in favor of a common cause model of the correlations manifested in Aspect-type experiments, Elby has argued against such a model. I have elsewhere criticized an earlier argument of Elby's which invoked a robustness condition. 17 More recently, Elby has given an argument which makes no appeal to a robustness condition. The new argument depends instead on two causal principles he calls Reichenbach's Requirement (Reich) and Causal Unidirectionality (Uni), which I quote. {c j } Reich

are all the partial causes of e only if p(e/{ cj},w)

=p(e/ (c j}) for all W (besides e and its effects, and besides "intermediate stages" connecting (c) to e).

Uni

An observer cannot (correctly) claim both that a is a partial cause of b and that b is a partial cause of a.

A further principle to which Elby's new argument appeals is EPR Explanatory Symmetry (ES), whose statement I have adapted to conform to my own notation. ES

If our physical description of an Aspect-type experiment is physically symmetric under L-wing H R-wing exchange, then our corresponding explanation must not introduce an asymmetry between the two wings.

The argument proceeds as follows. IS If eL were a partial cause of eR , then by ES, e R would be a partial cause of eL • But this would violate Uni. Hence 193


e L is not a partial cause of e R , and by symmetric reasoning nor is e R a partial cause of e L• The only plausible remaining causal story posits es as sole common cause of eL and e R • But within the framework of quantum theory, the chance of each possible outcome of e L depends only on es's preparing a pair of "particles" in the singlet state, and the outcome of eR ; (and symmetrically with 'L' and 'R' interchanged). Moreover, according to quantum theory, p(eL/eS' e R ) p(eL/eS )' But we have already shown that e R is not an effect of eL' and nor is eR an intermediate stage between es and e L• Hence the fact that p(eL/eS ' e R ) p(eL/eS ) is in violation of Reich. It follows that es cannot in fact be a cause of each of eL and e R• Hence there are no causal relations among es and each of e L and e R • This argument may even be strengthened so that it applies also to a wide class of "hidden variable theories" which supplement the quantum mechanical description of e s' provided these both satisfy the key inequality p(eL/eS' e R ) p(eL/eS ) and give a physical description of an Aspect-type experiment which is physically symmetric under L-wing H R-wing exchange. Does this argument of Elby' s establish that there are no direct causal relations between eLand eR , and that es is not a cause of each of eLand eR ? It does so only if Reich and Uni are justified as necessary conditions on causal relations. Once more I shall argue that the concept of causation is not sufficiently univocal to permit any convincing justification for these conditions. Elby illustrates the plausibility of Reich by means of an everyday example in which the failure of the equality p(e/ {c j}, w) =p(e/ {c j}) prompts the inclusion of w in the set of partial causes of e, thus ensuring conformity to Reich. He also offers the following general defense of Reich. Let {c j} be a set of partial causes of e, and w be a distinct event which is neither an effect of e nor an "intermediate stage" connecting {c j} to e, with p(e/ {c j}, w) p(e/{c j}). Presumably, not only w, but also the members of {cjL affect the probability of e. Why, then, does w, or some relevant event correlated with w, itself not count as a partial cause of e? I agree that apparent failures of Reich frequently motivate a successful search for additional causes, and so it is easy to support the plausibility of Reich by appeal to intuitions based on ordinary cases. But this is not enough to justify it as a necessary condition on causal relations even in an extraordinary case like that presented by the Aspect-type experiments. Elby's general justification apparently rests on a conception of causation according to which (ifw is not an effect of e and) ifp(e/w) > p(e), then either w, or some relevant event correlated with w, is a cause of e. Now typically, p(e/w) > p(e) if wand e are joint effects of a common cause c. If Elby's justification is to succeed in such a case even though p(e/c, w) p(e/c), then w must be correlated with some additional cause d of e such that p(e/c, d, w) = p( e/c, d). But the justification of the assumption that there is such a d cannot be derived from the conception alone, and must be sought elsewhere.

* *

*

*

*

194


Can Reich be given a convincing general justification? It is not necessary to begin a new investigation in order to show why this is unlikely. For Reich is closely related to two principles examined earlier whose general justification was shown to be problematic, namely robustness and factorizability. To see the connection to robustness, suppose that {cl } is a total set of causes of e, but that the robustness condition R fails for this total cause. Then there is a causal antecedent w of (at least one of) the cl ' and a modification w' of w which keeps fixed the {cl }, such that p(eI{cl }, w) p(e/{cl }, w'). If Reich held, then p(e/{ cJ, w) =p(e/{ cJ), and also p(eI{ CI}' w') =p(e/{ cl }). But then we would have p(eI{cl }, w) = p(eI{cl }, w'). Consequently, if the robustness condition does' not hold for every total cause, then Reich fails. Thus Reich implies that the robustness condition R holds for every total cause. 19 But section II pointed out how problematic is the justification of robustness. It can be no easier to justify Reich. Consider now the connection to factorizability. Suppose that {c l } compose a total common cause T of distinct events d and e, which are not directly causally connected to one another. If the factorizability condition F fails, then p(d&eff) * p(dff) . p(eff). It follows that p(eI{ c), d) * p(eI{ cl }). But dis not an effect of e, and nor is d an "intermediate stage" connecting {cl } to e. Hence Reich fails. What this shows is that if Reich holds, then any total common cause of a pair of events which are not directly causally connected to one another satisfies the factorizability condition F. But it was the burden of section III to show that while F is indeed a plausible requirement on a total common cause of such a pair of events, our notion of causation is not sufficiently definite to permit any convincing general justification of this requirement. The prospects for a general justification of Reich can be no better. Consider now whether Uni can be justified as a necessary condition on a causal relation. The problematic nature of Reich may make this seem superfluous, but there is a reason for doing so which will emerge in the final section, where I consider an account of the correlations manifested in Aspect-type experiments involving a causal relation that arguably conforms to Reich and ES, and which might therefore be thought to be ruled out by Elby's argument. At first sight nothing is more clearly part of our concept of causation than that it is an asymmetric, or at least antisymmetric, relation. Moreover, this appears to be an immediate consequence of the generative nature of the causal relation. As Anscombe says, ... causality consists in the derivativeness of an effect from its causes. This is the core, the common feature, of causality in its various kinds. Effects derive from, arise out of, come of, their causes. 20

But if f derives from e, then surely e cannot also derive from f1 A resolute empiricist should not be swayed by the force of this rhetorical question. He

195


will regard talk of generation as simply a metaphor, which may be more appropriate in some circumstances than in others. Typically, it will be appropriate only when a cause clearly precedes its effect. But philosophers have questioned the temporal priority of causes. It has been widely maintained that some causes are simultaneous with their proximate effects, and appeal to the concept of causation has rarely curtailed debate about the possibility of instantaneous action at a distance. That controversy also continues as to whether an effect could precede its cause is an even surer sign that temporal priority is not indisputably part of any reasonable concept of causation. Even if one accepts that causal priority is parasitic on temporal priority, consideration of alternative temporal structures suggests that symmetric temporal relations are not inconceivable. Consider, for example, Godel's model of general relativity. Although it is quite clear that this does not fit our world, it does seem to demonstrate the physical possibility that the (local) time appropriate to motion on an accelerated trajectory should have the structure of a closed but temporally oriented 100p.21 Does our concept of causation prevent us from saying that each of two distinct events involving a particle whose spacetime trajectory corresponds to such a loop would be a cause of the other (and that it would be so classified by any hapless observer in the Godel world)? If not, then we have here a violation of Uni. Even in a spacetime (such as Minkowski spacetime) with no closed timelike curves there may be distinct "frame-dependent" temporal orderings of a pair of events. In such a case, there is no privileged asymmetric temporal relation to provide a basis for an asymmetric generation relation. The empiricist may deny the appropriateness of talk of generation. But it does not follow that he will also deny that there is a direct causal relation between these events. If such a pair is correlated via an appropriately continuous connecting process, he may accept a violation of Uni, rather than deny that these events are directly causally related. In the final section I shall present one significant conception of causation which recommends that one deny Uni for this reason in circumstances which may in fact obtain in Aspect-type experiments.

v I have elsewhere offered an interpretation of quantum mechanics which entails a certain account of the correlations manifested in Aspect-type experiments. 22 I shall sketch this account here for several reasons. The account makes clear by example why it is impossible to draw any firm conclusions about causal relations in the experiments in the absence of a definite theo-

196


retical model. It shows how this particular model highlights an important hidden premise in Cartwright and Chang's argument connecting contiguity to factorizability. And it illustrates a conception of causation which admits symmetric causal relations that conflict with the requirement Uni (while, interestingly enough, conforming to Reich!) But the most important reason for presenting this account is to reinforce the skeptical conclusion, that even given a definite theoretical account of the correlations, no causal conclusions follow. For the concept of causation admits alternative equally valid conceptions, but these dictate conflicting causal conclusions. Let A be the "particle" detected on the left, and B the "particle" detected on the right. Then on my interactive interpretation of quantum mechanics. the pair AB emerges from es with a certain correlational property C (associated with the singlet state) which subsequently gives rise to the observed correlations. I call C a holistic property, since there are no properties of A, B which determine whether or not AB possesses C. The process which leads to the correlations is continuous, since each temporal cross-section reveals a corresponding stage ofthe process. But it is nonseparable, in that there are stages of the process which are not determined by any ascription of properties at spacetime points in the region within which the process occurs. This is a simple consequence of the fact that the process includes stages which involve the ascription to a spatially dispersed system (namely AB) of a holistic property such as C. Suppose first that e R occurs (invariantly) earlier than eL• Only the probability of the outcome of e R is fixed by the prior state of AB: each possible outcome has probability 112, which quantum mechanics predicts (as p(eies». After e R, the relevant properties of A have not changed. It is not the properties of A, but those of AB which explain the probabilistic correlation with the subsequent outcome of eL• One such property is C, which is conserved between es and eL; another is a property F that AB possesses between e R and eL just because B acquired a corresponding property at e R• C and F together fix the probability of each possible outcome of eL: these probabilities therefore depend on the outcome of e R, but the dependence is not consequent upon any alteration in the properties of A itself. The resultant probability is just the conditional probability which quantum mechanics predicts (i.e., p(eL/es' eR»: for almost every setting ofthe two detectors it differs from 112. This is illustrated in figure 2, in which SL represents a stage at which AB has C and F. If e L occurs (invariantly) earlier than e R, then an exactly symmetric account of the correlations is arrived at simply by interchanging L, A and R, B. The interesting case is that illustrated in figure 3, in which eL is spacelike separated from e R, and so these events have no invariant time ordering. In that case, a complete account is given by describing the entire continuous

197


/

...

..,

......

........

.... ........

.'

"""'"

'.

es FIGURE 2

process which links es' eo and eR, Part of this process is described by repeating the story of the previous paragraph for every family of spacelike hyperplanes parametrized by a Lorentz time according to which eRoccurs earlier than eL: as before, SL represents a stage intermediate between eRand eL, The rest of the process is described by repeating the symmetric story for every family of spacelike hyperplanes parametrized by a Lorentz time according to which eLoccurs earlier than eR: SR represents a stage intermediate between eL and eR, Note especially that this description is not redundant, since the two parts of the story are not equivalent: they are not mere interchangeable frame-dependent perspectives on the nonseparable process linking es' eo and eR, Rather, the process cannot be completely described without including both parts. This is a characteristic of such nonseparable processes. Note also that since all probabilities are defined relative to the relevant spacelike hyperplane, one cannot extract contradictory values for (say) p(eR/eS) from different parts of the description. Consider now Chang and Cartwright's argument for factorizability. This purports to derive that condition from the their contiguity, relativistic finite-speed, Markov, and cause-correlation conditions on causal relations. But note that this argument requires that if there is a continuous process linking es to spacelike separated eo eR, then there is a stage SL of this process defined on a region of a spacelike hyperplane confined to the backward light-cone of eo such that every point in that region is spacelike separated from eR, and symmetrically with Land R interchanged (see figure 1). But in the nonseparable model, there are no such stages. Every stage of the process linking e s to eL and eR is nonseparable, and includes properties

198


ascribed not to A but to AB: SL and SR in figs. 2 and 3 are examples of such stages. But even if B does not have a definite trajectory, the region of each spacelike hyperplane it occupies must always overlap the light-cone of e R• It follows that there is no stage of the process at which AB is confined to a region, every point of which is spacelike separated from e R• Consequently, within the nonseparable model, there is no stage of this process defined on a region of a spacelike hyperplane confined to the backward light-cone of eL , such that every point in that region is spacelike separated from e R• Moreover, I see no reason for Chang and Cartwright to deny that the nonseparable model I have sketched satisfies all their explicit premises. The nonseparable model involves a process which is continuous. And Chang and Cartwright, at least, would surely grant that this process is causal, insofar as it connects the acknowledged common cause es to its effects e L and e R• Thus both cause-correlation and contiguity conditions appear to be met. The Markov condition is also met by all stages of the process, provided one defines these stages on the appropriate hyperplanes. Does the nonseparable process really propagate at a speed less than or equal to that of light, as required by the relativistic finite-speed condition? I would argue that it does. Propagation is viewed as occurring continuously from stage to stage defined on a continuous family of spacelike hyperplanes, according to the increasing Lorentz time appropriate to that family. It follows that no stage of the process is spacelike separated from any other stage. Moreover, because the process is nonseparable, the stages do not decompose into spatially localized events: hence no event within one stage is spacelike separated from any event within another stage. This ensures conformity to

.... ......./ /

':>.:. .

....)< .' .' ,."

..........

..... "''''''

.'

.'

....

"'" '

.

'.

........

'

.

.....

FIGURE 3

199


the relativistic finite-speed condition on any reasonable understanding of nonsuperluminal propagation. It is nevertheless interesting to note that this process does indeed connect the localized events eL and eR, which were assumed to be spacelike separated! What causal conclusions can be drawn from the nonseparable model? One conclusion that may seem relatively noncontroversial in the light of the preceding discussion is that es is a common cause of eL and eR. For this conclusion apparently follows both on a means/ends conception and on a (nonseparable) continuous causal process conception of causation. But there are other conceptions which make it more problematic. If causation is just a matter of raising the chance of an event over what it otherwise would have been, then one might deny that es causes eR, on the grounds that the appropriate comparison is with a situation in which the pair AB is prepared not necessarily in the singlet state, but in an arbitrary quantum state. For symmetry considerations alone dictate that on any plausible measure over quantum states, the probability that a counterpart to eR will have outcome e R in an Aspect-type experiment on a pair in an arbitrary quantum state is 112, in which case es has no effect on the chance of eR. In my opinion, once one has shed the prejudice that if es is a common cause it must satisfy factorizability, the intuition that es is a common cause of eL and eRis quite strong. If this intuition is further supported by significant conceptions of causation like the means/ends and continuous causal process conceptions, then this should prompt one to consider modifying an alternative conception which appears to conflict with this intuition so that it conforms to it. And I think it is quite easy so to modify that conception of causation in accordance with which a cause is an event which is accompanied by an increase in the chance of its effect over what it would be in otherwise similar circumstances. I conclude that if the nonseparable model of the correlations manifested in Aspect-type experiments is correct, then es is indeed a common cause of ~ and eR. But note that the argument for this conclusion depended on details of the theoretical model: no causal conclusion could reasonably be drawn from the correlations alone. The more controversial question is whether or not e L and eRare directly causally related. I shall argue that this question cannot be answered even if one accepts the nonseparable model ofthe correlations manifested in Aspecttype experiments. But note first that the answer to this question must be model-dependent, since there is nothing incoherent about a nonlocal hiddenvariable model according to which a separable continuous process directly propagates from the absolutely earlier event (say eL ) to the absolutely later event (sayeR)' in such a way that the earlier event (together with es) causes the later event. Of course, in the spacelike separated case, such a model is in tension, if not outright conflict, with relativity, and for that and other reasons may be considered implausible or even experimentally refuted. But this

200


does not affect the internal coherence of the model. And that is all that is required to establish the point that there is at least one coherent model on which eL and eR are directly causally related, even in (at least some) situations in which they are space like separated. Suppose one favors a conception of causation according to which causal relations arise between events insofar as they are appropriately connected by a continuous, law-governed, process. I have already argued that this is true of events eL and eR on the nonseparable model sketched earlier. But someone might object that these events are not in fact connected in the appropriate way: that while that process may be thought to generate eL and eR from es' eL cannot be said to generate eR, and nor can eR be said to generate eL° Now in the case in which eL and eR are timelike separated, I see no force to this objection. The later of these events may be thought to be generated from the earlier by the succeeding stages of the process that connects them. On this conception, there is no basis for allowing that es is a cause of event eL (say), while eL is not a cause of the later event eRo But the case in which eL and eR are spacelike separated is more problematic. Here the temporal symmetry excludes an asymmetric account according to which one of these events generates the other via successive stages of the process that connects them. If eL and eR are causally related at all on the nonseparable model, it seems that they must be symmetrically causally related: but surely it is incoherent to suppose that each generates the other through the process that connects them? While I agree that this is an incoherent idea, I do not think it undermines the conclusion that spacelike separated eL and eR are directly causally related on this model. For the notion of one event generating another imports a metaphysics which is quite foreign to the pure conception of causation as corresponding to connection by a continuous, law-governed, process. The only residual content to the notion of generation within this conception is that the causally related events be such that, on some time ordering, there are stages of the process which continuously connect earlier cause event to later effect event. This is sufficient to rule out the invariantly later of two timelike separated events in Minkowski spacetime as cause of the invariantly earlier. But if a spacetime admits a "circular" time order (like the proper time along a closed timelike curve in Godel spacetime) there may be stages of a process which continuously connect a cause event e to an effect event f which is earlier than e, while still respecting this residual content. This residual content is also respected if one grants that, in the nonseparable model for spacelike separated eL and eR, there is a sense in which eL is a cause of eR, and eR is a cause of eLo To explain this sense, we first introduce an asymmetric but frame-dependent relation Chkf (read as "h is a partial cause of k relative to frame f'). For any frame f on whose Lorentz time

201


e Lprecedes e R, there is a continuous law-governed sequence of stages of the nonseparable process which, in f, successively link e L to eRe On the present conception of causation this amounts to saying that eL is a partial cause of eR relative to frame f. Symmetrically, eR is a partial cause of e L relative to frame r, on whose Lorentz time eR precedes e L. We may now use this frame-dependent notion to define two distinct frame-independent notions of partial cause, as follows: h is a weak partial cause of k (Whk) if and only if h is a partial cause of k relative to some frame; h is a strong partial cause of k (Chk) if and only if h is a partial cause of k relative to every frame. Finally we may say that h, k are symmetrically causally related (Shk) if and only if Whk & Wkh. It is SeL eRwhich expresses the sense in which eL is a partial cause of eR, and eRis a partial cause of eL • Of course, since in the spacelike separated case we have not-CeLeJhk and not-CeRe LIkh there is another sense in which neither of eo e R is an (asymmetric) cause of the other!23 Here we have the promised exhibition of a conception of causation which admits a symmetric causal relation (S), which is defined in terms of a relation of (weak) partial cause (W) that violates Uni while remaining consistent with Reich. But note that it is only the weak partial cause relation which violates Uni. The strong partial cause notion vacuously conforms to Uni. More suprisingly, the frame-dependent notion of partial cause also conforms to Uni! This is surprising since the frame-dependent notion also respects Reich and ES, while Elby has argued that no causal explanation can be given consistently with Uni, Reich, and ES. This points to a lacuna in Elby's argument. The argument assumes that "our" physical description of an Aspect-type experiment is physically symmetric under L-wing H R-wing exchange. But while this is true for a frame-independent physical description, it is not true for a frame-dependent physical description. A causal explanation of eRcan be given in frame f on whose Lorentz time eLprecedes e R in terms of the asymmetric frame-dependent causal relation CeLeRf just because the description of the experiment in frame f is not physically symmetric. I wish to make it quite clear that I do not endorse a conception of causation according to which causal relations arise between events insofar as they are appropriately connected by a continuous, law-governed, process as the uniquely correct conception of causation. On the contrary, other conceptions of causation are equally worthy of acceptance. One may well adopt a means/ends conception, for example. It is not hard to argue that on such a conception, eL and e R are not directly causally related, given the nonseparable model. For on that model, no possible manipulation of either of these events would have any effect on the chance of the other. To indirectly affect the chance of one event it would be necessary to directly influence (the chance of) the other: but that cannot be done, consistent with the constraints

202


imposed by the nonseparable model. My conclusion is rather that no conception of causation is uniquely appropriate to this case, and so even within the nonseparable model of the correlations manifested in Aspect-type experiments there is no determinate answer to the question as to whether or not eL and eR are directly causally related.

NOTES I. I wish to thank Jordi Cat, Andrew Elby, and Martin Jones for helpful correspondence, as well as the members of an fnformal History and Philosophy of Science Colloquium at the' University of Arizona. An early version of section II was presented at the British Soci.ety for Philosophy of Science annual meeting in September 1991. 2. See "The Logic of Quantum Mechanics" and "Time and Physical Geometry" (205) in his Philosophical Papers, Vol. I (Cambridge: Cambridge University Press,1974). 3. The literature on this violation is enormous. There is no better introduction than chapter 4 of Redhead's book Incompleteness, Nonlocality and Realism: a Prolegomenon to the Philosophy oj Quantum Mechanics (Oxford: Clarendon Press, 1987). 4. Midwest Studies in Philosophy IX, French, Uehling, and Wettstein, eds. (Minneapolis: University of Minnesota Press, 1984), 3-16. 5. The bibliography to Redhead( 1987) gives detailed references to several of their published papers. 6. By van Fraassen, for one. See, for example, his Quantum Mechanics: an Empiricist View (Oxford: Oxford University Press, 1991). 7. This is the view expressed by Cartwright in her Nature's Capacities and Their Measurement (Oxford: Oxford University Press, 1989). 8. See R. Healey, The Philosophy oj Quantum Mechanics: An Interactive Interpretation (Cambridge: Cambridge University Press, 1989). 9. N. Cartwright and M. Jones, "How to Hunt Quantum Causes," Erkenntnis 35 (1991): 205-31. 10. The later argument appears in his "Nonfactorizability, Stochastic Causality and Passionat-a-Distance," in J. T. Cushing and Ernan McMullin, eds. Philosophical Consequences oJQuantum Theory (Notre Dame, Ind.: University of Notre Dame Press, 1989), 145-53. II. This is Salmon's view: see his Scientific Explanation and the Causal Structure oj the World (Princeton: Princeton University Press, 1984). 12. Or, relativistically, on some spacelike hyperplane within the backward light cone of e. 13. On page vi of the preface to the 1989 edition of Redhead (1987). 14. Reichenbach himself gives such a statement on page 163 of his book The Direction oj Time (Berkeley: University of California Press, 1956): "If coincidences of two events A and B occur more frequently than would correspond to their independent occurrence, that is if the events satisfy [ p(A&B) > p(A).p(B) ], then there exists a common cause C for these events such that the fork ACB is conjunctive, that is, satisfies [the following] relations p(A&B/C) =p(AlC). p(B/C), p(A&B/-C) =p(AI-C). p(B/-C) p(AlC) > p(AI-C) , p(B/C) > p(B/-C)." (In stating these relations I have modified Reichenbach's notation to conform to that used in the rest of this paper.) 15. H. Chang and N. Cartwright, "Causality and Realism in the EPR Experiment," forthcoming in Erkenntnis.

203


16. We have P(eLI SL)

= P(~ I SL & e,) (Markov)

=P(eLI SL & e, & SR) (CRFS) = P(eLI SL & SR) (Markov) (Where (CRFS) abbreviates the contiguity and relativistic finite-speed conditions).

And P(e L I SL & e,)

= P(eLI SL & e, & SR & e R) (CRFS) = P(eLI SL & SR & e R)

(Markov)

It follows that P(~I SL & SR) =P(eLI e R &SL & SR) i.e., eL, e R are probabilistically independent, conditional on SL & SR' 17. This earlier argument is contained in Elby's paper "Should we Explain the EPR Correlations Causally?" Philosophy of Science, 59 (1992). My criticisms appear in "Discussion Note: Causation, Robustness and EPR," also Philosophy of Science 59 (1992).

18. See Elby, "A New Reason not to Explain the EPR Correlations Causally," unpublished manuscript. 19. Essentially this argument appears in Elby (1992). Elby there credits it to Frank Arntzenius. 20. G. E. M. Anscombe, Causality and Determination (Cambridge: Cambridge University Press, 1971). 21. See the discussion in D. Malament, '''Time Travel' in the GOOel Universe," PSA 1984, Volume Two, eds. P. D. Asquith and P. Kitcher (East Lansing, Mich.: Philosophy of Science Association, 1984),91-100. 22. In Healey (1989). 23. It is interesting to note that while the conception of causation presently under consideration licenses the conclusion that e L, ~ are directly causally related, it permits one to say either that each is a partial cause of the other, or that neither is a partial cause of the other. This is yet a further indeterminacy in the concept of causation. Note also that while the strong partial cause relation C conforms to Uni, it violates Reich, and also the factorizability condition F. In the light of Chang and Cartwright's argument it is ironic that a conception of causation based on continuous, temporally ordered, mediating processes should give rise to the idea of a common cause that violates factorizability!

204

Chomsky, Noam, Explaining Language Use , Philosophical Topics, 20:1 (1992:Spring) p.205


Explaining Language Use

Noam Chomsky Massachusetts Institute of Technology

In his John Locke lectures, Hilary Putnam argues "that certain human abilities-language speaking is the paradigm example-may not be theoretically explicable in isolation," apart from a full model of "human functional organization," which "may well be unintelligible to humans when stated in any detail." The problem is that "we are not, realistically, going to get a detailed explanatory model for the natural kind 'human being' ," not because of "mere complexity" but because "we are partially opaque to ourselves, in the sense of not having the ability to understand one another as we understand hydrogen atoms." This is a "constitutive fact" about "human beings in the present period," though perhaps not in a few hundred years.! The "natural kinds" human being and hydrogen atom thus call for different kinds of inquiry, one leading to "detailed explanatory models," the other not, at least for now. The first category is scientific inquiry, in which we seek intelligible explanatory theories and look forward to eventual integration with the core natural sciences; call this mode of inquiry "naturalistic," focusing on the character of work and reasonable goals, in abstraction from actual achievement. Beyond its scope, there are issues of the scale of full "human functional organization," not a serious topic for (current) naturalistic inquiry but more like the study of everything, like attempts to answer such pseudo-questions as "how do things work?" or "why do they happen?"

