Some would call the approach I am going to take in these lectures a Wittgensteinian approach, but even if my discussion will not be completely free of references to Ludwig Wittgenstein's later philosophy, I promise to avoid quotations from the Investigations and to employ Wittgenstein's own terminology as little as possible.²⁵ Perhaps this promise will relieve the anxieties some of you may feel! Specifically, my approach to Kim's challenging dilemma will be—not surprisingly—to question the intelligibility of the antecedent of (AUTOMATA). I will claim that this antecedent—once more,

(SOULLESS) Certain people do not have any mental properties, but all of their physical properties are the same as if they did and their physical environments are the same

—and also the conjunction of (SOULLESS) with "the denial of interactionism"—both fall short of full intelligibility.²⁶ And I shall argue that because neither of these supposed "possible states of affairs" is sufficiently intelligible, the question of what would happen if either of them were true is nonsensical. To draw an analogy, the statement "Cinderella's coach turned into a pumpkin" satisfies our criteria for intelligibility when understood as an utterance in a fairy tale (we know "what to do" with it, how to react to it, how to "play the game"); but if we detach the statement "sometimes coaches turn into pumpkins" from a very particular sort of context and try to discuss the question "What would happen if a coach turned into a pumpkin?" as if this were, for example, a serious scientific question, we would be talking nonsense. In the same way, one might enjoy a work of fiction in which someone was in love with an "automatic sweetheart," but that doesn't entail that talk of "possible worlds" in which some or all of us are automatic sweethearts makes sense.²⁷ But this will take a lot of unpacking, and I anticipate encountering certain misunderstandings.

The misunderstandings that I fear have to do with the erstwhile popularity and current well-deserved unpopularity of verificationism and logical behaviorism. Because the struggle against logical behaviorism that went on in the 1950s and 1960s, and the struggle against verificationism that is still going on,²⁸ have occupied much attention and still take up significant space in widely used collections of articles in the philosophy of mind, it is immediately expected that anyone who questions the intelligibility of a scenario like (SOULLESS) (or (SOULLESS) plus the denial of interactionism, or plus "the people in question do not behave any differently") must be a verificationist or a behaviorist. In the sequel I hope to dispel this misunderstanding by showing that there are quite different grounds for questioning the intelligibility of these scenarios, grounds that do not turn upon a prior commitment to a philosophical theory (such as verificationism) that is supposed to yield a general method for assessing the meaningfulness of an arbitrary statement or upon a commitment to the claims of logical behaviorism. In the next two lectures I hope to show that there are closely related grounds for questioning the intelligibility of reductionism (and of antireductionism as well, for that matter) and thus for finding both horns of Kim's dilemma very far from fully intelligible.

the question of the "independence"
of mental and physical properties

Let us consider (SOULLESS) once again:

(1) (SOULLESS) Certain people do not have any mental properties, but all of their physical properties are the same as if they did and their physical environments are the same.

Let us begin by looking at a prima facie reason for thinking that this makes full sense.

The following principle seems prima facie plausible:

(INDEPENDENCE) If A and B are two sorts of properties, and B-properties are not reducible to any A properties, then B properties are independent of A properties, in the sense that it is logically possible that the A properties should be present without the B properties.

Moreover, Kim himself clearly assumes the applicability of (INDEPENDENCE) to the case in which the A properties are the physical properties and the B properties are the mental properties in "The Myth of Nonreductive Physicalism."²⁹

If we agree that this principle is correct and applicable, then, on the assumptions we are making (at this stage in Kim's argument), our mental and our physical properties are independent. So why shouldn't one be able to conceive of one sort being present without the other, and to ask what "would happen if" just as Kim does?

With this question we begin to leave the harbor and sail out into deep waters. In a way, this should not be a surprise to those of you who are familiar with the great debates in analytic philosophy in the last half-century. For (INDEPENDENCE) assumes just the notions that Quine, in particular, attacked in his seminal papers "On What There Is" and "Two Dogmas of Empiricism": the idea of properties as having well-defined identity conditions and the metaphysical idea of "possibility." This is already enough to show that (INDEPENDENCE) is not a harmless truism. But my difficulties with (INDEPENDENCE) do not presuppose Quine's complete rejection of property talk and of possibility talk. I am concerned not with (INDEPENDENCE) in the abstract but with Kim's specific application of (INDEPENDENCE).

When we turn from (INDEPENDENCE) as an abstract principle to the implicit employment I believe it has in Kim's argument, we notice that one thing we must assume to use it to support the full sensicality of (SOULLESS) is that it makes sense to talk about reducibility as a relation that holds or fails to hold between mental and physical properties, that is, to suppose that we know what we are asking when we ask whether a given (so-called) mental property, e.g., thinking about the beauty of the roses in Safed, is or is not "reducible to" (or "nomologically coextensive with") a physical property. But the burden of my own arguments against "identity theories" (including my own former "functionalism") is that the notion of identity has not been given any sense in this context.³⁰ We cannot, for example (as I once thought we could),³¹ employ the model of theoretical identification derived from such famous successful reductions as the reduction of thermodynamics to statistical mechanics, because that model assumes that both the reduced theory and the reducing theory" have well-defined bodies of laws. (So that a proposed statement of identity, e.g., "temperature is mean molecular kinetic energy," can be established by showing that if we assume it the laws of the reduced theory—thermodynamics, in this example—can be derived from the laws of .the reducing theory.) We cannot sensibly ask whether the laws of "folk psychology" are or are not derivable from some set of assumed identities between the attributes spoken of in folk psychology and some set of computational properties (plus the description of the "program" that defines those computational properties), because, as I have argued elsewhere, the notion of a "computational property" depends essentially on what formalism the "program" is written in, and no one has the slightest idea what a formalism in which one can write programs that reduce folk psychology might look like. As long as we have not given a determinate meaning to computational property, all this talk of functionalism is just science fiction. No serious scientifically discussable issue of identity has really been raised.