205


Many questions, including those of greatest human significance one might argue, do not fall within naturalistic inquiry; we approach them in other ways. As Putnam stresses, the distinctions are not sharp, but they are useful nonetheless. In a critical discussion of "sophisticated mentalism of the MIT variety" (specifically, Jerry Fodor's "language of thought"), Putnam adds some complementary observations on theoretical inquiry that would not help to explain language speaking. He considers the possibility that the brain sciences might discover that when we "think the word cat" (or a Thai speaker thinks the equivalent), a configuration C is formed in the brain. "This is fascinating if true," he concludes, perhaps a significant contribution to psychology and the brain sciences, "but what is its relevance to a discussion of the meaning of cat" (or of the Thai equivalent, or of C)?-the implication being that there is no relevance. 2 We thus have two related theses. First, "language speaking" and other human abilities do not currently fall within naturalistic inquiry. Second, nothing could be learned about meaning (hence about a fundamental aspect of language speaking) from the study of configurations and processes of the brain (at least of the kind illustrated). The first conclusion seems to me understated and not quite properly formulated; the second, too strong. Let's consider them in tum. The concept human being is part of our common-sense understanding, with properties of individuation, psychic persistence, and so on, reflecting particular human concerns, attitudes, and perspectives. The same is true of language speaking. Apart from improbable accident, such concepts will not fall within explanatory theories of the naturalistic variety; not just now, but ever; not because of cultural or even intrinsically human limitations (though these surely exist), but because of their nature. We may have a good deal to say about people, so conceived, even low-level accounts that provide weak explanation. But such accounts cannot be integrated into the natural sciences alongside of explanatory models for hydrogen atoms, cells, or other entities that we posit in seeking a coherent and intelligible explanatory model of the naturalistic variety. There is no reason to suppose that there is a "natural kind 'human being' ," at least if natural kinds are the kinds of nature, the categories discovered in naturalistic inquiry. The question is not whether the concepts of common-sense understanding can themselves be studied in some branch of naturalistic inquiry; perhaps they can. Rather, it is whether in studying the natural world (for that matter, in studying these concepts, as part of the natural world), we view it from the standpoint provided by such concepts. Surely not. There may be scientific studies of some aspects of what people are and do, but they will not use the common-sense notions human being or language speaking, with their special role in human life and thought, in formulating their explanatory principles.

206


The same is true of common-sense concepts generally. Such notions as desk or book or house, let alone more "abstract" ones, are not appropriate for naturalistic inquiry. Whether something is properly described as a desk, rather than a table or a hard bed, depends on its designer's intentions and the ways we and others (intend to) use it, among other factors. Books are concrete objects. We can refer to them as such ("the book weighs five pounds"), or from an abstract perspective ("who wrote the book?"; "he wrote the book in his head, but then forgot about it"), or from both perspectives simultaneously ("the book he wrote weighed five pounds," "the book he is writing will weigh at least five pounds if it is ever published"). If I say "that deck of cards, which is missing a Queen, is too worn to use," that deck of cards is simultaneously taken to be a defective set and a strange sort of scattered "concrete object," surely not a mereological sum. The term house is used to refer to concrete objects, but from the standpoint of special human interests and goals and with curious properties. A house can be destroyed and rebuilt, like a city; London could be completely destroyed and rebuilt up the Thames in 1000 years and still be London, under some circumstances. It is hard to imagine how these could be fit concepts for theoretical study of things, events, and processes in the natural world. Uncontroversially, the same is true of matter, motion, energy, work, liquid, and other common-sense notions that are abandoned as naturalistic inquiry proceeds; a physicist asking whether a pile of sand is a solid, liquid, or gas, or some other kind of substance, spends no time asking how the terms are used in ordinary discourse, and would not expect the answer to the latter question to have anything to do with natural kinds, ifthese are the kinds in nature. 3 It is only reasonable to expect that the same will be true of belief, desire, meaning and sound of words, intent, etc., insofar as aspects of human thought and action can be addressed within naturalistic inquiry. To be an Intentional Realist, it would seem, is about as reasonable as being a Desk- or Sound-ofLanguage- or Cat- or Matter-Realist; not that there are no such things as desks, etc., but that in the domain where questions of realism arise in a serious way, in the context of the search for laws of nature, objects are not conceived from the peculiar perspectives provided by concepts of common sense. It is widely held that "mentalistic talk and mental entities [should] eventually lose their place in our attempts to describe an~ explain the world."4 True enough, but it is hard to see the significance of the doctrine, since the same holds true, uncontroversially, for "physicalistic talk and physical entities" (to whatever extent the "mental"/"physical" distinction is intelligible). Even the most elementary notions, such as nameable thing, crucially involve such intricate notions as human agency. What we take as objects, how we refer to them and describe them, and the array of properties with which we invest them, depend on their place in a matrix of human actions, interests, and intent in respects that lie far outside the potential range of naturalistic

207


inquiry. The tenns of language may also indicate positions in belief systems, which enrich further the perspectives these tenns afford for viewing the world, though in ways inappropriate to the ends of naturalistic inquiry. Some tenns, particularly those lacking internal relational structure (notably, socalled "natural-kind tenns"), may do little more than that, as far as the natural language lexicon is concerned. s The concepts of natural language, and common sense generally, are not even candidates for naturalistic theories. Putnam extends his conclusions to Brentano's thesis that "intentionality won't be reduced and won't go away": "there is no scientifically describable property that all cases of any particular intentional phenomenon have in common" (say, thinking about cats).6 More generally, intentional phenomena relate to people and what they do as viewed from the standpoint of human interests and unreflective thought, and thus will not (so viewed) fall within naturalistic theory, which seeks to set such factors aside. Like falling bodies, or the heavens, or liquids, a "particular intentional phenomenon" may be associated with some amorphous region in a highly intricate and shifting space of human interests and concerns. But these are not appropriate concepts for naturalistic inquiry. We may speculate that certain components of the mind (call them the "science-forming faculty," to dignify ignorance with a title) enter into naturalistic inquiry, much as the language faculty (about which we know a fair amount) enters into the acquisition and use of language. The products of the science-forming faculty are fragments of theoretical understanding, naturalistic theories of varying degrees of power and plausibility involving concepts constructed and assigned meaning in a considered and determinate fashion, as far as possible, with the intent of sharpening or otherwise modifying them as more comes to be understood. Other faculties of the mind yield the concepts of common-sense understanding, which enter into natural language semantics and belief systems. These simply "grow in the mind," much in the way that the embryo grows into a person. How sharp the distinctions may be is an open question, but they appear to be real nevertheless. Sometimes there is a resemblance between concepts that arise in these different ways; possibly naturalistic inquiry might construct some counterpart to the common-sense notion human being, as H 20 has a rough correspondence to water (though earth, air, and fire, on a par with water for the ancients, lack such counterparts). It is a commonplace that any similarities to commonsense notions are of no consequence for science. It is, for example, no requirement for biochemistry to determine at what point in the transition from simple gases to bacteria we find the "essence of life," and if some such categorization were imposed, the correspondence to some common-sense notion would matter no more than for (topological) neighborhood, energy, or fish. Similarly, it is no concern of the psychology-biology of organisms to deal with such technical notions of philosophical discourse as perceptual

208


content, with its stipulated properties (sometimes dubiously attributed to "folk psychology," a construct that appears to derive in part from parochial cultural conventions and traditions of academic discourse). Nor must these inquiries assign a special status to veridical perception under "normal" conditions. Thus in the study of determination of structure from motion, it is immaterial whether the external event is successive arrays of flashes on a tachistoscope that yield the visual experience of a cube rotating in space, or an actual rotating cube, or stimulation of the retina, or optic nerve, or visual cortex; in any case, "the computational investigation concerns the nature of the internal representations used by the visual system and the processes by which they are derived" (Stephen Ullman), as does the study of algorithms and mechanisms in this and other work along lines pioneered by David Marr. It is also immaterial whether people might accept the nonveridical cases as "seeing a cube" (taking "seeing" to be having an experience, whether "as if' or veridical); or whether concerns of philosophical theories of intentional attribution are addressed. A "psychology" dealing with the latter concerns would doubtless not be individualistic, as Martin Davies argues, but it would also depart from naturalistic inquiry into the nature of organisms, and possibly from authentic folk psychology as welI.7 To take another standard example, on the (rather implausible) assumption that a naturalistic approach to, say, jealousy were feasible, it is hardly likely that it would distinguish between states involving real or imagined objects. If "cognitive science" is taken to be concerned with intentional attribution, it may tum out to be an interesting pursuit (as literature is), but it is not likely to provide explanatory theory or to be integrated into the natural sciences. As understanding progresses and concepts are sharpened, the course of naturalistic inquiry tends towards theories in which terms are divested of distorting residues of common-sense understanding and are assigned a relation to posited entities and a place in a matrix of principles: real number, electron, and so on. The divergence from natural language is twofold: the constructed terms abstract from the intricate properties of natural language expressions; they are assigned semantic properties that may well not hold for natural language, such as reference (we must beware of what Strawson once called "the myth of the logically proper name," in natural language, and related myths concerning indexicals and pronouns).8 As this course is pursued, the divergence from natural language increases; and with it, the divergence between the ways we understand hydrogen atom, on the one hand, and human being (desk, liquid, heavens,fall, chase, London, this, etc.), on the other. But even a strengthened version of Putnam's first thesis does not entitle us to move on to the second, more generally, to conclude that naturalistic theories of the brain are of no relevance to understanding what people do. Under certain conditions, people see tachistoscopic presentations as a rotating cube or light moving in a straight line. A study of the visual cortex might

209


provide understanding of why this happens, or why perception proceeds as it does in ordinary circumstances. And comparable inquiries might have a good deal to say about "language speaking" and other human activities. Take Putnam's case: the discovery that thinking of cats evokes C. Surely such a discovery might have some relevance to inquiry into what Peter means (or refers to, or thinks about) when he uses the term cat, hence to "a discussion of the meaning of cat." For example, there has been a debate, in which Putnam has taken part, about the referential properties of cat if cats were found to be robots controlled from Mars. Suppose that after Peter comes to believe this, his brain does, or does not, form C when he refers to cats (thinks about them, etc.). That might be relevant to the debate. Or, take a realistic case. Recent studies of electrical activity of the brain (eventrelated potentials, ERPs) show distinctive responses to nondeviant and deviant expressions and, among the latter, to violations of (1) word meaning expectancies, (2) phrase structure rules, (3) the specificity-of-reference condition on extraction of operators, (4) locality conditions on movement.9 Such results surely might be relevant to the study of the use of language, in particular, the study of meaning. We can proceed further. Patterns of electrical activity of the brain correlate with the five categories of structure noted: nondeviance, and four types of deviance. But the study of these categories is also a study of the brain, its states and properties, just as study of algorithms involved in seeing a straight line or in doing long division is a study of the brain. Like other complex systems, the brain can be studied at various levels: atoms, cells, cell assemblies, neural networks, computational-representational (C-R) systems, etc. The ERP study relates two such levels: electrical activity of the brain and C-R systems. The study of each level is naturalistic both in the character of the work and in that integration with the core natural sciences is a prospect that can be reasonably entertained. In the context of Putnam's discussion, discoveries about the brain at these levels of inquiry are on a par with a discovery about the (imagined) configuration C, when Peter thinks of cats. In the case of language, the C-R theories have much stronger empirical support than anything available at other levels, and are far superior in explanatory power; they fall within the natural sciences to an extent that inquiry into "language speaking" at the other levels does not. In fact, the current significance of the ERP studies lies primarily in their correlations with the much richer and better-grounded C-R theories. Within the latter, the five categories have a place and, accordingly, a wide range of indirect empirical support; in isolation from C-R theories, the ERP observations are just curiosities, lacking a theoretical matrix. Similarly, the discovery that C correlates with use of cat would, as an isolated fact, be more of a discovery about C than about the meaning of cat-and for that reason alone would shed little light on the controversy about robots controlled from Mars. To take

210


another case, the discovery of perceptual displacement of clicks to phrase boundaries is, for now, more of a discovery about the validity of the experiment than about phrase boundaries. The reason is that evidence of other sorts about phrase boundaries (sometimes called "linguistic" rather than "psychological" evidence, a highly misleading terminology) is considerably more compelling and embedded in a much richer explanatory structure. If click experiments were found to be sufficiently reliable in identifying the entities postulated in C-R theories, and if their theoretical framework were deepened, one might rely on them in cases where "linguistic evidence" is indecisive~ possibly more, as inquiry progresses.1O For the present, the best-grounded naturalistic theories of language and its use are C-R theories. We assume, essentially on faith, that there is SOUle kind of description in tenns of atoms and molecules, though without expecting operative principles and structures of language and thought to be discernible at these levels. With a larger leap of faith, we tend to assume that there is an account in neurological tenns (rather than, say, glial or vascular tenns, though a look at the brain reveals glial cells and blood as well as neurons. ll ) It may well be that the relevant elements and principles of brain structure have yet to be discovered. Perhaps C-R theories will provide guidelines for the search for such mechanisms, much as nineteenth-century chemistry provided crucial empirical conditions for radical revision of fundamental physics. The common slogan that "the mental is the neurophysiological at a higher level"-where C-R theories are placed within "the mental"-has matters backwards. It should be rephrased as the speculation that the neurophysiological may tum out to be "the mental at a lower level"-i.e., the speculation that neurophysiology might, some day, prove to have some bearing on the "mental phenomena" dealt with in C-R theories. As for the further claims of eliminative materialism, the doctrine remains a mystery until some account is given of the nature of "the material"~ and given that account, some reason why one should take it seriously or care if successful theories lie beyond its stipulated bounds. For the present, C-R approaches provide the best-grounded and richest naturalistic account of basic aspects of language use. Within these theories, there is a fundamental concept that bears resemblance to the common-sense notion "language": the generative procedure that fonns structural descriptions (SDs), each a complex of phonetic, semantic, and structural properties. Call this procedure an I-language, a tenn chosen to indicate that this conception of language is internal, individual, and intensional (so that distinct I-languages might, in principle, generate the same set of SDs, though the highly restrictive innate properties of the language faculty may well leave this possibility unrealized). We may take the linguistic expressions of a given I-language to be the SDs generated by it. A linguistic expression, then, is a complex of phonetic, semantic, and other properties. To have an

211


I-language is something like having a "way to speak and understand," one traditional picture of what a language is. There is reason to believe that the I-languages ("grammatical competence") are distinct from conceptual organization and "pragmatic competence," and that these systems can be selectively impaired and developmentally dissociated. 12 The I-language specifies the form and meaning of such lexical elements as desk, work, fall, insofar as these are determined by the language faculty itself. Similarly, it should account for properties of more complex expressions: for example, the fact that "John rudely departed" may mean either that he departed in a rude manner or that it was rude of him to depart, and that in either case, he departed (perhaps an event semantics should be postulated as a level of representation to deal with such facts.l3) And it should explain the fact that the understood subject of expect in (I) depends on whether X is null or is Bill, with a variety of other semantic consequences: (1) John is too clever to expect anyone to talk to X.

And for the fact that in my speech, ladder rhymes with matter but madder doesn't. In a wide range of such cases, nontrivial accounts are forthcoming. The study of C-R systems provides no little insight into how people articulate their thoughts and interpret what they hear, though of course it is as little-and as much-a study of these actions as the physiology and psychology of vision are studies of humans seeing objects. A deeper inquiry into I-languages will seek to account for the fact that Peter has the I-language Lp while Juan has the I-language LJ-these statements being high-level abstractions, because in reality what Peter and Juan have in their heads is about as interesting for naturalistic inquiry as the course of a feather on a windy day. The basic explanation must lie in the properties of the language faculty of the brain. To a good approximation, the genetically determined initial state of the language faculty is the same for Peter, Juan, and other humans. It permits only a restricted variety of I-languages to develop under the triggering and shaping effect of experience. In the light of current understanding, it is not implausible to speculate that the initial state determines the computational system of language uniquely, along with a highly structured range of lexical possibilities and some options among "grammatical elements" that lack substantive content. Beyond these possibilities, variation of I-languages may reduce to Saussurean arbitrariness (an association of concepts with abstract representations of sound) and parts of the sound system, relatively accessible and hence "learnable" (to use a term with misleading connotations). Small differences in an intricate system may, of course, yield large phenomenal differences, but a rational Martian scientist studying humans might not find the difference between English and Navajo very impressive. The I-language is a (narrowly described) property of the brain, a relatively stable element of transitory states of the language faculty. Each

212


linguistic expression (SD) generated by the I-language includes instructions for performance systems in which the I-language is embedded. It is only by virtue of its integration into such performance systems that this brain state qualifies as a language. Some other organism might, in principle, have the same I-language (brain state) as Peter, but embedded in performance systems that use it for locomotion. We are studying a real object, the language faculty of the brain, which has assumed the form of a full I-language and is integrated into performance systems that play a role in articulation, interpretation, expression of beliefs and desires, referring, telling stories, and so on. For such reasons, the topic is the study of human language. The performance systems appear to fall into two general types: articulatory-perceptual and conceptual-intentional. 14 If so, it is reasonable to suppose that a generated expression includes two interface levels, one providing information and instructions for the articulatory-perceptual systems, the other for the conceptual-intentional systems. One interface is generally assumed to be phonetic representation (phonetic form, PF). The nature of the other is more controversial; call it LF ("logical form"). The properties of these systems, or their existence, are matters of empirical fact. One should not be misled by unintended connotations of such terms as "logical form" and "representation," drawn from technical usage in different kinds of inquiry. Similarly, though there is a hint of the notions "deep" and "surface grammar" of philosophical analysis, the concepts do not closely match. What is "surface" from the point of view of I-language is, if anything, PF, the interface with articulatory-perceptual systems. Everything else is "deep." The surface grammar of philosophical analysis has no particular status in the empirical study of language; it is something like phenomenal jUdgment, mediated by schooling, traditional authorities and conventions, cultural artifacts, and so on. Similar questions arise with regard to what is termed, much too casually, "folk psychology," as noted. One should regard such notions with caution: much may be concealed behind apparent phenomenal clarity. The complex of I-language and performance systems enters into human action. It is an appropriate subject matter for naturalistic theories, which might carry us far towards understanding how and why people do what they do, though always falling short of a full account, just as, a naturalistic theory of the body would fail to capture fully such human actions or achievements as seeing a tree or taking a walk. Correspondingly, it would be misleading, or worse, to say that some part of the brain or an abstract model of it (e.g., a neural net or programmed computer) sees a tree or figures out square roots. People in an ambiguous range of standard circumstances pronounce words, refer to cats, speak their thoughts, understand what others say, play chess, or whatever; their brains don't and computer programs don't-though study of brains, possibly with

213


abstract modeling of some of their properties, might well provide insight into what people are doing in such cases. An algorithm constructed in a C-R theory might provide a correct account of what is happening in the brain when Peter sees a straight line or does long division or "understands Chinese,"15 and might be fully integrated into a well-grounded theory at some other level of explanation (say, cells). But the algorithm, or a machine implementing it, would not be carrying out these actions, though we might decide to modify existing usage, as when we say that airplanes fly and submarines set sail (but do not swim). Nothing of substance is at stake. Similarly, while it may be that people carry out the action by virtue of the fact that their brains implement the algorithm, the same people would not be carrying out the action if they were mechanically implementing the instructions, in the manner of a machine (or of their brains). It may be that I see a straight line (do long division, understand English, etc.) by virtue ofthe fact that my brain implements a certain algorithm; but if I, the person, carry out the instructions mechanically, mapping some symbolic representation of the input to a representation of the output, neither I nor I-plus-algorithm-plus-extemal memory sees a straight line (etc.), again, for uninteresting reasons. 16 It would also be a mistake, in considering the nature of performance systems, to move at once to a vacuous "study of everything." As a case in point, consider Donald Davidson's discussion of Peter as an "interpreter," trying to figure out what Tom has in mind when he speaks. Davidson observes that Peter may well use any information, background assumption, guesswork, or whatever, constructing a "passing theory" for the occasion. Consideration of an "interpreter" thus carries us to full models of human functional organization. Davidson concludes that there is no use for "the concept of a language" serving as a "portable interpreting machine set to grind out the meaning of an arbitrary utterance"; we are led to "abandon ... not only the ordinary notion of a language, but we have erased the boundary between knowing a language and knowing our way around in the world generally." Since "there are no rules for arriving at passing theories," we "must give up the idea of a clearly defined shared structure which language-users acquire and then apply to cases." "There is no such thing as a language," a recent study of Davidson's philosophy opens, with his approval. 17 The initial observation about "passing theories" is correct, but the conclusions do not follow. A reasonable response to the observation, if our goal is to understand what humans are and what they do, is to try to isolate coherent systems that are amenable to naturalistic inquiry and that interact to yield some aspects of the full complexity. If we follow this course, we are led to the conjecture that there is a generative procedure that "grinds out" linguistic expressions with their interface properties, and performance systems that access these instructions and are used for interpreting and expressing one's thoughts.

214


What about "the idea of a clearly defined shared structure which language-users acquire and then apply to cases"? Must we also postulate such "shared structures," in addition to I-language and perfonnance systems? It is often argued that such notions as common "public language" or "public meanings" are required to explain the possibility of communication or of "a common treasure of thoughts," in Frege' s sense. Thus if Peter and Mary do not have a "shared language," with "shared meanings" and "shared reference," then how can Peter understand what Mary says? (Interestingly, no one draws the analogous conclusion about "public pronunciation.") One recent study holds that linguists can adopt an I-language perspective only "at the cost of denying that the basic function of natural languages is to mediate communication between its speakers," including the problem of "communication between time slices of an idiolect" (so-called "incremental learning"). 18 But these views are not well-founded. Successful communication between Peter and Mary does not entail the existence of shared meanings or shared pronunciations in a public language (or a common treasure of thoughts or articulations of them), any more than physical resemblance between Peter and Mary entails the existence of a public fonn that they share. As for the idea that "the basic function of natural languages is to mediate communication," it is unclear what sense can be given to an absolute notion of "basic function" for any biological system; and if this problem can be overcome, we may ask why "communication" is the "basic function." And the transition problem seems no more mysterious than the problem of how Peter can be the person he is, given the stages through which he has passed. Not only is the I-language perspective appropriate to the problems at hand, but it is not easy to imagine a coherent alternative. It may be that when he listens to Mary speak, Peter proceeds by assuming that she is identical to him, modulo M, some array of modifications that he must work out. Sometimes the task is easy, sometimes hard, sometimes hopeless. To work out M, Peter will use any artifice available to him, though much of the process is doubtless automatic and unreflective. 19 Having settled on M, Peter will, similarly, use any artifice to construct a "passing theory"even if M is null. Insofar as Peter succeeds in these tasks, he understands what Mary says as being what he means by his comparl;tble expression. The only (virtually) "shared structure" is the initial state ofthe language faculty. Discussion of language and language use regularly introduces other kinds of shared structure: communities with their languages, common languages across a broader culture, etc. Such practices are standard in ordinary casual discourse as well. Thus we say that Peter and Tom speak the same language, but Juan speaks a different one. Similarly, we say that Boston is near New York, but not near London, or that Peter and Tom look alike, but neither looks like John. Or, we might reject any of these assertions. There is

215


no right or wrong choice in abstraction from interests that may vary in every imaginable way. There are also no natural categories, no idealizations. In these respects, speaking the same language is on a par with being-near or looking-like. A standard remark in an undergraduate linguistics course is Max Weinreich's quip that a language is a dialect with an army and navy, and the next lecture explains that dialects are also nonlinguistic notions, which can be set up one way or another, depending on particular interests and concerns. Such factors as conquests, natural barriers (oceans, mountains), national TV, etc., may induce illusions on this matter, but no notion of "common language" has been formulated in any useful or coherent way, nor do the prospects seem hopeful. Any approach to the study of language or meaning that relies on such notions is highly suspect. Suppose, for example, that "following a rule" is analyzed in terms of communities: Jones follows a rule if he conforms to the practice or norms of the community. If the "community" is homogeneous, reference to it contributes nothing (the notions norm, practice, convention, etc., raise further questions). If the "community" is heterogenous, apart from the even greater unclarity of the notion of norms (practice, etc.) for this case, several problems arise. One is that the proposed analysis is descriptively inaccurate. Typically, we attribute rule following in the case of notable lack of conformity to prescriptive practice or alleged norms. Thus we might say that Johnny, who is three, is following his own rule when he says brang instead of brought; or that his father Peter is following the "wrong rule" ("violating the rules") when he uses disinterested to mean uninterested (as most people do). But only a linguist would say that Johnny and Peter are observing Condition (B) of the Binding theory, as does the "community" generally (in fact, the community of all language speakers, very likely). The more serious objection is that the notion of "community" or "common language" makes as much sense as the notion "nearby city" or "look alike," without further specification of interests, leaving the analysis vacuous. 20 For familiar reasons, nothing in this suggests that there is any problem in informal usage, any more than in the ordinary use of such expressions as "Boston is near New York" or "John is almost home." It is just that we do not expect such notions to enter into explanatory theoretical discourse. They may be appropriate for informal discussion of what people do, with tacit assumptions of the kind that underlie ordinary discourse in particular circumstances; or even for technical discourse, where the relevant qualifications are tacitly understood. They have no further place in naturalistic inquiry, or in any attempt to sharpen understanding. Alleged social factors in language use often have a natural individualistinternalist interpretation. If Peter is improving his Italian or Gianni is learning his, they are (in quite different ways) becoming more like a wide range of people; both the modes of approximation and selection of models vary

216


with our interests. We gain no insight into what they are doing by supposing that there is a fixed entity that they are approaching, even if some sense can be made of this mysterious notion. If Bert complains of arthritis in his ankle and thigh, and is told that he is wrong about both, but in different ways, he may (or may not) choose to modify his usage to that of the doctor's. Apart from further detail, which may vary widely with changing contingencies and concerns, nothing seems missing from this account. Similarly, ordinary talk of whether a person has mastered a concept requires no notion of common language. To say that Bert has not mastered the concept arthritis orjlu is simply to say that his usage is not exactly that of people we rely on to cure usa nonnal situation. If my neighbor Bert tells me about his arthritis, my initial posit is that he is identical to me in this usage. I will introduce modifications to interpret him as circumstances require; reference to a presumed "public language" with an "actual content" for arthritis sheds no further light on what is happening between us, even if some clear sense can be given to the tacitly assumed notions. If I know nothing about elms and beeches beyond the fact that they are large deciduous trees, nothing beyond this infonnation might be represented in my mental lexicon (possibly not even that, as noted earlier); the understood difference in referential properties may be a consequence of a condition holding of the lexicon generally: lack of indication of a semantic relation is taken to indicate that it does not hold. 21 Questions remain-factual ones, I presume-as to just what kind of infonnation is within the lexicon, as distinct from belief systems. Changes in usage, as in the preceding cases, may in fact be marginal changes of 1language, or changes in belief systems, here construed as (narrowly described) C-R systems of the mind, which enrich the perspectives and standpoints for thought, interpretation, language use, and other actions (call them I-belief systems, some counterpart to beliefs that might be discovered in naturalistic inquiry). Work in lexical semantics provides a basis for empirical resolution in some cases (particularly in the verbal system, with its richer relational structure), keeping to the individualist-internalist framework. Little is understood about the general architecture of the mindlbrain outside of a few scattered areas, typically not those that have been the focus of the most general considerations of so-called "cognitive ,science." There has, for example, been much interesting discussion about a theory of belief and its possible place in accounting for thought and action. But substantive empirical work that might help in examining, refining, or testing these ideas is scarcely available. It seems reasonable at least to suppose that I-beliefs do not fonn a homogeneous set; the system has further structure that may provide materials for decisions about false belief and misidentification. Suppose that some I-beliefs are identifying beliefs and others not, or that they range along such a spectrum, where the latter (or the lesser) are more readily

217


abandoned without affecting conditions for referring. Suppose, say, that Peter's infonnation about Martin van Buren is exhausted by the belief that he was (a) the President of the United States and (b) the sixteenth President, (a) being more of an identifying belief than (b). If Peter learns that Lincoln was the sixteenth President he might drop the nonidentifying I-belief while using the tenn to refer. If he is credibly infonned that all the history books are mistaken and van Buren wasn't a President at all, he is at a loss as to how to proceed. That seems a reasonable first step towards as much of an analysis as an internalist perspective can provide, and as much as seems factually at all clear. Further judgments can sometimes be made in particular circumstances, in varied and conflicting ways.22 It may be that a kind of public (or interpersonal) character to thought and meaning results from uniformity of initial endowment, which permits only I-languages that are alike in significant respects, thus providing some empirical reason to adopt some version of the Fregean doctrine that "it cannot well be denied that mankind possesses a common treasure of thoughts which is transmitted from generation to generation." And the special constructions of the science-forming faculty may also approach a public character (more to the point, for Frege' s particular concerns). But for the systems that grow naturally in the mind, beyond the instantiation of initial endowment as 1language (perhaps also I-belief and related systems), the character of thought and meaning varies as interest and circumstance vary, with no clear way to establish further categories, even ideally. Appeals to common origin of language or speculations about natural selection, which are found throughout the literature, seem completely beside the point. Consider the shared initial state of the language faculty of the brain, and the limited range of I-languages that are attainable as it develops in early life. When we inquire into lexical properties, we find a rich texture of purely internalist semantics, with interesting general properties, and evidence for fonnal semantic relations (including analytic connections).23 Furthennore, a large part of this semantic structure appears to derive from our inner nature, determined by the initial state of our language faculty, hence unlearned and universal for I-languages. Much the same is true of phonetic and other properties. In short, I-language (including internalist semantics) seems much like other parts of the biological world. We might well tenn all of this a fonn of syntax, that is, the study of the symbolic systems of C-R theories ("mental representation"). The same terminology remains appropriate if the theoretical apparatus is elaborated to include mental models, discourse representations, semantic values, possible worlds as commonly construed, and other theoretical constructions that still must be related in some manner to things in the world; or to the entities postulated by our science-forming faculty, or constructed by other faculties of the mind.

218


The internally determined properties of linguistic expressions can be quite far-reaching, even in very simple cases. Consider again the word house, say, in the expression "John is painting the house brown," a certain collection of structural, phonetic, and semantic properties. We say it is the same expression for Peter and Tom only in the sense in which we might say that their circulatory or visual systems are the same: they are similar enough for the purposes at hand. One structural property of the expression is that it consists of six words. Other structural properties differentiate it from "John is painting the brown house," which has correspondingly different conditions of use. A phonetic property is that the last two words, house and brown, share the same vowel; they are in the formal relation of assonance, while house and mouse are in the formal relation of rhyme, two relations on linguistic expressions definable in terms of their phonological features. 24 A semantic property is that one of the two final words can be used to refer to certain kinds of things, and the other expresses a property of these. Here too there are formal relations expressible in terms of features of the items, for example, between house and building. Or, to take a more interesting property, if John is painting the house brown, then he is applying paint to its exterior surface, not its interior; a relation of entailment holds between the corresponding linguistic expreSSlons. Viewed formally, relations of entailment have much the same status as rhyme; they are formal relations among expressions, which can be characterized in terms of their linguistic features. Certain relations happen to be interesting ones, as distinct from many that are not, because of the ways 1languages are embedded in performance systems that use these instructions for various human activities. Some properties of the expression are universal, others languageparticular. It is a universal phonetic property that the vowel of house is shorter than the vowel of brown; it is a particular property that the vowel in my I-language is front rather than mid, as it is in some I-languages similar to mine. The fact that a brown house has a brown exterior, not interior, appears to be a language universal, holding of "container" words of a broad category, including ones we might invent: box, airplane, igloo, lean-to, etc. To paint a spherical cube brown is to give it a brown exterior. The fact that house is distinguished from home is a particular feature ,of the I-language. In English, I return to my home after work; in Hebrew, I return to the house. When we move beyond lexical structure, conclusions about the richness of the initial state of the language faculty, and its apparently special structure, are reinforced. Consider such expressions as (2): (2) (a) he thinks the young man is a genius. (b) the young man thinks he is a genius. (c) his mother thinks the young man is a genius.

219


In (b) or (c), the pronoun may be referentially dependent on the young man; in (a) it cannot (though it might be used to refer to the young man in question, an irrelevant matter). The principles underlying these facts appear to be universal, at least in large measure;25 again, they yield rich conditions on semantic interpretation, on intrinsic relations of meaning among expressions, including analytic connections. Furthennore, in this domain we have theoretical results of some depth, with surprising consequences. Thus the same principles appear to yield the semantic properties of expressions of the fonn (1), mentioned earlier. Given the perfonnance systems, the representation at the interface level PF imposes restrictive conditions on use (articulation and perception, in this case). The same is true of the LF representation, as illustrated in (1) and (2), or at the lexical level, in the special status of the exterior surface for container words. A closer look reveals further complexity. The exterior surface is distinguished in other ways within I-language semantics. If I see the house, I see its exterior surface; seeing the interior surface does not suffice. If I am inside an airplane, I see it only if I look out the window and see the surface of the wing, or if there is a mirror outside that reflects its exterior surface. But the house is not just its exterior surface, a geometrical entity. If Peter and Mary are equidistant from the surface, Peter inside and Mary outside, Peter is not near the house, but Mary might be, depending on the current conditions for nearness. The house can have chairs inside it or outside it, consistent with its being regarded as a surface. But while those outside may be near it, those inside are necessarily not. So the house involves its exterior surface and its interior. But the interior is abstractly conceived; it is the same house if I fill it with cheese or move the walls-though if I clean the house I may interact only with things in the interior space, and I am referring only to these when I say that the house is a mess or needs to be redecorated. The house is conceived as an exterior surface and an interior space (with complex properties). Of course, the house itself is a concrete object; it can be made of bricks or wood, and a wooden house does not just have a wooden exterior. A brown wooden house has a brown exterior (adopting the abstract perspective) and is made of wood (adopting the concrete perspective). If my house used to be in Philadelphia, but is now in Boston, then a physical object was moved. In contrast, if my home used to be in Philadelphia, but is now in Boston, then no physical object need have moved, though my home is also concretethough in some manner also abstract, whether understood as the house in which I live, or the town, or country, or universe; a house is concrete in a very different sense. The house-home difference has numerous consequences: I can go home, but not go house; I can live in a brown house, but not a brown home; in many languages, the counterpart of home is adverbial, as partially in English too.

220


Even in this trivial example, we see that the internal conditions on meaning are rich, complex, and unsuspected; in fact, barely known. The most elaborate dictionaries do not dream of such subtleties; they provide no more than hints that enable the intended concept to be identified by those who already have it (at least, in essential respects). The I-variant of Frege's telescope operates in curious and intricate ways. There seems at first glance to be something paradoxical in these descriptions. Thus, houses and homes are concrete, but from another point of view, are considered quite abstractly, though abstractly in very different ways; similarly, books, decks of cards, cities, etc. It is not that we have confused ideas, or inconsistent beliefs, about houses and homes, or boxes, airplanes, igloos, spherical cubes, etc. Rather, a lexical item provides us with a certain range of perspectives for viewing what we take to be the things in the world, or what we conceive in other ways; these items are like filters or lenses, providing ways of looking at things and thinking about the products of our minds. The terms themselves do not refer, at least if the term refer is used in its natural language sense; but people can use them to refer to things, viewing them from particular points of view-which are remote from the standpoint of the natural sciences, as noted. The same is true wherever we inquire into I-language. London is not a fiction, but considering it as London-that is, through the perspective of a city name, a particular type of linguistic expression-we accord it curious properties: as noted earlier, we allow that under some circumstances, it could be completely destroyed and rebuilt somewhere else, years or even millennia later, still being London, that same city. Charles Dickens described Washington as "the City of Magnificent Intentions," with "spacious avenues, that begin in nothing, and lead nowhere; streets, mile-long, that only want houses, roads, and inhabitants; public buildings that need but a public to be complete; and ornaments of great thoroughfares, which only lack great thoroughfares to ornament"-but still Washington. We can regard London with or without regard to its popUlation: from one point of view, it is the same city if its people desert it; from another, we can say that London came to have a harsher feel to it through the Thatcher years, a comment on how people act and live. Referring to London, we can be talking about a location or area, people who sometimes live there, the air above it (but ~ot too high), buildings, institutions, etc., in various combinations (as in "London is so unhappy, ugly, and polluted that it should be destroyed and rebuilt 100 miles away," still being the same city). Such terms as London are used to talk about the actual world, but there neither are nor are believed to be things-in-the-world with the properties of the intricate modes of reference that a city name encapsulates. Two such collections of perspectives can fit differently into Peter's system of beliefs, as in Kripke's puzzle. 26

221


For purposes of naturalistic inquiry, we construct a picture of the world that is dissociated from these "common sense" perspectives (never completely, of course; we cannot become something other than the creatures we are 27). If we intermingle such different ways of thinking about the world, we may find ourselves attributing to people strange and even contradictory beliefs about objects that are to be regarded somehow apart from the means provided by the I-language and the I-belief systems that add further texture to interpretation. The situation will seem even more puzzling if we entertain the obscure idea that certain terms have a relation to things ("reference") fixed in a common public language, which perhaps even exists "independently of any particular speakers," who have a "partial, and partially erroneous, grasp of the language" (Michael Dummett);28 and that these "public language terms" in the common language refer (in some sense to be explained) to such objects as London taken as a thing divorced from the properties provided by the city name (or some other mode of designation) in a particular I-language, and from the other factors that enter into Peter's referring to London. Problems will seem to deepen further if we abstract from the background of individual or shared beliefs that underlie normal language use. All such moves go beyond the bounds of a naturalistic approach, some of them, perhaps, beyond sensible discourse. They also go beyond intemalist limits, which is a different matter. A naturalistic approach does not impose intemalist, individualist limits. Thus if we study (some counterpart to) persons as phases in the history of ideally immortal germ cells, or as stages in the conversion of oxygen to carbon dioxide, we depart from such limits. But if we are interested in accounting for what people do, and why, insofar as that is possible through naturalistic inquiry, the argument for keeping to these limits seems persuasive.29 We began by considering the (hypothetical) discovery that Peter's brain produces the configuration C when he thinks about cats. We then moved to the more realistic example of ERPs, and the still more realistic case (from a scientific standpoint) of C-R systems; one may think of their elements as on a par with C, though now real, not hypothetical, we have reason to believe. The same would be true of a naturalistic approach that departs from these intemalist limits, viewing Peter's brain as part of a larger system of interactions. The analogy would no longer be to the configuration C produced in Peter's brain when he thinks of cats, but to some physical configuration C' involving C along with something else, perhaps something about cats. We are now in the domain of the hypothetical-I know of no serious candidate. But suppose that such an approach can be devised and proves to yield insight into questions of language use. If so, that might modify the ways we study language and psychology, but would not bridge the gap to an account of people and what they do. We have to distinguish between a hypothetical extemalist naturalism of

222


the kind just sketched, and nonnaturalist externalism that attempts to treat human action (referring to or thinking about cats, etc.) in the context of communities, real or imagined things in the world, and so on. Such approaches are to be judged on their merits, as efforts to make some sense out of questions that lie beyond naturalistic inquiry, like questions about energy, falling stones, the heavens, etc., in the ordinary sense of the tenns. I have mentioned some reason for skepticism about recourse to communities and their practices, or public languages with public meanings. Consider further the other facet of externalism, an alleged relation between words and things. Within internalist semantics, there are explanatory theories of considerable interest that are d((veloped in tenns of a relation R (read "refer") that is postulated to hold between linguistic expressions and something else, entities drawn from some stipulated domain D (perhaps semantic values).3o The relation R, for example, holds between the expressions London (house, etc.) and entities of D that are assumed to have some relation to what people refer to when they use the words London (house, etc.), though that presumed relation remains obscure. As noted, I think such theories should be regarded as a variety of syntax. The elements they postulate are on a par, in the respects relevant here, with phonological or phrase structure representations, or the hypothetical brain configuration C; we might well include D and R within the SD (the linguistic expression), as part of an interface level. Explanation of the phenomena of (2) (repeated here) is commonly expressed in tenns of the relation R: (2) (a) he thinks the young man is a genius. (b) the young man thinks he is a genius. (c) his mother thinks the young man is a genius. The same theories of binding and anaphora carry over without essential change if we replace young in (2) by average, typical, or replace the young man by John Doe, stipulated to be the average man for the purposes of a particular discourse. 31 And to anaphoric properties of the pronouns in (3), (4): (3) (a) it brings good health's rewards.

(b) good health brings its rewards. . (c) its rewards are what make good health worth striving for. (4) (a) [there is a flaw in the argument], but it was quickly found. (b) [the argument is flawed], but it was quickly found. In tenns of the relation R, stipulated to hold between the average man, John Doe, good health, flaw, and entities drawn from D, we can account for the differential behavior of the pronoun exactly as we would with the young

223


man, Peter, fly ("there is a fly in the coffee"). The relations of anaphora differ in (4a,b), though there is no relevant difference in meaning between the bracketed clauses. And it might well tum out that these expressions, along with such others as "the argument has a flaw" (with the anaphoric options of (4a», share still deeper structural properties, possibly even the same structural representation at the level relevant to the internal semantics of the phrases, a possibility that has been explored for some years.32 The same is true in more exotic cases. It would seem perverse to seek a relation between entities in D and things in the world-real, imagined, or whatever; at least, one of any generality. One may imagine that the relation of elements of D to things in the world is more "transparent" than in the case of other syntactic representations, as the relation to sound waves is more "transparent" for phonetic than for phonological representation; but even if so, these studies do not pass beyond the syntax of mental representations. The relation R and the construct D must be justified on the same kinds of grounds that justify other technical syntactic notions; those of phonology, or the typology of empty categories in syntax. An occasional resemblance between R and the term refer of ordinary language has no more significance than it would in the case of momentum or undecidability. Specifically, we have no intuitions about R, any more than we do about momentum or undecidability in the technical sense, or about c-command or autosegmental in (other parts) ofthe C-R theories of syntax;33 the terms have the meanings assigned to them. We have intuitive judgments about the notion used in such expressions as "Mary often refers to the young man as a friend (to the average man as John Doe, to good health as life's highest goal)." But we have no such intuitions about the relation R holding between Mary (or the average man, John Doe, good health, flaw) and postulated elements of D. R and D are what we specify that they are, within a framework of theoretical explanation. We might compare Rand D to P and PF, where P is a relation holding between an expression and its PF representation (between "took" and [thuk], perhaps), though in the latter case the concepts fit into a much better-grounded and richer theory of interface relations. Suppose that postulation of Rand D is justified by explanatory success within the C-R theory of I-language, alongside of P and PF, c-command, and autosegmental. That result lends no support to the belief that some R-like relation, call it R', holds between words and things, or things as they are imagined to be, or otherwise conceived. Postulation of such a relation would have to be justified on some grounds, as in the case of any other invented technical notion. And if we devise a relation R' holding between linguistic expressions and "things," somehow construed, we would have no intuitions about it-matters become only more obscure if we invoke unexplained notions of "community" or "public language," taken in some absolute sense. We do have intuitive judgments concerning linguistic expressions and the

224


particular perspectives and points of view they provide for interpretation and thought. And we might proceed to study how these expressions and perspectives enter into various human actions, such as referring. Beyond that, we enter the realm of technical discourse, deprived of intuitive judgment. Take Putnam's influential twin-earth thought experiment. We can have no intuitions as to whether the term water has the same "reference" for Oscar and twin-Oscar: that is a matter of decision about the new technical term "reference" (some particular choice for R'). We have judgments about what Oscar and twin-Oscar might be referring to, judgments that seem to vary considerably as circumstances vary. Under some circumstances, Putnam's proposals abo~t "same liquid," a (perhaps unknown) notion ofthe natural sciences, seem very plausible; under other circumstances, notions ()f sameness or similarity drawn from common-sense understanding seem more appropriate, yielding different judgments. It does not seem to me at all clear that there is anything general to say about these matters, or that any general or useful sense can be given to such technical notions as "wide content" (or any other notion fixing "reference") in any of the extemalist interpretations. If so, questions arise about the status of what Putnam, in his Locke lectures, calls the "social co-operation plus contribution of the environment theory of the specification of reference," a fuller and more adequate version of the "causal theory of reference" developed in his paper "The Meaning of 'Meaning'" and Saul Kripke's Naming and Necessity, both now landmarks in the field. "Social co-operation" has to do with "the division of linguistic labor": the role of experts in determining the reference of my terms elm and beech, for example. Putnam provides a convincing account for certain circumstances. Under some conditions, I would indeed agree that what I am referring to when I use the term elm is what is meant by an expert, perhaps an Italian gardener with whom I share only the Latin terms (though there is no meaningful sense in which we are part of the same "linguistic community" or speak a "common language"); under other conditions, probably not, but that is to be expected in an inquiry reaching as far as all of "human functional organization," virtually a study of everything. As mentioned earlier, it is not clear whether the question relates to I-language or I-belief, assuming the theoretical construction to be valid. As for the "environment theory," it could contribute to specification of reference only if there were some coherent notion of "reference" (R') holding between linguistic expressions and things, which is far from obvious, though people do use these expressions (in various ways) to refer to things, adopting the perspectives that these expressions provide. There are circumstances in which the particular conclusions usually drawn seem appropriate, in which "same species," "same liquid," etc., help determine what I am referring to; and there are other circumstances in which they do not. 34

225


It also seems unclear that metaphysical issues arise in this context. To take some of Kripke' s examples, doubtless there is an intuitive difference between the judgment that Nixon would be the same person if he had not been elected in 1968, while he would not be the same person if he were not a person at all (say, if he were a silicon-based person replica). But that follows from the fact that Nixon is a personal name, offering a way of referring to Nixon as a person; it has no metaphysical significance. If we abstract from the perspective provided by natural language, which appears to have no pure names in the logician's sense (the same is true of variables, at least if pronouns are considered variables, and of indexicals, if we consider their actual conditions of use in referring), then intuitions collapse: Nixon would be a different entity, I suppose, if his hair were combed differently. Similarly, the object in front of me is not essentially a desk or a table; that very object could be any number of different things, as interests, functions, intentions of the inventor, etc., vary. To cite some recent work, Joseph Almog's judgment that the mountain Nanga Parbat is a mountain essentially might be intelligible under some circumstances, but contrary to what he assumes, his "coherentabstraction test" seems to me to permit us, under other circumstances, to deprive Nanga Parbat of this property, leaving it as the same entity: say, if the sea level rises high enough for its top to become an island, in which case it is no more a mountain than England is; or if earth is piled around it up to its peak, but a millimeter away, in which case it is not a mountain but part of a plateau surrounded by a crevice, though it remains the very same entity. 35 In summary, it is questionable that standard conclusions can survive a closer analysis of the technical notions "reference" (in some R'-like sense) or "specification of reference." There may well be justification for the notion R internal to C-R theories (basically a syntactic notion, despite appearances). But there seems to be little reason to suppose that an analogous notion R' can be given a coherent and useful formulation as a relation holding between expressions and some kind of things, divorced from particular conditions and circumstances of referring. If that is so, there will also be no reasonable inquiry into a notion of "sense" or "content" that "fixes reference" (R'), at least for natural language, though there is a promising (syntactic) inquiry into conditions for language use (including referring). As discussed earlier, naturalistic inquiry may lead to the creation of language-like accretions to the I-language; for these, an R'-like notion may be appropriate, as terms are divested of the I-language properties that provide interpretive perspectives and semantic relations, are dissociated from I-belief, and are assigned properties lacking in natural language. These constructed systems may use resources of the I-language (pronunciation, morphology, sentence structure, etc.), or may transcend them (introducing mathematical formalisms, for example). The I-language is a product of the language faculty, abstracted from other components of the mind; an idealization of course,

226


hence to be justified or rejected on the basis of its role in an explanatory framework. The picture could be extended, plausibly it seems, by distinguishing the system of common-sense belief from products of the scienceforming faculty. The latter are neither I-languages nor I-belief systems, and for these it may well be appropriate to stipulate a relation R'. Some of the motivation for externalist approaches derives from the concern to make sense of the history of science. Thus Putnam argues that we should take the early Niels Bohr to have been referring to electrons in the quantum-theoretic sense, or we would have to "dismiss all of his 1900 beliefs as totally wrong,"36 perhaps on a par with someone's beliefs about angels, a conclusion that is plainly absurd. The same is true of pre-Dalton chemists speaking of atoms. And perhaps, on the same grounds, we would say that chemists pre-Avogadro were referring to what we call atoms and molecules, though for them the terms were interchangeable, apparently. The discussion assumes that such terms as electron belong to the same system as house, water, and pronominal anaphora, so that conclusions about electron carry over to notions in the latter category. That assumption seems to be implicit in Putnam's proposal that "To determine the intrinsic complexity of a task is to ask, How hard is it in the hardest case?," the "hardest case" for "same reference" or "same meaning" being posed by such concepts as momentum or electron in physics. But the assumption is dubious. The study of language should seek a more differentiated picture than that, and what is true of the technical constructions of the science-forming faculty might not hold for the natural language lexicon. Suppose we grant the point nevertheless. Agreeing further that an interest in intelligibility in scientific discourse across time is a fair enough concern, still it cannot serve as the basis for a general theory of meaning; it is, after all, only one concern among many, and not a central one for the study of human psychology. Furthermore, there are internalist paraphrases. Thus, we might say that in Bohr's earlier usage, he expressed beliefs that were literally false, because there was nothing of the sort he had in mind in referring to electrons; but his picture of the world and articulation of it was structurally similar enough to later conceptions so that we can distinguish his beliefs about electrons from beliefs about angels. What is more, that seems a reasonable way to proceed. To take a far simpler example from the study of language, consider a debate some thirty years ago over the nature of phonological units. Structural phonologists postulated segments (phonemes) and phonetic features, with a certain collection of properties. Generative phonologists argued that no such entities exist, and that the actual elements have somewhat different properties. Suppose that one of these approaches looks correct (say, the latter). Were structural phonologists therefore referring all along to segments and features in the sense of generative phonology? Surely not. They flatly denied that, and were right to do so. Were they talking gibberish? Again, surely not.

227


Structuralist phonology is intelligible; without any assumption that there are entities of the kind it postulated, much of the theory can be reinterpreted within generative phonology, with results essentially carried over. There is no principled way to determine how this is done, or to determine the "similarity of belief' between the two schools of thought or what thoughts and beliefs they shared. Sometimes it is useful to note resemblances and reformulate ideas, sometimes not. The same is true of the earlier and later Bohr. Nothing more definite is required to maintain the integrity of the scientific enterprise or a respectable notion of progress towards the truth about the world, insofar as it falls within human cognitive capacity. It is worth noting that an analysis in these tenns, eschewing externalist assumptions on fixation of reference, is consistent with the intuitions of respected figures. The discussion of the meaning of electron, water, etc., projects backwards in time, but we can project forward as well. Consider the question whether machines can think (understand, plan, solve problems, etc). By standard externalist arguments, the question should be settled by the truth about thought: what is the essence of Peter's thinking about his children, or solving a quadratic equation, or playing chess, or interpreting a sentence, or deciding whether to wear a raincoat? But that is not the way it seemed to Wittgenstein and Turing, to take two notable examples. For Wittgenstein, the question whether machines think cannot seriously be posed: "We can only say of a human being and what is like one that it thinks," maybe dolls and spirits; that is the way the tool is used. Turing, in his classic 1950 paper, wrote that the question whether machines can think may be too meaningless to deserve discussion. Nevertheless I believe that at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.

Wittgenstein and Turing do not adopt the standard externalist account. For Wittgenstein, the questions are just silly: the tools are used as they are, and if the usage changes, the language has changed, the language being nothing more than the way we use the tools. Turing too speaks of the language of "general educated opinion" changing, as interests and concerns change. In our terms, there will be a shift from the I-languages that Wittgenstein describes to new ones, in which the old word think will be eliminated in favor of a new word that applies to machines as well as people. To ask in 1950 whether machines think is as meaningful as the question whether airplanes and people (say, high jumpers) really fly; in English airplanes do and high jumpers don't (except metaphorically), in Hebrew neither do, in Japanese both do. Such facts tell us nothing about the (meaningless) question posed, but only about marginal and rather arbitrary variations of 1language. The question of what atom meant pre-Dalton, or electron for Bohr

228


in 1900, seems comparable, in relevant respects, to the question of what think meant for Wittgenstein and Turing; not entirely comparable, because think, atom, and electron should probably not be regarded as belonging to a homogenous I-language. In all these cases, the internalist perspective seems adequate, not only to the intuitions of Wittgenstein and Turing, but to an account of what is transpiring; or what might happen as circumstances and interests vary. Perhaps one might argue that recent semantic theories supersede the intuitions of Wittgenstein and Turing because of their explanatory success. But that does not seem a promising idea; explanatory success will hardly bear that burden. In general, we have little reason now to believe that more than a Wittgensteinian assembly of particulars lies beyond the domain of internalist inquiry, which is, however, far richer and informative than Wittgenstein, Austin, and others supposed. Naturalistic inquiry will always fall short of intentionality. At least in these terms, "intentionality won't be reduced and won't go away," as Putnam puts it, and "language speaking" will remain not "theoretically explicable." The study of C-R systems, including "internalist semantics," appears to be, for now, the most promising form of naturalistic inquiry, with a reasonably successful research program; understanding of performance systems is more rudimentary but within the range of inquiry, in some respects at least. These approaches raise problems of the kind familiar throughout the natural sciences, but none that seem qualitatively different. Pursuing them, we can hope to learn a good deal about the devices that are used to articulate thoughts, interpret, and so on. They leave untouched many other questions, but it remains to be shown that these are real questions, not pseudo-questions that indicate topics of inquiry that one might hope to explore, but little more than that.

NOTES 1. Meaning and the Moral Sciences (London: Routledge & Kegan Paul, 1978). 2. Putnam, Representation and Reality (Cambridge, Mass.: MIT Press, 1988). 3. H. M. Jaeger and Sidney R. Nagel, "Physics of the Granular State," Science, March 20, 1992. 4. Tyler Burge, "Philosophy of Language and Mind," Philosophical Review. Centennial Issue, Jan. 1992. 5. See, among others, Julius Moravcsik, "Aitia as Generative Factor in Aristotle's Philosophy," Dialogue 14: 622-36 (1975); Thought and Language (London: Routledge, 1990). Chomsky, Reflections on Language (New York: Pantheon, 1975). Sylvain Bromberger, "Types and Tokens in Linguistics," in Bromberger, On What We Know We Don't Know (Chicago: University of Chicago Press, 1992). By "internal relational structure" I mean the selectional properties of such words as "give" (which takes an agent subject, theme object, and goal indirect object), lacking for "cat," "liquid," etc.

229


6. Representation and Reality. 7. Davies accepts Tyler Burge's position that work of the Marr school is concerned with "informational" representations with intentional content (hence with actual causal antecedents), but that position does not seem reconcilable with actual experimental practice or theoretical results (e.g., Ullman's rigidity principle); it is hard to see how it could be correct if only because, as Davies emphasizes, Marr's work did not reach to 3D model representation at all. Insofar as the study of visual perception does so (e.g., Elizabeth Spelke's work on object constancy in infancy), it keeps to visual experience, not perceptual content in the technical sense of philosophical discourse. Ullman, The Interpretation of Visual Motion (Cambridge, Mass.: MIT Press, 1979); Davies, "Individualism and Perceptual Content," Mind, October 1991. 8. P. F. Strawson, Introduction to Logical Theory (Methuen: London, 1952),216. 9. Helen Neville, Janet Nicol, Andrew Barss, Kenneth Forster, and Merrill Garrett, "Syntactically Based Sentence Processing Classes: Evidence from Event-Related Brain Potentials," Journal of Cognitive Neuroscience 3 (2) (1991): 151-65. 10. On some misunderstandings of these matters, see my "Language and Interpretation: Philosophical Reftections and Empirical Inquiry" (1986), in John Earman, ed., Inference, Explanation and Other Philosophical Frustrations (Berkeley: University of California Press, 1992); and my papers in Asa Kasher, ed., The Chomskyan Turn (Cambridge: Blackwell, 1991). II. It reveals such a rich vascular system, Richard Lewontin remarks, that to the fanciful stories concocted about evolution of cognition one might add the speculation that the brain evolved as a thermoregulator, cooling the blood as Aristotle thought and yielding human cognition as a byproduct; "The evolution of cognition" in D. N. Osherson and E. E. Smith, An Invitation to Cognitive Science, vol. 3 (Cambridge Mass.: MIT Press, 1990). 12. See Jeni Yamada, Laura (Cambridge Mass.: MIT Press, 1990), and John Marshall's foreword and sources cited. 13. See James Higginbotham, "On Semantics," Linguistic Inquiry 16: 547-93, 1985; "Elucidations of Meaning," Linguistics and Philosophy 12: 465-518,1989. 14. There is, again, no implication here that the actual performance systems will correspond closely to informal usage, or philosophical or other technical discourse. 15. Much less likely, even if the phrase can be given some meaning clear enough for the question to be sensibly raised. 16. The topic has been widely debated since John Searle's "Minds, Brains, and Programs," Behavioral and Brain Sciences 3 (Sept. 1980). It is not clear that any substantive issue has yet been formulated. 17. Davidson, "A Nice Derangement of Epitaphs," in E. Lepore, ed., Truth and Interpretation (Oxford: Blackwell, 1986); Bjorn Ramberg, Donald Davidson's Philosophy of Language (Oxford: Blackwell, 1989). 18. Jerry Fodor and Ernest Lepore, Holism: a Shoppers' Guide (Cambridge, Mass.: Blackwell, 1992). The interstage problem is held to arise only on the assumption of "semantic holism." 19. These procedures are not to be confused with principles of charity and the like, if the language-belief distinction is valid; see below. To be even minimally realistic, we should distinguish many cases. Thus what Peter does when Mary speaks a closely related language may have little relation to his procedure when she speaks an unintelligible one. Subsuming all such processes under "interpretation" or "translation" is not a good research strategy. 20. On Saul Kripke's development of this approach, and his conclusions about its relevance to linguistics, see my Knowledge of Language (Praeger, 1986), chapter 4.1. 21. In Representation and Reality, Putnam argues against the assumption that the lexical entry includes specific reference to expert judgment. The argument is based on tacit assumptions about common public language and translation that do not seem to me easy to defend, or

230


22.

23. 24. 25.

26. 27.

28. 29.

30.

31. 32.

33.

34.

35. 36.

even fonnulate. We might, however, accept the conclusion, considering reliance on expert judgment (among other options) to be a general property of a wide range of lexical entries, relating to the ways they enter into belief systems. See Stephen Stich, From Folk Psychology to Cognitive Science (Cambridge, Mass.: MIT Press, 1983). The basic problem-that any criteria we put forth are at once too strong and too weak-was outlined by Israel Scheffler, "On Synonymy and Indirect Discourse," Philosophy of Science 22.( 1): 3~. See references of note 10. Technically, we should speak of "I-rhyme," etc. See Howard Lasnik, Essays on Anaphora (Dordrecht: Kluwer Academic, 1989), particularly chapter 9. Interesting questions arise in the case of (c) ("backwards pronominalization") with regard to such matters as referential use of definite descriptions and old-new infonnation. See Akeel Bilgrami, Belief and Meaning (Cambridge, Mass.: Blackwell, 1992), for extensive discussion from a somewhat similar point of view. Putnam has frequently stressed that standards for inference and justification of belief are inescapably interest-relative. Furthennore, the particular character (and therefore limits) of human understanding impose choices of framework for theory that may be inappropriate, leaving problem areas that are inherently mysteries for humans (a general property of organisms). See my Reflections on Language, and Colin McGinn, The Problem of Consciousness (Blackwell: Oxford, 1991). Dummett, "Comments on Davidson and Hacking," in LePore, op. cit. Not in question, of course, is the fact that what people do depends upon events elsewhere in space and time; the question is whether naturalistic inquiry will be "Markovian," taking only the resulting state of the organism to enter into local current perfonnance. Thus memories may fade or be reshaped, but to understand what a person is doing here and now, we ask what is internally represented, not what may once have happened. Similarly, the growth of a cell to a finger or a bone of the forearm depends on elapsed time, but the study of the process keeps to such indicators as current gradients of chemical concentration that infonn the cell of such facts. That is standard, and it seems very reasonable, procedure. Whether the theories should be developed in these tenns is another matter. My point is simply to note that if they rely on notions of intended reference, referential dependence, etc., as more than a fa~on de parler, then something of the sort outlined here seems to be presupposed-not reference to things in (or believed to be in) the world. There are differences in backwards pronominalization; see note 25. See Mireille Tremblay, "Possession and Datives," Ph.D. diss., McGill University, 1991. The basic point about "systematically misleading expressions" in Ryle's sense is traceable at least to the eighteenth-century critique of the theory of ideas by du Marsais and later Thomas Reid; see my Aspects of the Theory of Syntax (Cambridge, Mass.: MIT Press, 1965), 199-200. Or about perceptual content in the special technical sense of philosophical discourse; see note 7 and text. The distinction Davies draws between "conservative" and "revisionary" interpretations of the technical notion is not clear, any mor~ than we can distinguish conservative and revisionary interpretations of electromagnetic force. Note Stich's observations (op. cit., 58, 62) about the inability of "most ears not previously contaminated by philosophical theory" to provide judgments at all in many such cases. The observation is not necessarily decisive; perhaps the facts of folk psychology can only be discerned by trained and guided intuition. With a richer theoretical context, that might be a reasonable sunnise. But there is virtually no theoretical context, hence little reason to regard the isolated judgments as meaning much. Almog, "The What and the How," Journal of Philosophy 5 (1991): 225-44. Representation and Reality.

231

Bilgrami, Akeel, Can Externalism Be Reconciled with Self-Knowledge? , Philosophical Topics, 20:1 (1992:Spring) p.233


Can Externalism Be Reconciled with Self-Knowledge?

Akeel Bilgrami Columbia University

I. IT ALL STARTED WITH PUTNAM It is a tribute not just to his impressive originality as a philosopher but to the systematic range of his thought that Hilary Putnam, who--along with Kripke, Marcus, Chastain, and no doubt some others-founded a certain conception of the nature of reference and meaning, was the first to then go on to draw the consequences it has for the nature of intentionality. The conception I have in mind is, of course, the so-called 'causal' conception of reference and meaning, sometimes also called 'direct reference'. The consequences 1 have in mind have amounted to a view of intentionality which is often called the 'extemalist' account of intentional content. My concern in this paper is with the latter and, in particular, with one very important and problematic theme that emerges from this externalist view of intentional content. This is the theme of self-knowledge of our intentional contents, sometimes also described as a 'first-person authority' that agents have over their intentional contents. (I will use the expressions 'self-knowledge' and 'first-person authority' more or less interchangeably throughout the paper except at one crucial point, where the latter will mark a special non-inferentiality of much of our self-knowledge. For this see note 22 and the text to which it attaches.) It is widely thought that there is at the

233


very least a prima Jacie difficulty that externalism raises for the idea that we have self-knowledge, in cases where we would otherwise think it intuitive that we do have it. Even those who deny that it poses a difficulty have tended to acknowledge that there is at least a prima Jacie problem that they must remove. I won't pause to sketch in detail Putnam's causal views of reference and meaning, nor to spell out the details by which it comes to have the externalist consequences it has in the philosophy of mind. These are well known and have been extensively elaborated in the voluminous commentary that Putnam has inspired. Let me simply begin with a general definition of externalism as one finds it in his paper "The Meaning of 'Meaning' ."1 Externalism is a denial of the following thesis: Intentional states, in the ordinary non-scientific sense that Aristotle intended when he devised the 'practical syllogism', that is, states such as beliefs and desires which we invoke to make sense of or 'rationalize' people's various doings, "do not presuppose the existence of any thing external to the agent who possesses such states" (136). The denied thesis may and has been called 'internalism' or, in Putnam's own phrase, 'methodological solipsism'; its denial has been called 'externalism'. I will call this externalism (G.E.) to mark the perfectly general definition of externalism just given. I believe that externalism, so generally and minimally defined, is an extremely important truth about intentionality. However, here is something well worth noting. Though (G.E.) is indeed a consequence of Putnam's causal conception of reference and meaning, there is no entailment in the opposite direction. That is, it is perfectly possible to accept externalism, generally defined as above (and as Putnam himself defines it) and not go on to embrace Putnam's views of reference and meaning nor the specific externalist picture of intentionality that these views have as a consequence. I have, elsewhere, proposed a specific externalism which falls under (G.E.) and which makes no appeal to those views of reference and meaning. 2 I say this at this point only to draw attention to the fact that we need to distinguish between a general characterization of externalism and specific externalist proposals which may fall under it. Putnam's views of reference and meaning give rise to a specific externalism, but it is quite uncompulsory for externalists, that is for those who subscribe to (G.E), to adopt these views and the externalism it gives rise to. (From now on, my use of the expression "Putnam's externalism" will be restricted to talk of his specific thesis rather than to his general definition.) This fact has gone more or less unnoticed, and almost everyone in the large class of contemporary philosophers of mind and language who thinks that there is something worthwhile in externalism takes it for granted that externalism must be cashed out in terms ofthese views. Since this amounts to such an orthodoxy today, I will call this specific version of externalism that falls under (G.E.), (O.E.).3 Because it is such an orthodoxy, (O.E.) is often simply conflated with (G.E.);

234


being an orthodoxy, its adherents cannot imagine any other specific externalist proposal than their own, and so they proceed as if their proposal is equivalent to the general thesis. I will argue that we would do well to guard against this conflation. For one of my principal and eventual claims in this paper will be that on the theme of self-knowledge of intentional states, externalists, that is, anyone who subscribes to (G.E.), would do better to shun (O.E.) for some alternative externalism which does not raise the difficult problems that (O.E.) raises for self-knowledge. I briefly sketch such an alternative externalism towards the end of the paper. But first, let me say something about why it is thought that externalism poses even aprim(lfacie problem for self-knowledge.

II. THE PRIMA FACIE PROBLEM EXTERNALISM POSES FOR SELF-KNOWLEDGE Before raising the problem, I should point out that here too not only was Putnam the first to point out what this problem is, but that Putnam accepted defeat in the face of the problem, and was the first to propose a highly influential way of living with it. However, some other externalists have been more defiant on Putnam's behalf and have tried to argue that the problem is only prima facie and that, on scrutiny, it is illusory. Let me explain all this. Why is it natural to assume, at least prima facie, that states which are constituted, in part, by things external to a person may sometimes not be known to the person who is in these states? In some recent papers Davidson asks exactly this question and answers it by invoking Putnam's twin-earth example.4 The idea is that a pair of internally identical twins on earth and twin earth respectively have different but, as far as they can tell, indistinguishable substances, which they both call 'water', in their environments; so, given externalism, they have different 'water' -concepts and 'water' -thoughts. Since they have different thoughts without really being able to tell the difference, they do not fully know what their own thoughts are. The example and its general point have been widely discussed, so I will not pause over details. The particular point that Davidson's discussion is intended to bring out is that if the contents of our thoughts are not constituted only by things internal to us in some suitably Cartesian sense, then it may seem at least prima facie natural to think that we will often not know what our thoughts are, since we may often not know crucial things about items external to us. He points out that Putnam grants this initially natural point as in the end undeniably true, and he criticizes Putnam for giving in to it too easily. Tyler Burge also raises this problem for externalism and he too thinks that we should hold out for a way of reconciling externalism and self-knowledge. 5

235


Putnam, of course, despite his concession, does not rest content with the conclusion that because of externalism we very often do not know the contents of our thoughts. To do so would be foolish, even in a thoroughly postFreudian climate. I say 'foolish' and mean it. However widespread psychological phenomena such as self-deception, self-censorship, inattention ... might be, these are psychological phenomena, which psychologists such as Freud, and many others, have studied; and though there may be some quarreling at the margins about the extent of their sway, we are intuitively willing to grant that such phenomena obstruct self-knowledge. What is foolish and unintuitive is to grant that a wholly non-psychological phenomenon, i.e., an entirely abstract philosophical doctrine about reference, should be seen as raising an obstruction to self-knowledge. At any rate it is foolish and unintuitive to see it as obstructing the very same sort of (or the very same sense of) knowledge state that the psychological phenomena just listed obstructs. And to see it as obstructing some other kind of (or sense of) state of selfknowledge would be to make equivocal the notion of self-knowledge in a way that adds unnecessary complication and brings no illumination. So, as I was saying, Putnam, laudably, does not rest content with a denial of self-knowledge of our thoughts and goes on, in effect, to claim that there are two notions of a thought's content, one internal and one external, and it is only the latter which raises a problem for self-knowledge; the former is always available to the agent (unless, of course, it is obstructed by one of the psychological phenomena listed above). Thus we find that Putnam, after conceding lack of self-knowledge, bifurcates the intentional aspect of the mind into two, a distinction now routinely described as a distinction between 'narrow' (internal) intentional content and 'wide' (external) intentional content. But Davidson and Burge eschew this bifurcation because they think that it arises, in part, from an unnecessary surrender in the face of the problem raised only prima facie by externalism for self-knowledge. I will argue that Putnam is right in thinking that the problem is crippling for the idea of a unified or unbifurcated mind, and that Davidson and Burge do not see this only because they do not raise the problem for self-knowledge that Putnam's externalism poses in an appropriately specific way. The right way of raising the problem reveals the problem to be an insuperable one. But my defense of Putnam will only be partial. It will only be a defense against Davidson and Burge. The eventual point of my criticisms of Davidson and Burge will not be that all externalists are stuck with Putnam's bifurcated conception of intentionality. My solution, as I said, will rather be that we need to abandon orthodox externalism (O.E.), and fashion a new and alternative kind of specific position which satisfies (G.E.) but for which the problem regarding self-knowledge does not arise even prima facie.

236


III. DAVIDSON'S WAY OF RECONCILING EXTERNALISM WITH SELF-KNOWLEDGE Davidson's criticisms of Putnam depend crucially on a particular diagnosis he offers for why philosophers have thought first-person authority is possible in the first place. He first points out that there is a long tradition of postCartesian thinking which persists today in which it is taken for granted that there are objects of thoughts. In Descartes and various traditional and modem empiricists these were usually taken to be objects in consciousness, available in an authoritative and complete way to the subject of the thoughts and unavailable ~n such a way to any other subject. Davidson says that though contemporary philosophers like Searle still cling to the Cartesian version of such objects of thought, others have moved them some distance from their Cartesian moorings in inner consciousness and have taken them to be either propositions, or sentences in the head, or sometimes even (as in the somewhat exceptional case of Gareth Evans's and John McDowell's Russellian version of singular thoughts) external objects with which we are directly acquainted.6 But despite these differences between them, they all share the common idea that there are such objects of thought and, according to Davidson, it is this idea which they all assume makes self-knowledge possible. If, after all, the contents of our thought are defined in terms of special kinds of object authoritatively and completely available to us, then there is no problem regarding our knowledge of them. Davidson then goes on to say that Putnam, in shunning internalism (for his wide contents, anyway), gives up on the internalist idea that there are such objects of thought. 7 But having given up on the idea, he immediately concludes that with it he must also give up on first-person authority over such contents, for it is the idea which is supposed to account for such authority. So Davidson charges Putnam with having retained the traditional and longstanding internalist assumption that objects of thought alone can explain firstperson authority. That is to say, though Putnam, in giving up on internalism, gives up on there being such objects of thought (for wide content) he does not take the radical step of giving up on the assumption that it is these internalist objects of thought alone which account for our self-knowledge of such contents. This combination of retaining the assumption while giving up on the objects forces him to give up on self-kllowledge for those contents. Having diagnosed, to his own satisfaction, why it is that Putnam gives up on a reconciliation between self-knowledge and externalism, Davidson then goes on to say that the reconciliation should not be hard to achieve, if we do take the radical step of also giving up on the assumption that objects of thought account for self-knowledge. He too denies that there are such objects of thought which have any such epistemological standing. He does

237


not deny that when we attribute thoughts we use our interpreted sentences or utterances to specify the thoughts. But he denies that to do this amounts in any way to attributing anything which is within the epistemological or psychological ken of the agent to whom the thoughts are being attributed. And, since he gives up not only on the objects themselves but also on the assumption that we need any such epistemic intennediaries to account for self-knowledge, he concludes the following: If with Putnam we reject internalism, and therefore give up on such objects of thought, we may nevertheless now not follow him in giving up on self-knowledge for those externally constituted thoughts. It only remains for Davidson to give some alternative account of selfknowledge from the one he attributes to the tradition, an account which makes no appeal to the traditional idea of objects of thought. The account he offers is, roughly and crudely, that interpreting another person and attributing thoughts to her requires the assumption that she has direct and noninferential knowledge of her own thoughts. If we did not do so, we would not have anything to interpret. Thus the very idea of interpretation requires that the interpretee has authority over her own thoughts. And since, for Davidson, to be a thinker is to be the object of interpretation, all thinkers have firstperson authority. I will not spend time discussing this alternative explanation at this point, since this is not primarily a paper on Davidson's positive views. s Since my present concern is only in his criticism of Putnam's handling of the problem that externalism poses for self-knowledge, it is enough to just record that Davidson's alternative account of self-knowledge does not invoke objects of thought, and can therefore live with externalism in a way that Putnam could not.

IV. BURGE'S WAY OF RECONCILING EXTERNALISM WITH SELF-KNOWLEDGE Burge suggests a slightly different way of avoiding the prima facie problem by pointing to a feature of what he calls 'basic cases' of self-knowledge. These are the cases expressed in sincere, first-person, present-tense, judgments about thoughts. In these basic cases, in knowing or thinking that one is thinking that p, one is also thinking that p. This is the feature he then exploits as follows. An agent's thought that p has certain necessary constitutive conditions, and since when one knows that one is thinking that p, one is also thinking that p, it follows that those constitutive conditions carry over to the self-knowledge too. Therefore these constitutive conditions cannot possibly obstruct self-knowledge. If the constitutive conditions are external to the agent, that makes no difference.

238


This point is correct. It assumes self-knowledge and shows that, if one has self-knowledge of one's thought that p, then (in the form of the basic cases, at least), what goes into the individuation of p carries over to the iterated thought that expresses the self-knowledge. It is true that he only has shown it for the basic cases, but that is quite enough to show that, in itself, externalism poses no, in principle, obstacle to self-knowledge. One might protest that his argument assumes that we have selfknowledge, where that is exactly what needs to be shown in the face of his externalism. But the protest is missing the point of Burge'S strategy. He is really claiming that the prima facie objection does not really get off the ground because, given his argument, there is no reason to think that externalism poses a problem for self-knowledge in the first place. One does not have to know all the constitutive conditions that go into the thought that p being the thought it is in order to have that thought. And so if one is unaware of the various external factors that go into its being that thought, it neither follows that one doesn't have that thought nor, given their common necessary conditions, that one doesn't have the iterated thought that expresses selfknowledge of the thought. By this strategy, it looks as if Burge is going to insist that one can never argue directly from externalism to a threat to self-knowledge. So far as I can see, the strategy, as it stands, looks reasonable and anybody who argues simply that Burge's (or Putnam's) externalism leads directly to absence of self-knowledge has not said enough to resist the strategy. All the same, I think there is a common flaw in the way that Davidson and Burge approach Putnam's externalism, and this allows them to run away with the impression that they have in their different ways reconciled it with the fact of first-person authority. The flaw lies in the way that the problem for self-knowledge is posed by them. Let me consider Burge first.

V. WHAT'S WRONG WITH BURGE'S RECONCILIATION? We concluded our exposition of Burge's reconciliation by pointing out that Burge had satisfactorily answered the primajacie objection (from considerations about self-knowledge) to externalism: he did so with his strategy of appealing to the unique feature of his basic cases which blocked any direct route from externalism to a denial of self-knowledge. But all this shows is that the prima facie objection is a badly posed objection, and that one needs to chart a more indirect route from externalism to lack of self-knowledge. One needs to raise the objection in a more roundabout way than is done by the prima facie objection.

239


In order to do so, let's put aside the question of self-knowledge for just a moment and notice once again that Putnam's externalism gives rise to a quite different prima facie problem. The indirect strategy says that selfknowledge is threatened not directly by the fact of externalism but by the fact that there is no way to avoid this other prima facie problem except by surrendering self-knowledge. The other prima facie problem is this. Because (as the twin-earth example above makes clear) Putnam is committed to an externalism that comes from a certain scientific essentialist view of natural-kind terms and concepts, his view is, on the face of it, sometimes going to attribute inconsistent thoughts to agents. It is sometimes going to attribute inconsistent thoughts when the agents are ignorant of the objective natures and the scientific essences, as for example, when an ignorant agent says "I have arthritis in my thigh" or "Water is not H20".9 If such an agent's concept of the disease or the substance is determined by its objective nature, the scientific essence, these predications amount to blatant inconsistency. That is, it is tantamount to attributing to him the belief that he has a disease of the joints only in his thigh (a blatant inconsistency, if we assume, as we may, that he knows that a thigh is not a joint), or attributing to him the belief that a substance with the chemical composition H20 is not H20. But these agents are not logical idiots. However, the only way to avoid the false and uncharitable conclusion that they are logically incompetent is to say that though they believe those things, as these externalisms require, they need not know that they believe them. Their ignorance of the full meaning of the terms or concepts sometimes makes them unaware of their own thoughts in which those concepts figure. And if they don't know that they believe these blatantly inconsistent things then they can't be accused of logical idiocy. But now we have avoided this accusation of attributing logical idiocy at the price of attributing lack of selfknowledge. So: we have arrived at the denial of self-knowledge in a somewhat more indirect way. This indirect way of posing the problem for Burge makes his remarks and his strategy for how externalism can accommodate self-knowledge unsatisfyingly general. They do not address the specific dilemma we have posed for Putnam's specific externalism: either agents must often be falsely accused oflogical idiocy, or, and this brings us to our subject, they must often be denied self-knowledge of their contents (for instance in the examples just mentioned) when intuitively there is no reason to deny it. 10 Actually, strictly speaking it is a trilemma since there is a third option, which is simply to opt for a bifurcation of content into two, one of which is not externalist (in Putnam's sense of externalism) and therefore does not have any problem with self-knowledge. This is, as I said, Putnam's own solution, but, as I also said, it amounts to accepting that self-knowledge cannot be reconciled with his externalism. Since Burge has explicitly resisted accepting this and has

240


denied the wisdom in bifurcation, II we may see the bifurcatory way out of the problems created by the first two horns of the dilemma as the third hom of a trilemma for him. It would be surprising if the case against Burge were so simple. So let me consider various responses one could make on his behalf in order to avoid the trilemma. 12 FIRST RESPONSE

In the reasoning which led to making the charge about inconsistent attributions (the first hom), there was one very important step. This step consisted in rewriting the occurrence of 'arthritis' and 'water' in the inconsistent belief attributions with 'a disease of the joints only' and 'a substance with the chemical composition H 20.' Without this step we could not attribute the blatantly inconsistent thoughts. Blatant inconsistency only arises if we claim that the ignorant agent thinks that it is this sort of substance which is not H20, or this sort of disease which has afflicted his thigh. Now someone may resist this 'rewrite' interpretation of Putnam's externalism, and in a recent paper Burge himself has introduced certain distinctions which someone may invoke in this resistance. 13 Here is the passage with the distinction: I distinguish between a lexical item and the explication of its meaning that articulates what the individual would give, under some reflection, as his understanding of the word. Call the former 'the word' and the latter 'the entry for the word'. I also distinguish between the concept associated with the word and the concept(s) associated with the entry. Call the former 'the concept' and the latter 'the conceptual explication'. Finally, I distinguish between a type of meaning associated with the word, 'translational meaning', and the meaning associated with its entry, 'explicational meaning'. For our purposes, the explicational meaning is the semantical analogue of the conceptual explication. The translational meaning of a word can be articulated through exact translation and sometimes through such trivial thoughts as my word 'tiger' applies to tigers, but need not be exhaustively expressible in other terms in an idiolect. (181)

He goes on to criticize the view (he calls it the "traditional" view) which claims that: ... a word's explicational meaning and its translational meaning are, for purposes of characterizing the individual's idiolect, always interchangeable; and that the individual's conceptual explication always completely exhausts his or her concept. This view is incorrect. It is incorrect because of the role that the referent plays in individuating the concept and translational meaning, and because of the role that non-explicational abilities play in the individual's application of the word and concept.

241


Accounting for a person's lexical entry or conceptual explication is relevant to detennining the nature of a person's meaning or concept. But the two enterprises are not the same. (181)

Now all this is relevant because someone may think that in the crucial step above, my insistence that Putnam's extemalist view amounts to taking the agent's concept of arthritis as involving an attribution of the concept of a disease of the joints only, is a running together of the agent's 'concept' with the 'conceptual explication'-just what Burge warns us against in the traditional view. An inconsistency is only attributed if the agent can exhaustively articulate (recall the definition of 'conceptual explication' in the first quotation) the concept of arthritis, in particular articulate that it is a disease of the joints only, which, being ignorant, he cannot. Thus the rewrite description is not part of his conceptual explication. Clearly since we want to avoid the first horn, that is, clearly since we are in advance convinced by considerations of charity that the agent is not being blatantly inconsistent, we may, armed now with these distinctions, attribute to him only the explicatory portion of the concept of arthritis and thereby ensure that no inconsistency is being attributed. In short, someone may want to defend Putnam against the charge of attributing inconsistencies to agents, such as in the arthritis example above, by saying that it involves a case of the conceptual explication not exhausting the concept. The trouble with this response, however, is ~at it has surreptitiously, that is to say with other words and labels, conceded that there is a second notion of content. Contents are composed of concepts. And now we are being told that there are two notions of concepts: 1) concepts proper, given by externalist reference and 2) concepts in the sense of conceptual explications that the agent can, on reflection, articulate. It is only when the second of these is attributed that an attribution of inconsistency is avoided. Contents composed of the former will often harbor internal inconsistencies when the agent is seriously misinformed about the reference, as in the examples above. The conceptual explication of 'arthritis' in this example of an ignorant agent will consist, say, of the agent's belief that it is a disease, that it is a painful disease, that it occurs more often in older people, etc., but not that it is a disease of the joints only. And it is this conceptual explication that avoids the first horn. This explication, of course, does not get us to the reference of 'arthritis' (in any official sense of 'reference') nor does it coincide with the explication of the experts. It is concepts proper which are tied to reference and to experts' explications. If one attributes contents composed of concepts proper to this agent, as Burge says we must, and if we want to avoid an intentional psychology for him that uncharitably and quite wrongly attributes blatant inconsistencies, we had better also attribute the other kind of content to him. We had better attribute contents with concepts in the sense of conceptual explications.

242


Hence an appeal to these distinctions to save Burge against the first hom succeeds only by accepting a bifurcation of content into two. It can only avoid the first hom by impaling him on the third hom. SECOND RESPONSE

Burge may resist my talk of the explicatory portions of concepts as forming the basis of a second notion of content. He may deny that it is necessary to talk that way in order to avoid the first hom and the charge of attributing inconsistencies. Recall again that the problem about attributions of inconsistent beliefs arose for externalism because we allowed ourselves to rewrite 'arthritis' in the representation ofthe agent's belief with, among other things, 'a disease of the joints only'. Only thus the inconsistency in his thinking, "I have arthritis in my thigh" (assuming still that he knows that the thigh is not a joint). Our justification for doing so was just Putnam's claim that the concept determined in terms of the scientific essences and objective natures of the external items must be attributed to the agent. That rewrite just is the concept which is so determined. It is this substitution or rewrite that the distinctions invoked by the first response were trying to finesse. Here is another, somewhat related, but not identical, reason why the inconsistency-inducing rewrites might be resisted. It might be said that the rewrite is only permissible if one believes in something like the analytic-synthetic distinction. Only if one thought that concepts or terms, such as 'arthritis', have definite meanings which include 'a disease of the joints only', is it permitted. But why, it will be protested, should we hitch this old cart horse of a distinction to a shiny new buggy like Putnam's externalism? Indeed Burge has explicitly denied that he embraces the distinction in another recent paper14 (and Putnam himself has, famously, opposed the distinction for a long time). It is unfair to him, then, to interpret his idea of the external element's constitutive relevance to concepts and contents as the idea that we take any definite belief or set of beliefs defining 'arthritis' and plug them into the representations of individual agents' beliefs whenever we might have said 'arthritis'. But without plugging it in we cannot have the first hom, we cannot have our inconsistent belief attributed. This response is fair enough only if we are told of an alternative way of saying what the agent's notion is than the one which is specified by the substitution of some set of beliefs capturing the objective natures. What, according to Putnam's externalism, are we attributing to the agent when we say that he believes that he has arthritis in his thigh, if it is not, among other things, the belief that he has a disease, which afflicts the joints only, in his thigh? The experts and most other people who are scientifically knowledgeable do, after all, think that that is what arthritis is, i.e., that is what it refers to; so what is left of Putnam's externalism, with its appeal to scientific essences, if the substitution is not allowed? How can we retain the relevance

243


of this external element (the objective natures that the experts are tracking) to this agent's concepts and contents and fail to specify things this way? The only alternative specification of his concept of arthritis, which retains the relevance of the scientific essences, is one which appeals to a very much more general belief of the agent himself, something like the belief: "arthritis is whatever the experts, call 'arthritis'." This metalinguistic specification is a convenient way of bringing in the scientific essentialist element without saying anything specific by way of a definition of 'arthritis', and it thus avoids saying anything that smacks of a commitment to the analytic-synthetic distinction. And because it does not say anything very specific it does not involve the agent in any inconsistency. But notice that this way of retaining the external element retains the wrong thing. The external now enters in a way that abandons Putnam's externalism. It, after all, enters mediated by a belief or description of the agent's, even if the belief is metalinguistically specified. It now is indistinguishable from a view of the external and referential element which is mediated by the descriptions of the agent. It is no longer an externalism based on the causal-theoretic view of reference which is familiar from Putnam. It is interesting to note that if someone, on Burge's behalf, accepted these metalinguistic ways of bringing in the external,IS Putnam's externalism would collapse with a much weaker and much more general thesis than his externalism: the thesis about the linguistic division of labour. This latter thesis plausibly emphasizes the fact that there is no specifying many of our concepts and many of the meanings of our terms except by noting that they are determined in part by our reliance on others in the society in which we live. But the thesis does not amount to Putnam's specific externalism at all since reliance can be captured by beliefs or descriptions of the relying agent-metalinguistically specified-such as the one mentioned in the last paragraph. It does not amount to a referential externalism because it does not attribute the same concept (of, say, arthritis or, to take a famous example from Putnam, of elm) to the relying agent as it does to the expert, the reliedupon agent. The latter's concept, since it does not tum on similar metalinguistically specified beliefs, but on more purely medical or botanical beliefs, is quite different from the relying agent's concept. The reference is no longer crucial in the specification of concepts, it is the differing beliefs or descriptions of the relied-upon and the relying agent which are doing the work, so the concepts attributed to them will be quite different. 16 I conclude then that Burge can only endorse this second line of response by abandoning the externalist thesis in Putnam that the response was supposed to defend. THIRD RESPONSE

It may still seem that I have not allowed Burge the most sympathetic exposition of his position so that he can make Putnam avoid the problem of 244


having to attribute an inconsistency and, thus, avoid the first hom. Someone may deny that the only way for a social extemalist to avoid a commitment to the analytic-synthetic distinction is to rewrite 'arthritis' along the metalinguistic lines suggested in my reply to the second response. Someone might say that there is no reason to plug in for 'arthritis', in the specification of the agent's content, either the metalinguistic rewrite ('whatever the experts call "arthritis"') or the inconsistency-inducing rewrite ('a disease of the joints only'). There is no need for any rewrite. Rather what is needed, it might be said, is simply the concept of arthritis. There is nothing else to say. So to the question: what does he think he has in his thigh, the answer simply is: arthritis. He is able to think this because he thinks, falsely, that one can have arthritis in one's thigh. That the concept is the concept of arthritis (which is a disease one can only have in the joints) allows us to say no more, Burge might say, than something harmless such as "He thinks, ofa disease which one can (in fact) only have in one's joints, that he has it in his thigh." No need to rewrite things at all, neither as the belief that he has a disease which is among other things a disease of the joints only, in his thigh, nor as the belief that he has the disease, which is called 'arthritis' by the experts, in his thigh. This third response, I believe, leaves things with a mystery. One is being left with what is sometimes called a purely 'disquotational' specification of the concept. If I were to persist, well what is it that you are attributing when you say that he has arthritis in his thigh, what does 'arthritis' in the de dicto specification of the contenttell us?-the answer will simply be that "'arthritis' refers to arthritis."17 To the question, what's that?, what does the right hand occurrence of the term convey, the protest will be that I am insisting on definitions, an insistence from the dark ages when the analyticsynthetic distinction was still in currency. Notice, though, that I am asking for what 'arthritis' in the specification tells us, I am not asking for further information about something which has already been conveyed by the disquotational specification. The protest, then, has got to be that even this minimal and initial demand smacks of a commitment to definitions and therefore to the analytic-synthetic distinction. This protest against my rewrite is not the same as another more clearly mistaken protest. It is not the protest which says that I am missing the obvious point that de dicto attributions do not allow of substitutions. Such a protest would itself be missing the obvious point that it cannot be that de dicto attributions allow of no substitutions. The prohibition of substitutions in de dicto attributions tum on the attributee not knowing of the coextensiveness of the expressions involved; and I am precisely protesting that there must be some, at the minimum one, and often more, expressions that can be substituted, or else it is not clear that any concept of arthritis should be attributed to him. And moreover if we admit that 'a disease of the joints only' is not one of the expressions to be substituted for 'arthritis' 245


because the attributee does not know of the co-extensiveness, then, as I said in my reply to the second response, that puts into doubt the relevance of Putnam's externalist concept of arthritis to de dicto attributions. The more serious protest under consideration charges my insistence on rewrites with the more specific charge of being committed to the idea of analyticity. I have no particular wish to defend the analytic-synthetic distinction and toward the end of the paper I will say something to make it absolutely clear that my view does not require any commitment to it. But even on the face of it, it does seem to me quite wrong to think: that if one persists with the question (what is it that the reference-giving assertion '''arthritis' refers to arthritis" is really saying?) one must be doing so because of a desire for definitions. All that the persistence reflects is a dissatisfaction with a trivially conceived disquotational specification of the term 'arthritis' when one is asking for the concept being specified in the content attribution. Disquotation cannot be a merely syntactic device, for if it were, it could never specify the concept of arthritis in content attributions. Indeed it could never specify any meaning of anything at all. Disquotation, if it is to be in the service of an account of meaning, is not a wholly trivial idea. It must be anchored in something which is not made explicit in the disquotational clause itself. The right hand side occurrence of 'arthritis', after all, is a use of the term. It expresses something. And so if disquotation is not a mere syntactic device, one does, if not explicitly at least implicitly,I8 get an answer to the question one is persisting with, the question what is it that one is saying when one says that an agent thinks that he has arthritis in his thigh. And then one wants to know, if the right hand side of the reference-giving statement expresses something, why-for Putnam's externalism-it does not express the inconsistencyinducing 'a disease of the joints only' , since that is what the scientific experts think arthritis is. What remains of his externalism, with its appeal to scientific essences, if that is not what the right hand side occurrence expresses? One answer to this criticism might be that the appeal to the idea that 'arthritis' refers to arthritis in specifying the concept is not an appeal to disquotation as a syntactic device at all. Rather the disquoted term conveys that the term 'arthritis' hooks up with something in the world by a causal relation which is unmediated by any description. It is hooked up with an object, and is not to be elaborated, as I am insisting, by the specification of an object under some beliefs or descriptions, which descriptions can then be plugged in. Here again I must confess, along with many others, that it is hard for me to understand how terms like reference (and truth) can possibly have a theoretical role to play in the study of concepts (and contents) if the right hand sides of the statements attributing reference (and truth-conditions) reduce us

246


to utter ineffability about what the concept is. This way of avoiding the triviality of disquotation as a syntactic device makes things trivial not by making something syntactic out of something semantic, but rather by making semantics ineffable and mysterious. The highly metaphysical nature of the hook-up has the same trivializing effect. Both the syntactic move and this move appealing to such an ulterior metaphysics of reference have the same consequence: they convey no infonnation at all as to what the concept being attributed is. Here is a way of bringing this out more vividly. Nobody can deny that there is, at least, a prima jacie difference between the sentence 'Venneer's View ojthe Delft is beautiful' and 'Venneer's View ojthe Delft is painted on a canvas', between 'Nehru was a good man' and 'Nehru was born in Allahabad'. Nobody can deny that predicates like 'is beautiful' and 'is good' are, primajacie, different from predicates like 'is painted on canvas' and 'is born in Allahabad'. The difference consists in the fact that, at least prima jacie, there is a difference in the way we think of how they stand in relation to concepts like truth and reference. The application of concepts like reference and truth is, at least prima jacie , problematic when we are dealing with evaluative predicates and sentences. No doubt many will eventually want to say that evaluative predicates like 'is good' and 'is beautiful' are susceptible to a naturalistic or an intuitionistic treatment and so there is not a serious difference between them and the other predicates, as far as applying the concept of reference and truth to them goes. But that is something we might eventually say, so it does not spoil the observation that there is aprimajacie difference. That is, it takes a lot of philosophical work to put oneself in an eventual naturalistic or intuitionistic position of being able to say that there is no serious difference. But if one took reference and truth in a disquotational way in either of the two senses we are considering (the syntactic or the highly metaphysical causal hook-up with objects), we will not be able even to acknowledge that there is aprimajacie difference between these two sorts of sentences and predicates. We will not be in a position to acknowledge that there is even a prima jacie problem in thinking of truth-conditions and reference as applying to the evaluative sentences and evaluative predicates respectively. If disquotation is a purely syntactic device then it is indifferent to the primajacie distinction we are marking between these two sorts of sentences and predicates. That is, it merely removes the quotes with no particular care about the meaning of what is inside of them, so it does not care if there are evaluative sentences inside of the quotes, for which 'is true' might, at least primajacie, be inappropriate. Equally, if descriptions and beliefs are wholly irrelevant to the hook up with objects which these disquotational assertions are supposed to convey, then there are no resources to state the problem that evaluative predicates might prima jacie raise for the application of notions

247


like truth and reference. For all we have said and can say, these predicates hook-up with objects (Goodness, Beauty) just like any other predicate, so there cannot even be a prima facie problem of this kind. But we started by saying that there is, undeniably, a prima facie problem; so there is nothing else to conclude but that this highly metaphysical way of thinking of the specification of our agent's concept, "'arthritis' refers to arthritis", is just as unsatisfactory as the syntactic treatment of the disquotation involved in it. It papers over a genuine prima facie problem and distinction. It does not even allow us to raise the problem or to make the distinction in a tentative, preliminary way. I have heard it said that concepts are 'primitive' things about which one can say no more than what this pure notion of disquotation allows. The disquotation does not convey any further explication because if it did it would deny the primitiveness of concepts. But 'primitive' here merely labels what I am finding mysterious. Some may object that surely some concepts have to be primitive. Not all concepts can be such that they get a rewritable explication, because explications themselves invoke other concepts and it would make things circular or infinitely regressive unless some concepts did not get explications. I think this view comes from an altogether uncompulsory foundationalist picture of a bedrock of concepts on which others are founded. No doubt, we hold some concepts steady in the background in order to explicate what others mean in the foreground. But that does not mean that the ones we held steady are primitive in the sense of not themselves susceptible to explication, holding other concepts steady. Illumination about concepts comes from this shifting dialectic between background and explication, and not from some foundationalist bedrock. Someone may be tempted to say that my complaint about how there is a mystery at the heart of the orthodox extemalist's appeal to disquotation is unfair because such a mystery arises for my own view as well. They may say that I just assume that there is no mystery about how the beliefs or descriptions (which form the agent's conception of the disease and which I insist on plugging in for the term 'arthritis') get us the meaning of 'arthritis'. Why should we find agents' conceptions less mysterious than direct unmediated reference? In short, as I have often heard it said, if direct 'reference' is mysterious why is 'sense' any the less mysterious? It is just as mysterious to say that words get meaning via sense as it is to say that they get it via reference. First of all, my view is not committed to any orthodox notion of sense, as will become clear later (see note 29 and the text to which it attaches). But that apart, I would deny that mystery also attaches to my view. What I find mysterious about the claim that there are in principle no beliefs or descriptions which can get substituted for concepts is perfectly easy to state, and state precisely; and, once stated, it is obvious that it is not to be found in my view.

248


Following Wittgenstein, I find mystery in these matters if something is altogether and, by its very nature, inexpressible. Whereof one cannot speak, thereof one is mystery-mongering. By inexpressibility here I don't mean merely that something cannot be explicitly listed because it would take longer than we have, but something which is by its very nature not the sort of thing which is expressible, something for which descriptions are the wrong sort of things to demand. (I admit that there are some very special and unusual contexts in which demonstratives are used which are an exception to my insistence that there be some descriptions available to the speaker; in these cases it is wordless, non-conceptualized skills by which we fasten and focus on particular referents-but I think it is possible to treat these as limiting and degenerate instances, and they do not affect the point I am making against inexpressibility. Being degenerate there is no lesson to be learnt from them for terms like 'arthritis' or 'water' or, for that matter, for all the other contexts in which demonstratives are used.) The principled inexpressibility I am opposing is implied by the appeal to disquotation and primitiveness in the third response. It is said that Wittgenstein himself took a rather glamorous view of the inexpressible at the end of the Tractatus. That mayor may not be so. I don't think we should take such a view. In not allowing that there is anything to be said in answer to the question: what does the right hand side of the occurrence of 'arthritis' in a reference-giving statement convey? the orthodox externalist is being mysterious, by this criterion of what is mysterious. But by the same criterion my insistence on beliefs or descriptions being brought in to answer the question precisely eschews this mysteriousness. Of course, I grant that for any term in any of the descriptions given to answer such questions, more such questions can be raised. And the point is that more such answers appealing to further descriptions will be given. There is no reason to admit to inexpressibility and reduce oneself to silence. Therefore there is no mystery, at least as defined above. It is true that I have traded mystery for other things such as a seemingly infinitely regressive appeal to descriptions. But I have already said that that difficulty is inherent in anti-foundationalism, a difficulty which can be overcome in routine pragmatist ways by tentatively holding some descriptions unquestioned in the background while answering questions about others in the foreground. This is all of course much messier than the view that concepts are primitive and that they must be purely disquotationally characterized. But life is a mess, and theoretical and philosophical reflection on these matters would do well to acknowledge the mess and keep faith with it, rather than produce an artificially tidy theory with a mystery at its very core. I have stated clearly what I mean by 'mysterious' and I have found the view I am opposing-the view that permits no rewrite-guilty of it, in precisely the way that my own view-which demands a rewrite-is not. If someone is going to turn on me and say that I too am being mysterious in

249


insisting on descriptions and beliefs, they must have in mind another sense of what is mysterious. And I don't know what that is, so I cannot answer the charge. I conclude, then, that this third response on behalf of Burge, which tries to deal with the first hom of the dilemma by invoking purely disquotational specifications of concepts, and which thereby avoids any kind of rewrite of the concept of arthritis, won't work either. It avoids the first hom at the cost of making it wholly mysterious what concept, and, therefore, what content is being attributed to the ignorant agent in our trilemma. It remains for me to say something about why my insistence that there must be some form of implicit rewrite in order to specify a concept, does not, at least on my view of intentional content, bring with it any commitment to an analytic-synthetic distinction. I cannot do so right away, since I need to present my own externalist conception of intentionality before I have the resources to do so. I will do that at the end ofthe paper. To sum up this section on Burge: I have criticized his strategy for solving the problem for self-knowledge, raised by externalism, by showing that the strategy is only effective against a much too direct way of raising the problem. I then presented a more indirect way of raising the problem against which his strategy does not work because it is too general. And then I spent some time defending my indirect way of raising the problem against three responses which claimed that raising it that way was unjustified. I now tum to Davidson's way of denying that externalism raises a crippling problem for self-knowledge.

VI. WHAT'S WRONG WITH DAVIDSON'S RECONCILIATION? Davidson, recall, says that Putnam's repudiation of internalism in tum repudiates the idea of objects of thought. And, given an assumption that he says Putnam shares with internalists (that only such objects will account for selfknowledge), he says Putnam gives up on self-knowledge for externally constituted contents. The trouble with the criticism is that it takes Putnam to be saying that a mere denial of internalism is sufficient to give up on first-person authority. But it does not seem to me that Putnam's idea that we may often not know what we believe turns on just simply repudiating the internalist position with its commitment to internal objects of thought. Rather it turns on the specific externalist commitments which flow from his (and Kripke's and Burge's) views on reference and meaning.

250


Putnam is committed to an externalism that comes from a certain scientific essentialist view of natural-kind terms and concepts. And it is this, when it is extended to a thesis about concepts and intentional contents, that gives rise to the problem for self-knowledge. If many of my concepts-say, the concept of water-are fixed by the objective natures of kinds in the environment, then my intentional contents which are composed of these concepts-say, the belief that water will quench thirst-will not be something that I will have (full) self-knowledge of, if I have no knowledge (or only partial knowledge) of their objective natures. That is, if I do not have the appropriate knowledge (of chemistry) in this case, and the chemical facts about water at least partly determine my concept of water, then-if my indirect strategy against Burge is effective-it follows as a consequence that I will at best have only partial knowledge of the contents which contain my concept of water. The assumption about objects of thought accountingfor self-knowledge is irrelevant to this consequence. If Putnam were to join Davidson in denying that objects of thought account for self-knowledge, could he avoid this consequence of his externalist view? I don't see how he can. Here again I must invoke my indirect strategy to show that he cannot. Suppose a chemical ignoramus on Earth believes that water is not H20. It would seem that if we take Putnam's externalist view this person believes in something blatantly inconsistent. But, to repeat, that is an absurdly uncharitable conclusion to come to. A chemically ignorant person is chemically ignorant, not logically deficient. How can Putnam, then, get out of this absurd conclusion that seems to follow from his externalism? Only, I suggest, by saying that this person does not know (or does not know fully) what he believes. If he does not know what he believes then attributing a blatantly inconsistent belief is not absurdly uncharitable. But this has landed Putnam with just the rejection of self-knowledge which we had advertised as unavoidable. No amount of denying that objects of thought account for self-knowledge is going to help Putnam avoid this conclusion. I also think that Putnam's specific externalist commitments better explain why it is that, for Putnam, self-knowledge fails to hold only in some cases (cases where an agent does not know the right chemistry or some other of nature's essences). If one gave Davidson's explanation for why Putnam gives up on self-knowledge, then self-knowledge should fail to hold much more comprehensively than Putnam seems to want to say; more comprehensively, because if one believed that self-knowledge of thoughts was a result of thoughts having inner objects and one also thought that there are no inner objects of thought, then presumably one would not restrict one's denial of self-knowledge to the sorts of cases (involving natural kinds) Putnam discusses. 19

251


The final proof of the point that Putnam's denial of self-knowledge turns on his specific externalist thesis (rather than on an assumption he shares with internalists) will be that one can formulate a specific externalist view of content that is an alternative to Putnam's externalism and which does not deny self-knowledge in this way. I will briefly sketch such an externalism below, an externalism in which the external world constitutes concepts without having any truck with the scientific essentialism or rigid designation that is essential to Putnam's view of certain terms and concepts that go into the specification of contents. My point for now is that there are externalisms which threaten self-knowledge and there are externalisms which do not. Davidson's diagnosis for why an externalist need not abandon selfknowledge does not distinguish between these externalisms and is, as a result, an unsatisfyingly general diagnosis for a problem that has a much more specific source. 20 This unsatisfyingly unspecific diagnosis in Davidson feeds into the similarly unsatisfying positive suggestion he goes on to make about how to make Putnam compatible with self-knowledge, once we give up on objects of thoughts. His suggestion is that there can be no denying the presumption of first-person authority or self-knowledge because without it agents could not be said to be "interpretable at all." In one of his papers,2! he says: When we have freed ourselves from the assumption that thoughts must have mysterious objects, we can see how the fact that mental states as we commonly conceive them are identified in part by their natural history not only fails to touch the internal character of such states or to threaten first person authority; it also opens the way to an explanation of first person authority. The explanation comes with the realization that what a person's words mean depends in the most basic cases on the kinds of external objects and events that have caused the person to hold the words to be applicable; similarly for what the person's thoughts are about. An interpreter of another's words and thoughts must depend on scattered infonnation, fortunate training, and imaginative sunnise in coming to understand the other. The agent herself, however, is not in a position to wonder whether she is generally using her own words to apply to the right objects and events, since whatever she regularly applies them to gives her words the meaning they have and her thoughts the contents they have [my emphasis]. Of course, in any particular case, she may be wrong about what she believes about the world; what is impossible is that she is would be wrong most of the time. The reason is apparent: unless there is a presumption that the speaker knows what she means, i.e., getting her own language right, there would be nothing for an interpreter to interpret. To put the matter another way, nothing could count as someone regularly misapplying her own words. First person authority, the social character of language, and the external determinants of thought and meaning go naturally together, once we give up the

252


myth of the subjective, the idea that thoughts require mental objects. (455-56) This positive suggestion manifestly fails to answer the difficulty that Putnam's externalism poses for self-knowledge, because Davidson's description of the difficulty ushers out the real underlying source of the difficulty. Because Davidson nowhere mentions the relevance of the specific sorts of referential commitments in Putnam's externalism, the question he addresses in his positive explanation is something quite else than the question Putnam's externalism generates. Davidson's positive remarks answer a question that has nothing specifically to do with Putnam's externalism: what, in general, explains the undeniable fact that agents whom we are interpreting by and large have non-inferential self-knowledge (first-person authority) of their own thoughts, given that in our interpretations we are not specifying objects of thought within their epistemological ken, but looking instead to external objects in their environments? And it answers this question, as the passage I emphasized makes clear, by saying that if there are going to be meanings for the interpreter to interpret it is only because the interpretee knows her own thoughts in ways that are quite different from the way the interpreter comes to know them, i.e., directly and not via interpretation. Putting aside for a moment whether this answers the new question adequately, the relevant point for now is that the question itself, though it does focus on the interpreter and the agent's environment rather than on the internal objects ofthe agent's thought, is nevertheless a quite different question from the initial question we were interested in (and which Davidson himself posed), which was: how is self-knowledge compatible with Putnam's externalism? Notice that this initial question is not particularly about the non-inferential nature of selfknowledge. It is about whether one has full self-knowledge (whether inferential or non-inferential is not the point) of what one thinks, even when, as in the case of the chemically or medically ignorant agent, one's thoughts are composed of concepts (arthritis, water) which one may not have full knowledge of.22 I think that Davidson is under the impression that because his explanation is given from the point of view of what even the third person or interpreter must acknowledge about the first person or interpretee's authority over his own states, it is an explanation which does answer the initial question. But Putnam's externalism-that is, (O.E.)-is a much more specific doctrine than an externalism which says that contents are constituted by the deliverances of a third person or interpreter who looks to an agent's environment. Putnam's externalism, in the initial question we are interested in, claims that the interpreter makes essential appeal to the objective natures of natural kinds in the agent's external environment as constituting his contents. Davidson's positive explanation of self-knowledge makes no mention of this 253


appeal at all. The question which his positive explanation does address is a more general question: how can any kind of extemalist--anyone who gives up on internalism and thereby believes in something very general like (G.E.)-avoid making self-knowledge inferential from knowledge of the external world and, therefore, how can he avoid giving up on the special noninferential authority that attaches to self-knowledge? In other words, how can we show that just because what we think is constituted partly by external things, it does not follow that we know what we think partly by way of inferences from knowledge of those external things. But Putnam's externalism is, as I have been saying, much more specific and much less minimal than (G.E.) and it raises instead our initial question which has no particular focus on the non-inferential character of self-knowledge. Davidson offers no help with this question, and that is not surprising because, as my indirect strategy demonstrates, once one adopts Putnam's externalism there is no help against the threat to self-knowledge (except for the unwelcome help offered by the other two horns of the trilemma). I have tried to display how both Burge and Davidson's efforts at dealing with this initial question, how Burge and Davidson's reconciliation between Putnam's externalism and self-knowledge, fail to see the precise way in which Putnam's externalism makes it natural and right for Putnam to abandon self-knowledge (of wide, or externally constituted, contents). If I'm right, anyone who adopts Putnam's externalism has no way to avoid the threat to self-knowledge. But now the question remains whether it is compulsory for all externalists,-i.e., all those who subscribe to (G.E.), which is Putnam's own general characterization of externalism,-to adopt Putnam's specific externalist proposal which has such widespread currency today. That is, is it compulsory for them to adopt (O.E.)? In the next and last section I will deny that that is compulsory by sketching an externalism about intentional content which is quite different from (O.E.). My externalism raises no problem, not even prima facie, about self-knowledge.

VII. HOW TO RECONCILE EXTERNALISM WITH SELF-KNOWLEDGE The challenge to be met is: how shall we appeal to the external elements in an agent's environment in the determination of his intentional contents without giving up on the natural and intuitive assumption that he or she has knowledge of those contents? My answer, for reasons of space, will have to be very crude and smudged, and I will have to leave the refinements and qualifications to another place. 23 In my externalist account, this challenge is met by imposing a crucial constraint on the way the externalist element

254


detennines content and, I claim, that there is no chance that the challenge can be met without constraining the external element in this way. The constraint can be stated thus: (C): When fixing an externally detennined concept of an agent, one must do so by looking to indexically fonnulated utterances of the agent which express indexical contents containing that concept and then picking that external detenninant of the concept which is in consonance with other contents that have been fixed for the agent.

The emphasis in (C) is the heart of the constraint but some other rather basic things need to be clarified before I come to it. First of all let me say at the very outset that though I formulate the .constraint and the whole question of externalism in the context of how we fix an agent's concepts and contents, this does not mean that I am interested only or even primarily in the epistemological question: how do we find out about another's concepts and contents? That would be to get things the wrong way round. If we assume that meanings and concepts and intentional contents are public then we may assume that whatever it is that goes into their constitution is publicly available, i.e., available to another. So looking at how these are fixed by another does not by any means amount to a mere question about how we find about them as opposed to a question about what is their nature. It is because their nature is such that it is partly constituted by external things that this external constitution should emerge in the answering of the question: what goes into the fixing of concepts by another? So one should not be misled by my talk of concept-fixing and content-attribution to get things the wrong way round. Second, let me say what is meant by 'concepts'. Assuming a certain close relation between intentionality and meaning, I use 'concept' to talk of the counterpart to 'term' in just the way that 'content' is thought of as the counterpart to 'sentence'. Just as the content of an agent's belief that arthritis is painful is cashed out in terms of the meaning of the sentence 'arthritis is painful' , so also the concept of arthritis which, in part, composes that content is cashed out in terms of the meaning of the term 'arthritis' which, in part, composes that sentence. In assuming this close relation between intentionality and meaning I follow most others who writ~ on this subject.24 Third, the fixing of externally detennined concepts takes place, as one might expect, by looking at the indexical contents (and the utterances expressing them) in which they occur. One should expect this because it is indexical contents and utterances ('It's cold', 'Here comes Maggie', 'That's a Bosendorfer', 'This is a gazelle') which can be correlated most obviously with saliencies in the external world that are supposed to be detennining the concepts they contain. It is with these that one can most easily identify the external detenninants. Indexical contents and utterances are, therefore, the

255


points of entry into another's mind, since they are the first clear clues to the external (and public) sources of another's meanings. One fixes concepts first by correlating them, or at any rate, by correlating the terms in the utterances in which they are expressed, with items in the environment of the agent who thinks and expresses them. Thus these correlations consist in finding some sort of salient item regularly present in the environment when indexical utterances with a given term expressing some concept are uttered. The centrality of indexical utterances and thoughts, however, should not give the impression of any commitment to anything like 'direct reference'. Indexicals are central to externalism because they are essential clues to an agent's perceptions of and responses to things and events around him. This centrality is not lost if it is unaccompanied by doctrines of direct reference. Fourth, clearly one has to distinguish between those indexical utterances which are sincere and literal and those which are not, else we will not correlate them with the right item in the environment or perhaps we will not find any sort of item regularly present. Only the former must be correlated with external saliencies. If I were uttering a lie or a metaphor when I said 'He's a beast' the correlated external salience would not help much in fixing the concept correctly. There is, however, no algorithm for sifting out the sincere and literal utterances from the lies, the metaphors, the stage performances, etc. Nor is there any reason to assume that a concept or term occurs more frequently in sincere and literal utterances than in this vast variety of other uses. But one may assume that the lies and metaphors and so on, even if more frequent, will bring no single item in the environment to the fore since they will be used with an indefinite variety of motives; so over time if there is a correlation at all with something in the environment then we can proceed on the assumption that we had got hold of the sincere utterances and, therefore, have the right external determinant. All this is, of course, a dynamic process of theory building (of concept attribution or fixing) and as with all theories, one may revise earlier attributions if they tum out to fail to provide a coherent theoretical picture due to having fastened on some non-literal or insincere indexical utterances in the early phases. Fifth, obviously, not all concepts will be externally determined. Not, for example, the concept of a unicorn nor the concept of the number five. But these concepts will in one way or another be related to others which will be determined externally. In the case of the concept of the unicorn, it will be related by composition to other concepts which are externally determinedthe concept of a horse, the concept of a hom, etc.; not so in the case of the concept of the number five, which will not be composed of externally determined concepts but rather related to externally determined concepts by more complicated relations. Finally, the heart of the externalism, the constraint itself. What makes one an externalist at all is that in looking to fix concepts one

256


looks to correlations with external things. But that is just externalism in the very general sense defined as (G.E.). My constraint on externalism gives rise to a more specific version of this externalism. It asserts that in selecting the item in the environment which is supposed to fix the concept that is being expressed, one has not only to pick what is salient and is correlated with that term but one has to be very careful how one describes this external determinant of the concept. One has to describe it in a way that fits in with the other contents one has attributed to the agent. Expressions such as 'fits in with' or, as in the original formulation, 'in consonance with' are vague, so let me give a sharper intuitive sense to what I have in mind, by way of an example. Finding the right gross external item to correlate an agent's concept with requires no more than a) shared similarity standards, presumably wired into us all, so that what is grossly salient to him is not wholly at odds with what is salient to us, and b) Mill's methods. These two things, however, will only get us the relevant gross and regular saliencies in the environment and no more. But it is the selection, not of the regular gross salience in the environment, but the right description of it, which the constraint is meant to address; and its claim is that the right description turns on looking to other beliefs attributed to the agent. So, to give a very crude example, let us suppose that, by an application of Mill's methods and by shared similarity standards, we have noticed that there is a pedal regularly present under the left foot of an agent when he utters indexical utterances with a certain term. Then the constraint says not to describe this external determinant of his concept expressed by that term as a 'clutch' if one has no confidence that he has other beliefs, however rudimentary, regarding the inner workings of an automobile. If, by this method in which external items determine concepts, one could not describe the determinant that way, one would not end up attributing the determined concept of a clutch to the agent and so no contents relating to clutches. For some agents we may find ourselves, for similar sorts of reasons, withholding from the external determinant even the description 'pedal' (i.e., in such a case the regular gross salience will not even get the initial description 'pedal') with similar consequences for concepts and contents. Notice that there is a strong element of anti-foundationalism built into this constraint. One may have thought otherwise, misled by my insistence that there is something basic about the indexical utterances which correlate with saliencies in the environment. One may have thought that this insistence on seeing the indexical utterances as a starting point makes for an externalist foundationalism. 25 The thought is not intended at all and the constraint makes that clear. Even for this basic level, this ground floor of indexically formulated utterances and beliefs by which one enters another's mind, the external determinants we pick out and correlate them with have to be

257


described only as they cohere with the other beliefs. Hence the method at the ground floor is a convenient starting point in the building of a theoretical structure of another's beliefs and concepts, but it is built like Neurath's boat, and conclusions about the ground floor are just as mediated by belief, and just as revisable as we proceed with the theorizing if they will not lend themselves to a coherent theoretical picture. There are no Archimedean points in this externalism. Some links with the external world may be more direct than others, i.e., less mediated by surrounding beliefs than others-for instance those that determine the concept of a pedal rather than those that determine the concept of a clutch. But there are no unmediated causal links. There is no getting away from this mediation and the appeal to external causality is not intended to stop or interrupt a holistic (by which I only mean anti-foundationalist26) pattern. Anti-foundationalist holism and externalism are both absolutely essential features of content and they are perfectly compatible. My constraint's appeal to the agents' other beliefs in fastening on an appropriate description of the external determinant is bound to give the impression that I am, after all, an internalist. For it will seem that I am insisting on an internalist filter upon the external. This impression would be quite wrong. There is, to begin with, something misleading, in fact downright false, in thinking of the filter as internal since the belief contents of an agent which provide the filter will contain concepts, which are themselves externally determined. There is nothing internal about the filter at all, if internal is defined in contrast with externally constituted content. But, that apart, I think the impression comes from a tedious and misguided oscillation between a false pair of choices: a highly direct externalism and internalism. The plain fact is that if we take the definition of externalism with which we we started at the beginning of the paper, i.e., (G.E.), seriously and if we refuse to fall for the uncritical conflation of externalism, generally defined, with (O.E.), then any position which is incompatible with the scenario entertained in Descartes's First Meditation is 'externalist' in a perfectly clear sense of that term. And my position is certainly incompatible with that scenario since it explicitly entails a denial of the claim that one could have the thoughts one has if there were no external world. That is, it explicitly denies something entailed by that scenario. In short my position is not internalist because it satisfies the minimal definition for externalism in (G.E). There is one more crucial element in my overall account of content which needs to be briefly summarized. I will call it the 'locality' of content. It is an element forced by one of the effects of my constraint. As I described it, the constraint requires that no concept be attributed along externalist lines, unless it is in consonance with the other contents of the agent to whom it is being attributed. This leads to a rather alarmingly fine-grained concept.

258


After all, the beliefs of an agent that are relevant to (and that he associates with) the concept or term 'clutch' or 'water' or 'arthritis' or 'lemon' might be very numerous and very diverse. In my picture of things this does not matter. The reason why it does not matter is that these attributions of concepts are attributions of things that do not go directly into the attribution of specific contents to explain behavior. All that these attributions do is to provide a pool of resources which one uses in a selective way in order to attribute specific contents in the explanation of behavior. Let me explain with an example. Suppose you know a fair amount of chemistry and I know none, i.e., you have various chemical beliefs and I don't. The concept of water attributed to you in accord with the constraint requires that it be different from the concept attributed to me. Your beliefs which constrain the description of the external substance with which your 'water' -utterances are correlated will be different from mine since you have beliefs about water's chemical composition which I do not. The case is exactly like the one about the concept of clutch and pedal for two agents who respectively do and do not have beliefs about the workings of automobiles. At this level-what we might call the 'aggregative' level of concept attribution-it is unlikely that any two people will have the same concept of anything since it is unlikely that they have all the same beliefs associated with the term which expresses that concept. This is the level at which theories of meaning do their work: they specify the concepts or the term-meanings of an agent along the lines of this constrained externalist method. At the meaning-theoretic level, as I've been saying, the concepts are very finegrained and they are hardly ~ver shared by people. But this does not matter since it is not these concepts, so thought of, which go into the contents that explain action. Action explanation always takes place at another level than this meaning-theoretic level. It takes place at a much more local level. Here the entire aggregate of beliefs that an agent associates with 'water' are not all relevant. One distills out of the aggregate of resources provided by the meaning theory only those beliefs that are relevant to the action explanation at the local level. Thus if you and I are both drinking some substance from the kitchen tap because we want to quench our thirst with the cheapest available drink, we may in this locality both be attributed the same content: " ... that water will quench thirst". In this locality of explanation, your chemical beliefs are simply not among the beliefs selected from the specifications for 'water' in the meaning theory (for your idiolect) for use in the local explanation of your behavior. That is to say, the local concept of water, which goes into the specification of the content which explains your behavior in this locality, is to be thought of as a small selection of beliefs from the aggregate of beliefs you associate with 'water'. In the locality, one selects only what is needed to make the explanation; and

259


the chemical beliefs are not needed for the concept of water in your belief that water quenches thirst, which explains your drinking water. Thus the whole aggregate of beliefs that makes up a concept at the meaning-theoretic level is usually not needed in localities, only selections from them which make up local concepts are needed. So although our idiolects are never likely to be the same for any single concept, in many localities-for instance in the locality being considered in the example-we may nevertheless share many concepts, and therefore contents. There will obviously be other localities in which we will not share contents because in those localities we will find it necessary to use your chemical beliefs from the overall pool in order to capture your local concept of water which composes the content which explains your action. I call this thesis the 'locality' of content. (It should be obvious to the reader that one of the implications of this thesis is that the great importance that has been give to the 'theory of meaning' in the last few decades is highly exaggerated. Content is attributed and behavior is explained only in localities, and the aggregative deliverances of a theory of meaning have no direct role to play in them. 27 The deliverances of such theories, then, have no psychological reality since the concepts they specify do not pull their weight in any explanation of an agent's behavior. Only local concepts compose contents, so only local concepts pull their weight in explanations. The meaning-theoretic specifications have no other function than to summarize aggregates of beliefs associated with each concept, from which to make local selections of concepts.) It should be obvious from this brief sketch why my externalism does not entail a threat to self-knowledge in the way that orthodox externalism does, so I will be very brief. First of all notice how my constraint immediately distinguishes my externalism from those I reject. Putnam and Kripke would look to the external paradigmatic samples with their 'objective', scientifically revealed, 'natures' (the right chemical composition for 'water', the right DNA for 'lemon' ... ) in fastening on the right external determinant; Burge would look at the external determinant as it is mediated by the beliefs of the expert in that community; but my constraint instead insists that we look to the external determinant only as it is mediated by the beliefs of the agent whose concepts are being fixed. For my externalism, a concept is not fixed by correlating it with an external item as God sees that item, or as the expert in given societies sees it, or as the expert at the end of inquiry sees it, or even as the radical interpreter sees it. It is quite wrong to think that externalism must adopt any of these ulterior methods, since to do so entails among other things an insuperable problem for self-knowledge which can only be solved by Putnam's bifurcation of content, i.e., by the introduction of a supplementary internalist notion of content to handle the problem of self-knowledge which

260


such ulterior notions manifestly cannot handle. Without surrender to internalism, my constraint does not allow this problem to arise in the first place. To see this clearly, think again of my indirect strategy for raising the problem for self-knowledge against Burge's and Davidson's efforts at reconciling Putnam's externalism with self-knowledge. Self-knowledge, by this indirect strategy, was only under threat because threatening it is the only way which will allow Putnam's externalism to refrain from attributing certain blatant inconsistencies to agents. But my externalism, unlike these others, has no problems with inconsistent attributions in the first place. There is no first hom, so the trilemrna 1 raised for orthodox externalisms like Putnam's never gets ging for my view. My constraint sees to it that external items which determine concepts do not determine concepts that are at such odds with an agent's other beliefs that she will fall into the situation of uttering or thinking inconsistent thoughts just on the basis of the concepts attributed to her. Agents, on my view, may think thoughts that we specify as 'water is not H20' or 'I have arthritis in my thigh', but the concept of water or arthritis in these cases will not be determined by the experts' beliefs or by scientific essences. The sorts of inconsistency, which follow upon these other externalist views of concepts, therefore, are simply not entailed if one applies my constraint. If the agent lacks certain chemical or medical beliefs he will not be attributed the same concept of arthritis or water as the society's and its experts. Thus if he goes on to say things like 'I have arthritis in my thigh', etc., this will not amount, even primajacie, to inconsistency. My constraint has the effect of bringing in the external determining item under descriptions, or more properly under beliefs, of the agent. We have already seen in countering the second response made on behalf of Burge to the trilemrna 1 posed for him that if the external items (in that case it was the social external items, i.e., the community's expert's knowledge) enter routed through an ignorant agent's own beliefs (specified, in that case, by metalinguistic specifications such as "water is whatever the experts, call 'water"') then the threat of inconsistent attributions to the ignorant agent and the eventual threat of attributing absence of self-knowledge to him is avoided. The social enters into the concept without losing respect for the fact that the relevant individual agent himself knows very little about the chemical composition of various substances. Where the external items are not taken to be social, then the beliefs will not be metalinguistically specified, they will just be ordinary beliefs of the chemically ignorant agent, beliefs such as, for example, "water is the substance that comes out of the kitchen tap", etc. The point is that for chemically ignorant agents the concept of water will not be determined by an external substance under any chemical descriptions or beliefs. So if such an agent were to brashly think that water is not H 20, he would not be thinking something inconsistent. If he is not thinking something inconsistent, then the first hom in the indirect strategy

261


which leads eventually, in the second hom, to a threat to self-knowledge, is unavailable. Neither, therefore, is the second hom available. There just is no threat to self-knowledge. This way of avoiding the threat to self-knowledge within externalism is, of course, very different from Burge's and Davidson's. It is different because it addresses a problem that arises specifically for orthodox externalism in a way that their solutions do not. They run away with the impression that they have avoided the threat because in discussing self-knowledge they raise the threat in a way that does not bring out why the specific orthodox externalist position really gives rise to the threat. Once one sees via the indirect strategy why orthodox externalism gives rise to the threat in a very specific way, then only does one see clearly why my constraint on externalism alone will provide for the compatibility of externalism with self-knowledge. Only one issue remains. Does my constraint's insistence that beliefs of an agent must enter into the constitutive work the external items do smack of definitions and analyticity. This was an issue that I promised to deal with in my reply to the third response on behalf of Burge. And quite apart from that response, any insistence that concepts are to be conceived in terms of descriptions or beliefs must address the charge of a commitment to analyticity. The so-called 'cluster' version of the descriptive theory of terms was an early response to a roughly similar charge made in a slightly different setting. But my response is quite different. My response makes vital use of what I just called the thesis of the locality of content, which is an essential aspect of the overall externalist conception of intentionality that I am offering as an alternative to orthodox externalism. The charge of analyticity was made against my claim that Putnam's externalism is obliged to rewrite some description such as 'the disease which afflicts joints only' or 'the substance which has the chemical composition H20,' for 'arthritis' and 'water' respectively in various judgements (with those terms or concepts) that individuals make in our physical and social environments. The charge was that to insist that there be such a rewrite was to hanker for analytic definitions of these terms. My insistence on the rewrite flowed merely from the fact that I think that the alternative which denies all rewrites leaves unexplained and mysterious what the concept in question is. Now, it is only if one holds Putnam's externalism with its view of natural kinds and their scientific essences (or if one holds Burge's social externalism) that the insistence on the rewrite looks as if it will lead to definitions, and therefore to what looks like a commitment to the analytic-synthetic distinction. This is because on Burge's view, or on the scientific-essentialist view, there is some privileged belief or beliefs of the expert, or some paradigm instance of a natural kind, which fixes the reference and meaning of agents' concepts and terms. Thus I am claiming that the insistence on the rewrite (an insistence forced by the mystery attaching

262


to any :ie~ ~hat .denies. it) by i.tself does n~t co~t one to the anal)tic, synthetIc dIstmctIOn. It IS only If one combmes the msistence with certai accounts of concepts or the meaning of terms that one is cOmmitted to ana,I\. .. 28 IytICIty. If instead one combines the insistence on rewrites with an external'l account committed to the locality thesis like the one Ijust sketched, there :: no such commitment. For me a concept, at the level at which concepts ar; attributed by a theory of meaning, is nothing but an indiscriminate an~ aggregative carrier of all the beliefs an agent associates with the tem which expresses the concept. There are no weights placed on some beliefs overoth_ ers which might give the impression of definitional status or of the notion of criteria (versus merely symptom) or anything like that. There is, ofcours{; another level over and above the meaning-theoretic level, the local level. Bu~ the whole point of distinguishing between the aggregative, meaning_ theoretic level and the local level was to allow that there can be lots of different localities at the local level. This means that there is no single rewrite because there is no single locality. There is, therefore, no definition. Different localities, different explanatory contexts, dictate different selections or distillations from these aggregates, but there is no fixed concept, no canonical selection or distillation, at the local level. Different localities wi1l distill Out different beliefs and thus sanction different rewrites. Though there might be overlap and coincidence in what is distilled in different localities there is no saying in advance what these are. Thus there are no weights imposed on any of the beliefs within the aggregate of beliefs at the meaning-theoretic level such that some beliefs are more important than others and must be distilled out in all localities. So the rewrites I insist on will have no fixed criteria, if one adopts my picture of concepts and contents. They cannot, therefore, count as a commitment to analyticity.29 By contrast, once one forces the need for rewrites, as I have via my charge of mystery, the orthodox accounts I oppose are bound to provide the more rigid definitional rewrites which give rise to analyticity. Given their accounts, the rewrites will be canonical~.g., the scientific essences, the expert's beliefs, etc. No wonder they worry about and resist having to allow rewrites. So, my response to those who charge my demand for a rewrite with a commitment to analyticity is simply this. There is nothing in the demand itself which makes that commitment. The demand merely flows from the inability to specify what a concept is if one did not meet the demand. Only if one were already given to certain specific views on meaning and concepts. would the commitment be made with the rewrites. My view of meaning and concepts makes no such commitment, but the view of rneaning of the exterf nalisms I am opposing, given the need to rewrite, does. 0 way show-b~ to was first The I have had two aims in this paper. 263


defence of Putnam against Burge and Davidson-that Putnam's specific externalism (and others like it) which I called orthodox externalism, pose an insuperable problem for self-knowledge. The second was to show that Putnam's specific externalism was not the only specific proposal which satisfied his general characterization of externalist doctrine, which I called (G.E.). I offered an alternative specific proposal which posed no problem for self-knowledge. I have restricted myself to the question of self-knowledge and externalism and tried to show how Putnam's externalism cannot be reconciled with self-knowledge, while the externalism I offer can. That in itself does not amount to a decisive repudiation of Putnam's externalism since there are many other things that motivate philosophers to adopt externalism which I have not considered here and which might favor Putnam's more orthodox externalist view. So various questions remain: what are the motivations that have led Putnam and others to adopt orthodox externalism, are these good motivations, and if good can they not also be fulfilled by my own externalism?30 These questions must await an answer elsewhere.3l

NOTES 1. H. Putnam, "The Meaning of Meaning" in Language, Mind and Knowledge: Minnesota Studies in the Philosophy of Science 7, K. Gunderson, ed. (Minneapolis: University of Minnesota Press, 1975). 2. See A. Bilgrarni, Belief and Meaning (Oxford: Blackwell, 1992), chapter 1. See also A. Bilgrami, "An Externalist Account of Psychological Content," Philosophical Topics 15 (1987). 3. Putnam's externalism is not the only orthodox externalism. Burge in T. Burge, "Individualism and the Mental" in Midwest Studies in Philosophy 6: Studies in Metaphysics, P. French, T. Uehling, H. Wettstein, eds. (Minneapolis: University of Minnesota Press, 1975), has given a more socially oriented version of Putnam's views of reference, and drawn consequences for a more socially oriented externalist view of intentional content. Fodor in chapter 4 of 1. Fodor, Psychosemantics (Cambridge, Mass: MIT Press, 1987) has given a more information-theoretic version of externalism. Despite these differences between them, all these externalisms raise at least a prima facie problem for self-knowledge. In this and other respects they all stand in contrast to the externalism I propose in the last section of this paper. There are other externalists, such as McDowell and Davidson, whose externalisms I oppose because I believe they too contain an implicit and hidden bifurcation of content. But I will not discuss them here since they cannot really be classified with the orthodox externalisms I am primarily discussing. I discuss McDowell's and Davidson's externalisms in detail in chapter 4 of Belief and Meaning. 4. See D. Davidson, "Knowing One's Own Mind," Proceedings and Addresses of the American Philosophical Association 60 (1967). Also "What is Present to the Mind" in Consciousness E. Villanuova, ed. (New Jersey: Rowman and Littlefield, 1991). 5. See T. Burge, "Individualism and Self-Knowledge," Journal of Philosophy 85 (1988). 6. See the caveat about Evans and McDowell in the next note. For Evans's Russellian view see G. Evans, Varieties of Reference (Oxford: Oxford University Press, 1982). See also 1. McDowell, "De Re Senses" Philosophical Quarterly 34 (1984) and J. Searle, Intentionality (Cambridge, England: Cambridge University Press, 1983).

264


7. We need to record here that the case of Evans's and McDowell's commitment to external objects of thought is a very exceptional case. But for them, the objects of thoughts idea is an internalist idea. Most externalists who follow Putnam do give up on objects of thought for their externalist contents, and it is they that Davidson has in mind when he says that to adopt externalism is to give up on objects of thought. 8. I have written on the subject of objects of thought, and on Davidson's views on it, in "Objects of Thought" in Consciousness, E.Villanuova, ed. (New Jersey: Rowman and Littlefield, 1991). 9. It must be admitted that arthritis will not be counted a natural kind in the blue-chip sense that water is; it will seem to be somewhat more constructed. Since this difference does not affect the point I am making in any really deep way I shall ignore it. 10. For my reason for saying that it is intuitive not to deny it in these cases, see the point above of the foolishness and unintuitiveness of denying it unless psychological obstructions are involved. II. Burge denies it by denying the need or the coherence of the idea of a second notion of internal or narrow content in his "Individualism and Psychology," Philosophical Review 95 (1986). 12. In earlier papers such as "Comments on Loar," Contents of Thought (Proceedings of the Oberlin Colloquium in Philosophy, 1985) R. H. Grimm and D. D. Merrill, eds. (Tucson: University of Arizona Press), and also in (Bilgrami 1987), op. cit., I had raised the problem for self-knowledge via this indirect strategy but had failed to fortify it with replies to the responses which follow. Brian Loar in "Social Content and Psychological Content," the paper I was commenting on in (Bilgrami 1985), op. cit., offers a different indirect strategy which implicitly raises a very similar problem for self-knowledge. By invoking the intentional states of Kripke' s Pierre who travels from Paris to London with a different conception of London in each place, he shows how inferences involving relations between these states which involve the different conceptions of London can failed to be made by him. Given cenain intuitive connections between self-knowledge and inference (for which see note 10), we can easily raise a similar problem about self-knowledge, though Loar himself does not. 13. T. Burge, "Wherein is Language Social?" in Reflections on Chomsky, A. George, ed. (Oxford: Basil Blackwell, 1989). 14. T. Burge, "Intellectual Norms and Foundations of Mind," Journal of Philosophy 83 (1986). 15. Burge himself does not accept this metalinguistic ploy. In (Burge 1979), op. cit., he very explicitly rejects these metalinguistic specifications (96-97) as a way of avoiding his externalist conclusion. 16. It is interesting that Putnam who first formulated the thesis of the 'linguistic division of labour' did not notice and still has not claimed that the thesis can be given this reading, which makes it much weaker than his externalism and Burge's anti-individualism. For more on this way of accommodating reliance and the division of linguistic labour with individualism about content, see (Bilgrami 1987), op. cit. 17. Recall that this is rather like what Burge explicitly says about 'tiger' when he is talking of the 'translational' concept in the first long quotation cited during the discussion of the first response. 18. The implication here is that disquotational truth may be exploited in meaning specification only if somewhere implicitly in the overall theory in which such specifications are lodged, there is something being conveyed about what the right hand side of disquotational clauses express. In Davidson's use of such clauses in his truth-theoretic conception of meaning, for instance, this implicit information emerges in the recesses of the radical interpretation process. 19. Putnam casts his externalist net a little wider than I am indicating here. It is more than natural-kind terms that will raise a problem for self-knowledge. Since he shares Kripke's causal views of reference about proper names, judgments with proper names or singular concepts could raise a similar problem. Burge casts the net even wider by bringing in a

265


20.

21. 22.

23. 24.

25.

26. 27.

28.

266

social externalism over and above a scientific essentialist one, thereby including terms and concepts for artificial kinds as well, such as 'sofa', 'aluminium', etc. This is a common flaw in discussions of this subject. See, for instance, Crispin Wright in "Wittgenstein's Later Philosophy of Mind: Sensation, Privacy and Intention," Journal of Philosophy 86 (1989): 630, especially his long footnote 6. Here Wright moves unconsciously from a discussion of the specific externalisms of Putnam and Burge to remarks about why externalism need not threaten self-knowledge, remarks which talk much more generally about externalism rather than about Putnam and Burge's specific externalist views. A genuine defense of their externalisms against the charge that they threaten selfknowledge must take up their specific externalisms in detail and must respond to them. See Davidson (1987), op. cit. Sometimes the term 'first-person authority' is restricted to describing the special noninferential character of much of our self-knowledge, and it may be wise therefore not to use it as roughly equivalent to the term 'self-knowledge' as I have been doing in this paper. My only excuse for doing so is that this paper is not primarily concerned with the problem of reconciling the non-inferential nature of self-knowledge with externalism. This paper is primarily concerned rather with a prior and harder problem raised by the indirect strategy, which is the problem of reconciling self-knowledge at all with externalism. The indirect strategy makes it clear that the latter reconciliation must be had first before the question of the former reconciliation even arises, because the strategy puts into doubt that we know what we think at all (whether inferentially or non-inferentially). Philosophers who have discussed the question of self-knowledge and externalism have failed to keep these two issues apart. And, in partiCUlar, I'm saying that Davidson, who raises both issues, fails to keep them apart and fails to see that his positive remarks address only the former issue, not the latter. I'm not necessarily denying that Davidson (and Burge), if they are allowed to pretend that there is no prior and harder problem raised by the indirect strategy, would have given satisfactory solutions to the less hard problem about noninferentiality. I'm only insisting that they have not correctly addressed the prior and harder problem. I have discussed the question of non-inferentiality of self-knowledge in the appendix to (Bilgrarni 1992). See (Bilgrarni 1992), op. cit. The close relation is not intended to convey the obvious falsehood that an agent's utterances of sentences always express the contents that the meanings ofthose sentences cash out. But the close relation is conveyed by the fact that a sincere, non-self-deceived etc., utterance of (or assent to) a sentence by an agent is an utterance of something whose meaning gives the content of the intentional state that is expressed by that utterance. The fact that sentences are often not asserted this way does not spoil the connection, though it obvi0usly gives those who believe in the connection the task of producing an appropriately nuanced formulation of the connection. The impression may come from such superficial things as my use of expressions like 'ground floor' and 'points of entry into another's mind' , and less superficially from the fact that more traditional internalist foundationalist views, such as Russell's Logical Atomism, also stress the indexical utterances, but with an accompanying stress on correlations with inner experiential items rather than with external items in the environment. I stress this qualification to distinguish the holism I intend by my constraint from other well-known doctrines such as Davidson's. In case such a suspicion has been created, this should make it clear that I am not surreptitiously committing myselfto two notions of content, one local and the other aggregative or meaning-theoretic. As Ijust said, there is only one notion of content and that occurs at the local level. There are, to be sure, two notions of concepts, but one of them does not play any role in contents and therefore, as I say below, has no psychological reality. It is merely an aggregate pool of resources from which to select concepts which do playa role in composing contents and which therefore do have psychological reality. Actually, even if Burge denies that he is privileging the expert's belief over the other beliefs, the fact is that so long as the expert's belief is one among the several beliefs that


go into something like a 'cluster' rewrite, the inconsistency~f the first horn-will still get attributed to the agent who says or thinks "I have arthritis in my thigh." 29. It should be clear from all this that I am not committed to the notion of sense in any standard conception of the term, not the Fregean conception, nor the cluster-theoretic theoretic conception. 30. It would be surprising if the motivations for something as minimal as (G.E.) were the same as the motivations for (O.E.). In spelling the motivations for my own specific version of externalism in (Bilgrarni 1992), op. cit., I stick as closely as possible to the motivations that anyone might have for something as minimal as (G.E.). These motivations are altogether different from those that have inspired (O.E.). 31. I try to answer these and other questions in (Bilgrami 1992), op. cit.

267

Sidelle, Alan, Identity and the Identity-like , Philosophical Topics, 20:1 (1992:Spring) p.269


Identity and the Identity-like

Alan Sidelle University of Wisconsin-Madison

I One of the central legacies of Putnam and Kripke's work on reference and modality is the view that empirically discovered identities-both singular and general--can issue in knowledge of necessary truths not knowable through reflection alone. I While they are concerned to argue that the mechanisms of reference preclude the truth or falsity of certain identity statements-'Hesperus is Phosphorus,' 'Water is H20' (as well as certain essential predications-'Necessarily, cats are animals')--from being known a priori, and that if these statements are true, they are necessarily so, they show little interest in the question of how such empirical identities can come to be known. 2 This does not mean that they would not find this question of philosophical interest, but suggests that they think the results of such inquiry would not bear on their central contentions-all that matters is that there can be such knowledge, or perhaps even only that there can be such statements, since the claim that we can know necessary a posteriori truths is not quite as important as that there are such truths. I do not, however, think that this is the case, and it is one of my projects in this paper to convince you of this. I shall argue that the way in which we can come to know of empirical identities requires a priori knowledge which implicates more semantic structure

269


into the tenus of such identities than the causal theory allows, or at least typically acknowledges, 3 and further, that it is this semantic structure, rather than a mind-independent modal structure of reality, which is responsible for the modal status of these statements. Indeed, this semantic structure is needed for the statements to be true, not merely for them to be known. If I am right, then this investigation directly bears on the key semantic and modal contentions of these influential philosophers, or at least on their ability to unseat the empiricist semantics and metaphysics which are their target. 4 I hope the inquiry into knowledge of empirical identities is also of independent interest, beyond its bearing on these important semantic and modal issues; presenting some sort of positive picture is this paper's other main project. To aid in our investigation, we will also be considering two other relations which have recently been the topic of much philosophical discussion: constitution and supervenience. 5 Constitution is a relation which holds between individuals (or a particular and some matter, but I shall not be concerned with that), and supervenience a relation between properties. They are of interest to us here because they can both 'look' like identity.6 That is, these relations may hold in cases where, in the actual world, their relata share all historical and spatio-temporal properties, having no actual differences-just as the relata ofthe identity relation do. 7 For this reason, I call them 'identitylike' relations. 8 Their introduction is often motivated by claims that for all their actual similarity, the relata involved sometimes differ modally, and thus, if identity holds necessarily where it holds at all, these items cannot be identical. Thus, it is sometimes claimed that even if the occurrence of some mental state is always accompanied by the occurrence of some brain state, the best that can be said is that the mental state supervenes on the brain state, since the mental state can occur in the absence of the brain state. By the same token, a person is not to be identified with his body, or a ship with a certain collection of planks, because in each case, these objects can exist independently of each other; thus, we may say only that the body constitutes the person, and the planks the ship. I hope it is clear why such relations might be of interest to our investigation into the empirical discovery of identity-I want to ask: In the absence of actual differences between individuals a and b, or properties P and Q (which is, of course, a necessary condition for their identity), what can justify us in believing these pairs to be identical, rather than to stand in some identity-like relation? The very fact that we do judge some such pairs to be identical, and others to stand in constitution or supervenience relations, is. itself rather noteworthy. What legitimates our claims that in some cases, we have one thing, and in others, we have two? Now, there are cases in which the relata of constitution and supervenience differ actually-a ship has undergone a change in planks, or a certain mental state is found to be accompanied by brain states of different types.

270


I will return to these later. But advocates of constitution and supervenience are typically 'hard line'-that is, they will assert that these relations, rather than identity, hold even in cases in which there are no (apparent) actual differences, so for simplicity, it is these with which I will begin my inquiry; as we shall see, nothing hinges upon anyone's willingness to take this stand, although again, it is without doubt the common position.9 Since we are pursuing our inquiry concerning empirical identities by asking what justifies our differential judgments that identity or identity-like relations hold, we will also reach some findings about the epistemology of constitution and supervenience. I hope this will also be of some interest. Considering them together will lead to a more robust conception of how the semantic structure of language guides our metaphysics-not only as it concerns modality, but also criteria of individuation and ontology generally. lbis is part of the larger 'reply to Kripke and Putnam on behalf of the empiricist' motivation of the paper. I will conclude with a preliminary investigation into two 'big metaphysical pictures', which investigation I think our findings should make us naturally want to pursue.

II There is something of a consensus that water is identical to H 20, and at least a strong amount of resistance to the suggestion that fear, say, is identical to any state of the central nervous system. Similarly, Hesperus, Phosphorus, and Venus are all unproblematically identified, while there is a fair amount of skepticism about this tree and the wood with the tree's shape. What accounts for these differences? The answer we may now have come to expect might be: We can imagine fear occurring without the corresponding neural state, and the tree being chopped down and hacked up with its wood being made into a box. Thus, they are possibly distinct, and thus not identical. One might respond that we can imagine water where there is no ~O, and Hesperus and Phosphorus turning out to be different bodies, but it will be answered that in these cases, we are not really imagining what we thinkwhat we are imagining is not water, and we are misidentifying either Hesperus or Phosphorus in the latter imaginative act. \0 . It should be readily apparent that even bolstered with this 'reinterpret the imagination' move, we have not had our question answered. Our 'What accounts for the differences?' question is just shifted onto the question of when 'apparent imaginings' are veridical. Why are we successfully imagining the tree and its wood (fear and its CNS state), when we try to imagine one without the other, but not Hesperus and Phosphorus (water and ~O)? The answer which immediately comes to mind is that in one set of cases, we

271


are trying to imagine as distinct items which are identical-and thus necessarily so-while in the other set, the items actually are distinct, and so can easily be imagined so. But if we accept this answer, the appeal to what we can imagine does us no good, for we will not be able to know what we are imagining until we know whether the objects of our imaginative project are identical or not. Now, with this failure of a simple appeal to imaginative differences to do the work, it might be thought that we should attempt to establish the identities more directly, without imaginative appeals. But by the very nature of the case, it is hard to see how this could be done. For we are supposing that in each of the cases, there are no actual differences between the relata in question. Thus, all the actual evidence we (could) have for the identity of water and H 20 should be paralleled by evidence we (could) have for identifying fear and its eNS state, and similarly for the case of individuals. And it should be emphasized that those who maintain that there are cases of identity-like relations are prone to eschew actual considerations-it is sufficient for them that we can imagine the items distinct. Thus, we have a confidence in the veridicality of imagined differencesor to put it a bit differently, in the reality of possible differences-underlying rejections of identity. Or, perhaps more perspicuously, what is being rejected is not identity per se, but the possibility of identity. Insofar as it is possible differences which are being proffered against identity, it is being claimed that identities in these cases could not be discovered, viz., 'I don't care what you actually find-nothing could suffice for identity in this case'. It is not so much that empirical research is done to find possible identifications, and then the proposition faces the final test of the imagination, but rather the imagination tells us whether we've got a possible identification, and if the answer is 'No', empirical research is never, properly speaking, into whether the items are identical, but only into what supervenes on (or constitutes) what. This should not be very surprising. Once we accept the necessity of identity, the acceptance of any identity statement (between rigid designators) carries modal weight. There is some tendency, I think, in the philosophical community, to think that what the necessity of identity does is make it, in a way, easier to discover necessary truths: we empirically ascertain an identity and infer the modal conclusion by modus ponens. But this is a doubleedged sword. If we cannot accept the modal conclusion, we must then apply modus toliens and conclude that the identity does not in fact hold. Rather than making necessities easier to come by, the necessity of identity can make identities harder to come by. Given their modal weight, we should not be surprised to find out that the justified acceptance of an empirical identity statement is not simply a straightforward empirical matter.

272


What I am claiming, then, is that no amount of empirical research can, of itself, count for a claim of identity over a claim of constitution or supervenience. All that can be established is 'identity or supervenience', with the possibility of identity determined by imaginative appeals. The difference between the cases of water and H 20, or Hesperus and Phosphorus on the one hand, and fear and some CNS state, or a tree and its wood on the other, is that in the former cases we are willing to redescribe our imaginings (or discount them-see note 10) if the empirical evidence turns up positive, while in the latter cases we are not. What we need to ask, then, is what accounts for this willingness, or lack thereof, or by the same token, what accounts for our certitude that in some cases, identity is just out of the question?

III As we have already seen, our willingness to redescribe the contents of our imaginations cannot be a matter of whether we have become antecedently convinced that the objects in question are identical. There is no independent test for whether the relation in question is identity, as opposed to some identity-like relation. What makes us willing, in certain cases, to entertain the possibility of identity (or equivalently, what makes us willing to redescribe our imaginings if the empirical data turns out favorably)? One interesting feature of identities which we accept, or are willing to entertain, can be seen in the familiar water-H 20 case. While water turned out to be H20, it could have turned out to be HzS04' or XYZ, and if it had, we would presumably be as willing to accept these as H 20 as identifications with water. On the other hand, it seems unlikely that we would consider 'liquid enjoyed by George Washington' as a candidate for identity with water, even if it turned out that all and only water (the items we call 'water') had this feature. What this suggests is that there are 'candidacy' criteria which anything must meet if it is to be a candidate for identification with our subject of inquiry. And I suggest that in order to be such a candidate, something must have identity conditions which are compatible with those for our sUbject. 11 Identity conditions are compatible if, and only if, when spelled out precisely, they might be the same. (Compatibility of cQnditions is required here, rather than sameness, because at the time an identification is being considered, it will be unclear what the precise identity conditions are-if, for example, the identity conditions for elements are to be given by their atomic structure, all atomic structures are candidates for identification with some element, but before we make the proper identification, we cannot specify, say, 'atomic number 79'. If we had to know that the fully specified conditions were the same, we would already know whether the identity held.) So,

273


for instance, if it is true that water is ~O, the identity conditions for water involve, say, chemical structure or deep explanatory features. Since these are shared by H 20, H 2S04• and XYZ, they are all candidates for identification. However, the identity conditions for being liquid enjoyed by George Washington revolve around a certain man, and thus presumably don't make the same explanatory demands. When we switch over to trees and their wood, the idea is similar. The identity conditions for trees involve certain biological requirements which are not shared by hunks of wood. Now, it is not clear just what should be said about the relation between our knowledge of identity conditions and our imaginative experiments. Do we know that fear is not individuated neurally because we can imagine people in fear being in different neural states? Or do we stick to our guns when asked to redescribe the imagination (such a redescription would involve saying that at least one of the fellows is not really afraid) because we know how fear is individuated, and thus that this can be a case of fear even though the neurology is different? This might be thought important because if our knowledge of identity conditions is empirical, then in the end we will have an empirical difference between identity and 'mere' constitution and supervenience. By investigating the world, we can find out what sorts of identity conditions its items have, and will then be able to judge whether the identity conditions for proposed candidates for identification are compatible. What I now wish to argue is that this cannot be right. Our basic knowledge of identity conditions is non-empirical and a priori.

IV The question of whether, in cases of empirical identifications, our knowledge of the identity conditions for the object(s) involved is a priori may seem peculiar. After all, it is empirical identifications under consideration here. However, the fact that it is empirical, say, that water is identical to H 20, does not show that at a more general level of specification, the identity conditions for water might not be a priori. It could be part of our use of the term 'water' that we intend to be (rigidly) picking out the chemical structure, if there is one, or otherwise deepest explanatory feature, of (enough of) these clear, liquid, freezable, and steamable samples. If this were so, it would be analytic and a priori that water is individuated by the deep structure of (enough of) these samples (if there is one), although it would be empirical that water is individuated by the H 20 composition in particular, that is, that Hz0 is that deep structure. Thus, empirically known identities may be underlain by a priori knowable identity conditions, more generally specified. 12 And this indeed fits our bill, for we are looking for knowledge of iden-

274


tity conditions which is of a more general sort. How else could we know that something was a candidate for identity with a certain item, but not know whether it was in fact identical to that item? If we are to have knowledge of the compatibility of identity conditions which is prior to our knowledge of the concrete identities-which I have suggested needs to be the case if we are to be able to distinguish identity from the identity-like-this knowledge has to be of identity conditions specified generally, or at least, not in their full specificity. Now this, of course, does not settle the case that this knowledge is a priori, nor thus that at some level, identity conditions are given analytically. All that we have shown thus far is that even in the realm of empirical identifications, the basic sort of identity conditions may still be analytic ..But couldn't it in fact be the case that matters are empirical all the way up? I think not, and the reasons have already been laid out. Recall that we are trying to determine how we know whether the relation between two items which do not differ actually is one of identity, or only of constitution or supervenience. We have been suggesting that the key element concerns our knowledge of the identity conditions, or criteria of individuation, for the items involved, in particular, our knowledge of whether, for these two items, these conditions are compatible. But now, even if somehow our knowledge of identity conditions might be thought to be empirical all the way up, how could we ever, for the cases under consideration, come to an empirically justified conclusion that the items in question do not have compatible identity conditions? Supposing our knowledge of identity conditions to come only from empirical evidence, how could two items which were empirically equivalent-actually indistinguishable-justify for us the belief that their identity conditions were not compatible? Indeed, how could we gain empirical justification for any view other than that their criteria are just the same? It seems that if we are ever to have justification for believing that two such items do not have compatible identity conditions, and thus, if we are ever to be justified, in these cases, in distinguishing the identical from the constituted or supervenient, this justification must be non-empirical, that is, a priori. This is, I think, to be expected. In actual practice, in these sorts of cases, everybody who urges that we do not always have cases of identity appeals to imaginary cases, to thought experiments. Nobody says 'We'll have to see what our research shows us' . And these sorts of thought experiments are not empirical. What they give us insight into is our concepts, into our decisions about how we want to use our terms. That we will not call XYZ 'water' (if this is so), after discovering that actual water is composed of ~O, and that we will call the appropriate sort of state 'fear', even if not accompanied by its actual world eNS counterpart, tells us something about how we use the words 'water' and 'fear'. These imaginings cannot be thought to represent, or be based (solely) on our empirical findings, because by hypothesis, there

275


are no empirical findings which could justify our rendering discrepant judgments about whether something was water (fear) and whether it was I\O (the relevant eNS state). If water is identical to I\O, and fear merely supervenes on our eNS state, it is because we have decided to use the word 'water' in such a way that it will apply to something in any possible situation just in case it has the actual deep structure of these samples, and because we have decided to use 'fear' in such a way that it will apply just in case some state has, say, the right functional structure, or internal character, and do not use eNS terms with the same constraints. The difference between identity and the identity-like resides in the relations between the general individuating standards we have decided to associate with our terms. An objection is looming. Hasn't the deck been stacked in favor of our analysis by focusing on this limiting case, where there are no empirical differences between the items? Don't constitution and supervenience first find their homes in cases where this is not so, and then get extended to the cases we have been discussing? And further, if this is right, then may not the extension itself be fully legitimate-after all, once we have established an identitylike relation, why need there be any limit to how closely the items actually resemble each other? In some cases, this may end up sufficing for identitybut need it? Indeed, could we not even extrapolate from cases of non-identitylike supervenience and constitution (see note 8 for this locution) to cases in which actual differences are lacking, just by noting that the items involved are of the same type as in other non-identity cases? We could learn from nonidentity-like cases that the objects involved were of different types, and so couldn't be identical. We need, then, to look at cases of constitution and supervenience which are not so identity-like, i.e., in which the items are not actually indiscernible.

v Let's suppose that when people say they are in pain, there is no single neurophysiologically recognized type of state which they are all in. There are rather, suppose, four or five such types of state. Or note that mouse traps do not always have the same physical structure and are not always made of the same kind of materials. Here we have cases in which, excluding disjunctive identities, we seem to have no hope of maintaining property identities; being in pain can at best supervene, for some portion of the population, say, on being in eNS state P, and being a mousetrap can at best supervene on the having of a certain physical structure. Similarly, consider a ball of clay which, after sirting around in the potter's shop for a while, is shaped into an urn. Since the ball of clay antedates the urn, they too cannot be identical; the ball of clay, at best, constitutes the urn.

276


In these cases, our knowledge that the items involved are non-identical seems to be unproblematically empirical. We can detect actual differences between the items and apply Leibniz's Law to finalize the result. The relations between the items can still be rendered in English with the word 'is'"This piece of clay is an urn", or "In humans, being in pain is being in eNS state P"-but the 'is' is the is of constitution or supervenience, not identity. But is this as unproblematic as it seems? Let us begin with constitution. In order to know that the piece of clay antedates the urn, one needs to know that the clay which is now co-located with the urn did not come into existence when the urn was made, and that the urn did not in fact exist in a preurn phase. To pursue the latter, someone who was keen on maintaining an identity-perhaps to hold on to the view that no two physical objects can wholly occupy the same place at the same time-might say that indeed the ball of clay is (i.e., is identical to) the urn, but that this object, which is now an urn, was not always an urn. It is only contingently an urn, and was not, then, antedated by the ball of clay. Thus, the apparently unproblematic empirical knowledge that the ball of clay and the urn have different times of origin requires some further information, perhaps that urns are always, or non-contingently, or necessarily begin as, urns. But these are far from unproblematic bits of empirical information; rather, they concern the identity conditions for urns, which we have suggested are, at least at some level, not empirical at all. One may search for something less problematic, but the search, I believe, will be futile. To put the point concisely: Epistemologically speaking, Leibniz's Law is a device for allowing us to make inferences once we know that some A and B are identical. It is epistemologically inert for the purpose of getting us empirical derivations of the non-identity of things. Though it has the logical capacity for this, it requires as input that A has some property which B lacks, and surprising as it may seem, this is not a purely empirical matter. Knowing, for instance, that an urn has just come into existence requires knowing something of the existence conditions for things which are urns. For another example, knowing that building A is completely made of brick and building B completely made of steel would suffice for us to know that these two buildings were non-identical. But if their identity is in question, we need to ask how we know that building A is completely made of brick. If 'building A' and 'building B' pick out the same building, then building A is brick here, and steel there (likewise for B). Our knowledge that building A is completely made of brick requires that we know the boundaries of the building. We may, of course, stipulate that by 'building A', we mean the object with these spatial boundaries. But while this gets us our knowledge that the building is all brick, this knowledge comes after our knowledge that buildings A and B are distinct. To return to the clay, I can offer no more than an exhortation to

277


reflection as proof that no purely empirical methods, i.e., methods not crucially dependent on a priori knowledge of identity conditions, will yield that the clay and the urn are not identical. However, let me consider one more attempt. Many of us, from some experience in elementary school, are familiar with aborted pottery projects. We know of urns that are made, looked over, and smashed back into unformed clay. Does this not show that an urn is not identical to the piece of clay from which it is constructed-and perhaps even yield enough knowledge of the identity conditions involved to let us know when an urn has come into existence? Plainly not. Of course, on one reading, we do know that an urn comes into existence when something of a certain shape, perhaps designed with certain intentions, is formed. But this does not tell us when an entity which is an urn comes into existence, any more than knowledge of peroxide can tell us the origins of individuals who are bleach-blonds. A bleach-blond-call him 'David Lee'-may redye his hair, but this does not show us that David Lee is not identical to the person who went in for, or emerged from, the bleach-redye process. Similarly, without independently knowing that urns exist for only so long as they are urnish-and that this is false of lumps of clay-we cannot know that the smashed lump of clay is not identical to the urn. Thus, far from being able to use cases of non-identity-like constitution to give us an unproblematic hold on the difference between identity and constitution, the very knowledge of identity conditions we found we needed to appeal to in the identity-like case are required for us to even get a case of non-identity-like constitution.

••• What about supervenience? Can we not find empirically unproblematic cases of non-identity-like supervenience? This may seem simpler, since we are not confined, as in the case of constitution, to one entity on each side of the relation (the constituted and the constitutor), so the prospects for maintaining that there is identity seem bleaker. Some mousetraps have physical structure M, others N, others Q. Thus, being a mousetrap is not identical with having physical structure M, nor N, nor Q. To pursue this along the path we have treaded above, we need to ask whether we can (purely) empirically establish both that all of the items in question are mousetraps, and that physical structures M, N, and Q are distinct physical types. This immediately raises the question of whether it can be just an empirical matter whether two or more tokens are of the same type. Certainly, once it is clear, for any type F, what makes something count as an F, we can investigate particulars and find out that some of them are F, and others are not. But whether this is just empirical will depend on our knowledge of these criteria

278


of F-hood. And, as might be expected, I do not see how such knowledge can be completely devoid of a priori elements, and at just the crucial points. Consider how such investigations are undertaken. On the simplest 'purely empirical' model, we collect a bunch of things we think are F(s), and try to find out what they have in common. But there will be infinitely many features which these particulars share. So we need to somewhat circumscribe our investigation. On the face of it, the way this is done is by either having already some idea of what the salient F-making features are, or by having some general strategy or view as to what sorts of features are typically 'boundary making', or perhaps, are boundary making for the sort of investigation being undertaken. Thus, for instance, with water we are looking for some sort of deep explanatory structure, which we have found to be ~O. If it had turned out that these samples were predominantly XYZ, then presumably XYZ would have whatever claim H20 actually has to being the 'essence' of water. What if most of the samples were ~O, but a substantial minority were XYZ? One outcome would be that lots of what we thought was water was not in fact water, but 'fool's water'. Another view would be that we had discovered that there were two types of water. Within this view are two possible expansions-(a) we discover that there is some deeper explanatory microstructure shared by both ~O and XYZ, and hold this to be the essence of water, or (b) we discover nothing of the sort, and take the 'essence' of water either to be disjunctive, or to consist in various superficial and/or functional features, though perhaps holding the essence of the different types of water to reside in ~O and XYZ. Why have I gone into all these options? Well, what I think it indicates is that we enter into our 'empirical' boundary investigations looking for a certain sort of thing. If we discover a large enough commonality of the sort we are looking for, we will consider this to be the boundary, or essence, of our kind. If we don't, as Putnam has said, we revert to our fall-back criteria, which will, most likely, involve the more overt features which we focus on in our everyday ascriptions of F-hood.13 But we do not, again, just throw a bunch of samples together and see what they have in common. So the question is: What is it which guides these lines of investigation? Certainly, if it were not for our empirical belief that things have explanatory microstructures, we would not look there. But it cannot simp~y be this belief which allows us to deem such a structure, once found, to be the essence or boundary of, or identical to, the kind we are investigating. There are lots of features which we look for in our inquiries which we would not if we didn't think things had these features, but which we do not take to individuate kinds. Nor, again, could it simply be that we think such features are of causal or explanatory importance. We need to know why these are individuating-particularly when we know that not all kinds the members of which share a deep structure are in fact bounded by that structure (e.g., pain, in the identity-like

279


case).14 It again seems to me, then, that at bottom, we must come to something which is not a belief, but a convention, a decision, or if one prefers, knowledge of such a decision: namely, a decision to bound this, or that, or these kinds by this or that sort of feature. 15 One can investigate water samples forever: one cannot discover that what makes something water is the composition H 20 unless one has decided, roughly, that the criteria for properly applying 'water' is to be determined by the deep structure, if there is one, of these samples (or, 'most of the samples we call 'water'). And as it is for water, so it is for mousetraps, and for physical structures M, N, and Q. Physicists can investigate particulars, find them to be similar or different in various ways, and determine (or antecedently know) that these differences are causally important. But to know that these differences put things in different categories, make them different kinds of things, they must know that physical kinds are bounded by this sort of difference. Of course, I presume any physicist will have such knowledge; indeed, it might be claimed that it is, at least in part, constitutive of physics that it draws its boundaries as it does. The question, though, is whether this is an empirical matter, and I am claiming that it is analytic that physical kinds are bounded by physically salient similarities and differences. This may not be very bold or shocking, but it is enough to establish that the knowledge that M, N, and Q are distinct physical types is not completely an empirical finding. As with particulars, knowledge of supervenience requires knowledge of boundary or identity conditions, at least generally specified, and this knowledge is not~annot be, I am arguing-knowledge (in the first instance) about the world, but rather, is of general individuating conventions, rules for the application of terms. Perhaps the argument can be made more plain by approaching the point from a slightly different angle. Suppose physicists have found items of a certain physical construction, and call it 'M'. Have they, by this act, determined of any other token physical construction that it is not (an) M? This is the picture we seem to get from Putnam and Kripke; we introduce, or regulate, our use of a general term via certain exemplars, and the world then takes over in determining the extension of the term. This is what allows us to then undertake an empirical investigation of what the kind, or property is-that is, what it is identical to, or what it is essentially (and ditto, of course, for the case of individuals, though this is more Kripke's turf than Putnam's). This is how there can be empirical investigation of property identities; where general terms get their reference through meanings, the identity conditions will be analytic and knowable a priori, and all we could find out empirically would be things like the coextension of different types. Now, I have no trouble with the claim that our physicists have determined that certain things are not 'M'-but not by this act alone. As is by now familiar, and as I have said, there are too many features any set of exemplars have in common for ostension alone to determine extension-in this case, one important feature shared

280


by all M's is being a mousetrap. If 'M' in fact does not cover tokens which we later come to identify as N's, it is because it is understood to be part of how terms are introduced in physics that salient physical features determine their extensions. And while the agreement to do so may not be separable from the agreement to do physics, it is nonetheless a conventional understanding. One could investigate the exact same tokens, have the same sort of naming ceremonies, have the other features of the world be the same-and yet be individuating kinds differently, such that the things we (now) call N's would fall within the extension of 'M'-'M' could express the property of being a mousetrap. Perhaps one would not be doing physics-but this, of course, depends on how the sciences are individuated-which, one may have guessed, I also regard as a conventional matter. Kripke and Putnam do allow that in introducing these terms, we intend them to be terms for 'this kind of thing'; I am claiming that the intention needs to be a bit more specific, and that it is this specification, and not the 'real essences' of things, which determines where the boundaries are set. 16 Thus, the same considerations which suggested to us that cases of identity and identity-like relations differ only in virtue of the sorts of identity conditions we have conventionally associated with the terms in question also raise their head when we turn to cases of non-identity-like supervenience and constitution. Knowledge of whether items are identical is posterior to knowledge concerning their identity conditions, and the sense in which we 'discover' that items are identical, or stand only in constitution or supervenience relations, must be taken in this light. Taken by itself, the empirical component of 'empirical' discoveries of identity does not and cannot establish that the relation in question is indeed identity, nor then, that the relation is necessary. Whatever empirical data is pertinent to these discoveries could have been exactly the same, with opposite results concerning the matter of identity; it is our a priori knowledge of identity conditions, given by analytic criteria, which makes the difference between identity and the identity-like, and it is on these conditions that all the metaphysical weight rests. The crux of the distinction-and consequently, the fulcrum of the truth-value of identity and identity-like statements themselves-is semantic, and not factual.

Before moving on to a final speculative section, let me review our findings and make good on my earlier claim that this investigation would have a bearing on the broader semantic and metaphysical position of Putnam and Kripke. As our findings themselves are in direct conflict with one aspect of this

281


position, it may be best to present them in contrast with this. Empirical identity statements, on the view that Putnam and Kripke (and many others, before and after) hold, are relatively unproblematic. Names for objects, kinds, and properties may be introduced in a variety of situations, and it may be the case that on different occasions, we introduce different names for the same item without being aware that it is the same item. If, as a matter of fact it is, then the identity statement between these terms is true, and, if the terms are names, or at least rigid, the statement is necessarily true. The naming process itself can be pretty ostensive and at any rate carries no semantic baggage in the form of meanings. In short, the names are just labels, and the truth of an identity statement is just a matter of whether we have, with two different names, labeled the same thing. Further, since the names are just labels without (analytic) semantic content, we can investigate the object of our reference without any linguistically given information about the essential properties or identity conditions of the object in question, and so can find out, empirically, what these essences and identity conditions are. On the view I have defended, this picture is too simple. We came at the problem by asking how one can empirically discover these identities, worrying, in particular, how we can know that Hesperus is identical to Phosphorus, rather than constituting it (or being constituted by it, or being distinct and co-constituted by something else). I claimed that we must have a priori knowledge of the identity conditions, at least generally specified, for the objects of inquiry. In some cases, this knowledge will determine a priori that a certain identity statement cannot be true, no matter how the world turns out. This will be the case, for instance, in identity statements between names for statues and those for pieces of clay-for we know that whatever winds up filling in the values of the generally specified conditions here, the resulting fully specified conditions cannot tum out to be the same: it will always be possible for each to exist separately from the other (although, if Kripke is right, it may be the case that the statue has to originate in its particular clay-this too, of course, will be contained in the conditions). Consequently, the only possible empirical discovery here is that the items are related in some identity-like way. In other cases, the identity conditions are known to be compatible, so that an identity is possible-we need to look at the world to find out what in fact takes the values of these conditions. 'Water is H 20' is a possible identity statement, if, as Putnam's discussion suggests, 'water' is used such that its reference in all worlds is to be determined by sameness of deep explanatory structure, as 'H20' is. So is 'Iron is Zinc'. This one turns out to be false, because the deep structures of the samples by which the term was introduced tum out to be different. But the a priori knowable specification of the identity conditions for iron and zinc only tells us that these are also individuated by 'deep explanatory structure', so prior to investigation, we do not know that this is false. Further, since this knowl-

282


edge concerns individuation across worlds as well as within a world, finding the values of these conditions can tell us what is essential to particular objects. So there can be 'empirical discovery of essence' as well as empirical discovery of identity, supervenience, and constitution. But it is rather misleading, on the account I have given, to put the point simply like this and leave it. Because with respect to all these matters, there is a crucial a priori element, which element plays the decisive role in justifying our move from some empirical finding to the claim of essentiality or to identity, supervenience, or constitution. The empirical findings themselves leave these matters completely underdetermined-they are not things we do, or can, simply 'find' in the world, through empirical investigation. Further, and perhaps more importantly, this epistemic finding reflects the fact that in order for an identity statement (at least, between rigid designators)--or for that matter, a statement of mere supervenience or constitution (or essence)-to be true, or even to have a determinate truth-value, the terms of that statement must be analytically associated with identity conditions, at least generally specified. Without this, the terms will hover in reference between any number of possible objects. If 'Hesperus' is to refer to a certain planet, rather than the stuff of which it is composed, or if 'pain' is to refer to a qualitative psychic state rather than the neural state with which that psychic state always occurs, it must be that we (at least implicitly) associate certain identity conditions with our use of these terms. The terms cannot both be 'mere tags', and at the same time, have reference determinate enough to decide between identity, or merely identity-like relations holding. So, on a 'pure' causal theory, where there are no analytically given constraints upon the reference of a term, we can have neither the empirical discovery of identity and essence (or constitution or supervenience), nor even the truth of empirical identity (supervenience, constitution) statements, or essential predications. Our results bear not merely upon the epistemology of identity and the rest, but upon the semantics for the terms involved as well, and upon what is needed for the truth of these statements. Any term that can appear as a subject in such statements-which presumably includes all nouns-if the statement is to have a truth-value, cannot be a mere tag, as Putnam and Kripke would have it, but must have (at least generally specified) analytically given identity conditions. For Putnam, at least, this is an important result, since so much of his writing is aimed at the.rejection of philosophically significant analyticity. While Putnam is not so Quinean as to deny that there is an analytic/synthetic distinction, or that there are analytic truths, there is, in his view, none that will, in his lovely phrase, "bake ... philosophic bread and wash ... philosophic windows."!? But it is precisely such baking and washing which these analytic identity conditions are being called upon to-and which I have argued they do-provide. Further, along these lines, Putnam has long been

283


Quinean (I cannot say for certain whether he still is, but he has certainly convinced many others to be) in respect of finding philosophy continuous with, or maybe even a branch of, natural science, breaking sharply with the twentieth-century empiricist conception. And the spread of the idea of science discovering identity, identity conditions, and essence-traditionally taken to be the a priori province of philosophy-has certainly done much to promote this view throughout the philosophical community. If I am right, however, then at least in these matters, philosophy still plays a sharply distinct and a priori role. It is linguistic or conceptual analysis, not science, which informs us of identity conditions, and the role of science is simply to find the values that the variables specified by these conventions take. Philosophers are not doing science, nor scientists doing what has traditionally been taken to be the job of philosophers. My argument is finished, and I hope I have also indicated at least some of the bearing our findings have on the larger views of Putnam and Kripke. Before concluding, however, I would like to make a start on a project which I think our results should make us interested in undertaking. So far, we have been fundamentally talking about epistemology and semantics, about identity (and identity-like) statements and our knowledge of them. But what of the relations of identity, supervenience, and constitution themselves? Are they, as well as the statements which assert their obtaining, somehow products of our conventions? What sort of a metaphysical interpretation ought to be given to our findings? If these relations are not just out in the world, awaiting our discovery, what is going on? In the next and final section, I will outline two 'big metaphysical pictures' which seem compatible with our account, and which I hope will be seen as worthy of further exploration.

VII As I see it, there are two ways of understanding the need for identity condition-specifying conventions associated with our terms. On one view (,Picture I '), these conventions are needed to select, from among the many mind-independent objects which are potential objects of reference, those to which we are referring. Since there are so many of them, overlapping allover the place, with indefinitely many of them sharing all actual properties and differing only modally, we need to specify identity conditions to refer determinately to one of them. On the other view ('Picture 2'), the conventions playa more metaphysically robust role. Rather than selecting from among the many objects out there waiting to be referred to, the conventions articulate (or create, or construct-but 'articulate' seems to me better) objects from the independently inarticulate world. We need these conventions not because

284


there are too many items out there to refer to otherwise, but because there are too few: in fact, none. Let me say a bit more about both of these views. According to Picture 1, the world is populated by entities, replete with their own identity conditions, and which stand in numerous relations, among them identity, constitution, and supervenience. These are all given mindindependently. However, they are dense. That is to say, wherever you have an object, you have indefinitely many, differing only in their identity conditions. There are particulars which share all their (extensional) properties at all times (like our urn and its clay, when they are created and destroyed together), particulars which share all their properties at a given time (and so which can differ at different times, when they do, because their transtemporal identity conditions differ), and properties which are coextensive (also, either always, or at a time). Because of this, if we are to pick out, to refer to, some entity as opposed to another with which it stands in some identity-like relation, we need to stipulate identity conditions. Only thus can we pick out, say, an urn as opposed to the lump of clay which constitutes it. This is why I say that on Picture 1, our conventions select objects of reference-the objects are all out there, identity conditions and all, standing in the relations we are discussing, and our choice of identity conditions merely allows us to distinguish among them, and so to choose to which we will refer with a particular term. Without analytically associated identity conditions for our terms, we could not determinately refer to one of these objects, nor then, make true statements about them or assert the obtaining of the relations of identity, supervenience, or constitution which hold among them. Thus, we cannot 'merely' discover that these relations hold between the objects they do hold between, since the very holding of the relations prevents our singling the objects out by 'non-identity-condition-specifying' means. But while these conventions are needed in order for us to say or know these things (or even think them), neither the relations themselves, nor the objects they hold between, are in any way dependent upon the conventions. The moral of our investigation remains at the level of epistemology and semantics. Picture 2, on the other hand, disavows the notion that the only work our analytic criteria are doing is that of specification. It is not, as Picture 1 would have it, that when we point to stuff in a glass and say 'water' , that there is in the glass (at least) both water (H20) and thwater (more superficially individuated), and that we need to somehow determine which of these we are trying to get at. Rather, there is just stuff in the glass, which stands in innumerable similarity relations to other portions of the world, and by indicating, say, 'deep explanatory structure', as individuating, we have articulated a portion of the world. We are, of course, capable of other articulations, and it is in virtue of this that there can be even identity-like supervenience and constitution, that there can be two distinct individuals at a single place and time. The world, on this picture, is inarticulate; that is to say, it does not

285


contain items with their own identity conditions, which is to say that it does not, as such, contain individuated items, which are, I take it, the sorts of things about which we regularly talk with our use of nouns. On this picture, then, analytic principles of individuation do not determine which of the many pre-articulated! individuated items in the world about which we are to talk, but rather carve up the world into these items. Identity and the identitylike relations both arise out of this practice. As might be expected, I favor Picture 2, although the argument of this paper can at best be claimed to suggest it, rather than argue for it. ls Before saying something of why, though, let me first just make a few clarificatory remarks about Picture 2. Many philosophers find this position intuitively unattractive, and I believe this is in large part due to mistaken views (perhaps shared by some advocates of the picture) about what this position asserts or is committed to. First, it might seem to be suggested that if we never had a term or concept to pick out the stuff in lakes according to deep structure, then water would not have existed, and that, perhaps, prior to philosophers' recent introduction of 'thwater', there was no thwater. 19 This would indeed be unacceptable, but it is neither implied nor intended. The world is capable of being cut up in so many ways, and whenever we consider such a cut (some principle of individuation), we are considering the world cut that way, i.e., so articulated. An articulation will specify both actual conditions which must be met by something in order to be (an) F, and identity conditions for tracing Fs through space, time, and possible worlds. 20 If there are portions of the world which meet the actual conditions, then there are Fs (or there is F where 'F' is a mass noun), and to say that Fs would or would not exist in a given situation (e.g., if we had not had such-and-such conventions) is to say that there would or would not be portions which met those conditions in that situation. 2l But the conditions are not met by these portions meeting anything beyond the actual conditions, for, on Picture 2, there is nothing in the world to do that. So, for example, there is no segment of the world, considered in itself, which is H 20 in all possible worlds. However, there is a portion which is actually H 20 (and there would be portions which were ~O in [many] worlds in which we had no 'deep structural' conventions), and it is by meeting this that this portion of the world is (would be) its water. This portion, once so picked out, can then be said to be necessarily ~O-but this is the work that is done in the process of individuation. Briefly, when we evaluate a counterfactual (or a statement about past times before our conventions were introduced)--even one about what would be true if we had not had the conventions we have-we do so, as Kripke rightly insists, in our language, governed by whatever conventions govern these actual statements. Given how we actually use 'water', or how we have

286


specified the meaning of 'thwater', the question of whether there would have been (or was) water or thwater is simply the question of whether there would have been (or was) H 20, or wet, clear, drinkable, etc., stuff. Thus, a proponent of Picture 2 can acknowledge that F(s) would have existed (when this is true) even if we had not had the practices we do. (True in the case of water, false in the case of unicorns.) He will merely deny that this means that there are portions of the world out there with identity conditions of their own, maintaining rather that there are portions of the world with certain actual features, the having of which is quite independent of our conventions, and that full-blown identity conditions can be given which would actually pick out these portions, as Ol~tlined above. Second, Picture 2 does not claim that cases of supervenience and constitution are 'really' cases of identity. This might appear to be the case, since it seems like that view says that you can have just one thing, and then by applying different identity criteria, you can get overlapping entities. This might not be so bad, so long as this didn't lead one to say that "really, the urn and the lump of clay are identical." However, more correctly, the problem is that there is no way of talking about 'one thing' here. Of itself, the portion is no more, and no less, an urn than a lump, and is devoid of identity criteria. Indeed, if it had such criteria, we would be able to ask whether it was identical to the urn or to the lump, and as it couldn't be both (since they are not identical), one could not maintain that they were 'really' identical. This in fact is part of why Picture 2 emerges as a candidate view from the considerations we have been discussing-if one attempts to represent identitylike relations as arising from the application of different criteria of individuation to single objects, rather than to stuff, one will misrepresent these relations as relations of identity. Thus, it is important to see that on Picture 2, the world 'of itself, if you will, is thingless, so that the concept of identity applies no more (and no less) than those of constitution and supervenience. Finally, and what I hope goes without saying, this picture is not idealist. The world is all out there, we do not think it up, when we die it does not go away. It's full of stuff and features-they are just not articulate, that is, pre-individuated. After all, while convention plays a large ontological role on this view, it is no part of the view that all truths are analytic. The conventions carve the world up, but what is true or false (aside from those statements made true by our carvings-like 'Bachelors are unmarried') depends on more than how we have carved things up. While our conventions determine what Mars would have to contain in order for there to be water there, they are silent about whether these conditions are met. So while idealism and verificationism are, strictly, compatible with Picture 2, they are neither entailed nor supported by it. All the implausibility of those positions may be

287


granted~r

urged-by proponents of Picture 2 as by anybody else. From a God's-eye point of view, the world, as seen by Picture 2, is just as seen by anyone else--except that there are no 'lines of individuation' to be seen. So, in conclusion, let me just briefly indicate a couple of things I think Picture 2 has going for it over Picture 1, asserting again that our conclusions about analytic criteria of individuation, and the epistemology of identity (and constitution and supervenience) are wholly independent of this issue. The big point, to my mind, in favor of Picture 2 is that it does not postulate either real, mind-independent identity conditions, nor what comes with them, real, mind-independent modal properties. First of all, we have seen that our knowledge of these things requires, at some deep level, a priori knowledge, and that we need to bring analyticity into the story, which analyticity suffices to determine these boundaries. Thus, at least pending some special argument, it is superfluous to postulate mind-independent identity conditions in addition-they do no further work and could not do the work on their own. Further, there is the question of their 'surdness'. That is, they seem to be metaphysical danglers which cannot be rooted in anything, or at least anything natural. For, as we have noted throughout, there are no actual differences between the relata of identity-like relations, and those that may be urged (see note 7) seem to similarly dangle. What is it about an urn that makes it the case that it has different identity conditions and different modal properties from the clay of which it is made, and which came into existence at the same time it did?22 While we can sometimes distinguish 'further' facts of qualitatively identical items by appealing to contextual features, the two items here will be contextually the same as well. So unless identity conditions are just 'there', we have at least a naturalistic argument to the effect that identity conditions come from us, rather than finding their home in mind-independent reality. Combined with the above 'superfluity-givenanalyticity' consideration, we have some reason to prefer Picture 2. In addition~r perhaps just elaboration-identity-like relations are less mysterious on this view. Indeed, their mysteriousness is nothing different from the mystery just discussed concerning identity conditions. Those who claim to find nothing odd about overlapping entities typically point out-as they seemingly must-that these things just have different identity conditions. But this should be cold comfort, if we have no idea how, given all their other similarities, their identity conditions can differ. If we adopt Picture 2, identity conditions do not supervene on 'mind-independent' facts, nor are they 'in the world' at all, so to speak. So distinctness 'in a place' makes some sense. Our criteria of individuation for individuals, kinds, etc., are more fine grained than our criteria for space-time locations, a fact which may be of some interest, but which is no deep mystery. Picture 2, then, I think, offers us better hope for understanding identity conditions, modality, supervenience, constitution, and so, as a whole, individuals, kinds, properties, etc.-

288


the whole individuated world-than does Picture 1. But further argument must wait for another occasion.23 I hope this final section may spark some interest in the articulation and investigation of these larger world views. If I am correct about our having these individuative conventions, it would seem natural and of some importance to pursue these ontological inquiries. While the epistemology of identity, or the difference between identity and the identity-like may seem like rather localized (though certainly interesting and important) issues, I think that serious thought will reveal that they (and many other such issues) are in fact of very wide-ranging importance to our understanding of the world, our means of representing it, and the relation between the two. I am pleased for this to be appearing in honor of Hilary Putnam, who has done so much to help us see such forests through the trees. 24

NOTES 1. Saul Kripke, Naming and Necessity (Cambridge: Harvard University Press, 1980); Hilary Putnam, "The Meaning of 'Meaning'," in Mind, Language and Reality (Cambridge: Cambridge University Press, 1975). 2. One of the more noteworthy efforts in this direction is David Lewis, "Psychophysical and Theoretical Identifications," Australasian Journal of Philosophy L (1972): 249-58. 3. If causal theorists do allow that the sort of semantic connections I shall argue for can obtain consistently with their theories, this is fine with me. But this will undermine two important negative claims commonly supposed to result from the causal theory, namely that there are no philosophically interesting analytic truths, and that the truth of the causal theory undermines the central contentions of-and so replaces, rather than modifies-traditional semantics. So long as the semantic connections, analyticity, and the preservation of the central claims of traditional semantics are acknowledged, it is of little moment whether one wishes to call the new theory 'causal (but not purely causal)' or '(modified) traditional'. The dialectical relation between traditional and causal theories, when semantic relations are acknowledged, is discussed further in chapter 6 of my Necessity, Essence and Individuation (Ithaca: Cornell University Press, 1989). 4. This qualification reflects the possibility that for some, the central important contentions of Kripke and Putnam are just that (a) all or most important referring expressions do not have purely qualitative, analytically given, necessary and sufficient conditions determining their reference, and (b) there are necessary a posteriori truths. These are no doubt important claims, and counter to the empiricist tradition; and on them my discussion will cast no doubt. However, these claims themselves do not undermine the central empiricist contentions concerning reference and modality, which are a commitment to philosophically significant analyticity-discovery of which constitutes the central activity of philosophers-and the belief that modality is grounded in conventions, rather than a mind-independent modal structure of reality. Since I think their work has been taken to undermine these positions-so many philosophers proceed and talk as if this has been established-I take their central contentions, at least as interpreted by the philosophical community, if not themselves, to be (a') there is no important semantic-analyticity determining---

The Philosophy of Hilary Putnam (Philosophical Topics, Volume 20, Number 1, Spring 1992)

The philosophy of Hilary Putnam (Philosophical Topics)

Hilary Putnam (Philosophy Now)

Hilary Putnam Contemporary Philosophy

Hilary Putnam (Contemporary Philosophy in Focus)

Hilary Putnam (Contemporary Philosophy in Focus)

Philosophical Papers: Volume 2, Mind, Language and Reality (Philosophical Papers Hilary Putnam, Vol 2)

Philosophical Papers: Volume 2, Mind, Language and Reality (Philosophical Papers Hilary Putnam, Vol 2)

Journal of Semitic Studies, Volume 55, Number 1, Spring 2010

Journal of Semitic Studies, Volume 54, Number 1, Spring 2009

AJS Review 17, 1 (Spring, 1992)

Topics in Stereochemistry, Volume 20

Think. Philosophy for Everyone Volume 10, Number 27, Spring 2011

Acta Numerica 1992 (Volume 1)

Acta Numerica 1992: Volume 1

Asian Music - Journal Volume 42, Number 1, Winter - Spring 2011

The Philosophical Writings of Descartes: Volume 1

Weird Tales 304 (Spring 1992)

Meaning and Method: Essays in Honor of Hilary Putnam

Concepts of Indoctrination (International Library of the Philosophy of Education Volume 20): Philosophical Essays

The Philosophy of Daniel Dennett (Philosophical Topics, vol. 22, nrs. 1 & 2, Spring & Fall 1994)

Number theory: Paris 1992-3

Topics in Stereochemistry, Volume 1

Nancy Springer - Number 20

Nancy Springer - Number 20

Handbook of Philosophical Logic Volume 1

Journal of Managerial Psychology, Volume 20, Number 3 4, 2005

East-West Journal of Mathematics (Volume 1, Number 1)

The Methodology of Scientific Research Programmes: Volume 1: Philosophical Papers (Philosophical Papers Volume I)

Буддизм России №1 (1992)

Philosophical Analysis and Education (International Library of the Philosophy of Education Volume 1)

The Philosophy of Hilary Putnam (Philosophical Topics, Volume 20, Number 1, Spring 1992)