To be sure, Kim’s purpose is not primarily to defend functionalism (although, as we shall see in a later lecture, one of his crucial arguments presupposes functionalism). What Kim proposes is rather that each mental property is "strongly supervenient" on a physical property, at least in the case of each species (or, perhaps, each relevant neurological "structure"). In my final lecture I shall argue that this has only the appearance of being clearer than unanalyzed talk of identity. But I shall withhold arguing this until that lecture and devote the remainder of this lecture to trying to see what follows if I am right about this.

Well, the relevance of the claim I am making to (INDEPENDENCE) is not difficult to see. Suppose we reformulate (INDEPENDENCE) as follows:

(INDEPENDENCE?) If A and B are two sorts of properties, and the question as to whether B properties are or are not reducible to any A properties has not been given any clear sense, then B properties are independent of A properties, in the sense that it is logically possible that the A properties should be present without the B properties.

I do not think that any philosopher would regard (INDEPENDENCE?) as having any plausibility at all! For if the question of reducibility (or nomological coextensiveness) has not been given any clear sense, then it would be natural to suppose that the question of independence has also not been given any clear sense.

What this shows, if I am right, is that if we assume (INDEPENDENCE), we have already assumed that all the propositions in question make full sense: the propositions about the "reducibility" or "nonreducibility" of physical and mental properties as well as the propositions about the "independence" of the mental and the physical properties. No wonder it then seems obvious that one can sensibly talk about people who are supposed to have all their physical properties and none of their mental properties!

What I want to bring into focus now and in the rest of these lectures is the way in which different philosophical pictures of how language functions and what meanings are (or, better, what knowledge of meanings consists in) will affect our attitudes toward just about every philosophical debate. This is something that Charles Travis, a philosopher whose work is not nearly as well-known as it should be, has written about extensively and also something that Stanley Cavell has written about at length, especially in his masterpiece, The Claim of Reason. If you will forgive what may seem a digression, I want to say something about certain general issues raised by these thinkers before I return to the issues about reducibility and mental causation.

"words have meaning only in the stream of life"³²

The difference between the two philosophical pictures I just referred to is laid out in a very clear way in Charles Travis’s book on Wittgenstein’s later philosophy, The Uses of Sense.³³ (Travis’s discussion-review of Grice’s philosophy of language, "Annals of Analysis,"³⁴ is perhaps the best short introduction to these issues as well as a powerful criticism of Grice’s procedures.) Travis presents what I am calling two "philosophical pictures" as different conceptions of what the semantics of an utterance in a natural language might be. One conception (of which Paul Grice was a leading representative) Travis calls the "classical" conception; the other—less familiar—conception, which he attributes to Wittgenstein and Austin, he calls "speaking-sensitive semantics," because the heart of this second conception is the claim that the content of an utterance depends on the particular context in which it is spoken, on the particular "speaking."

The second conception does not deny that words have "meanings" (that is, that there is something that is rightly called "knowing the [or a] meaning" of a word and that this knowledge constrains the contents that can be expressed using the word with what can be regarded as that particular meaning). What it denies is that meaning (or the knowledge in question) completely determines what is being said (what is supposed to be true or false, or if anything is being said that is true or false) when a sentence is used to make an assertion.³⁵

A couple of examples may clarify the issue. I certainly know the meaning of the words there, coffee, a lot, is, on, the, and table. But that knowledge by itself does not determine the content of the sentence There is a lot of coffee on the table; in fact, the sentence, simply as a sentence, doesn’t have a determinate content apart from particular speakings. Moreover, the truth-evaluable content of the sentence "There is a lot of coffee on the table" is highly occasion-sensitive: depending upon the circumstances, the sentence can be used to say that there are many cups of coffee on a contextually definite table (There is a lot of coffee on the table; have some), or that there are bags of coffee stacked on the table (There is a lot of coffee on the table; load it in the truck), or that coffee has been spilled on the table (There is a lot of coffee on the table; wipe it up), etc.

A different sort of example: I have an ornamental tree in my garden with bronze-colored leaves. Suppose a prankster paints the leaves green. Depending upon who says it and to whom and why, the sentence The tree has green leaves, said with my tree in mind, may be true, false, or not clearly either!

(Responding to the coffee example, a philosopher of language of my acquaintance—one wedded to Grice’s distinction between the standard meaning of an utterance and its conversational implicatures—suggested that the "standard meaning" of "there is a lot of coffee on the table" is that there are many (how many?) molecules of coffee on the table. But if that is right, the "standard" sense is a sense in which the words are never used!)

The classical conception does not, of course, deny that the exact reference of some words (e.g., the tense indicators and the familiar indexicals) is speaking sensitive; but it treats speaking sensitivity as a special phenomenon, easily adjusted for. The classical picture of the semantics of a natural language is Tarskian: truth conditions are recursively associated with all the sentences of a natural language (as in Davidson’s celebrated adaptation and modification of Tarkski’s work). The claim Travis makes is that speaking sensitivity, far from being a special phenomenon, is the norm.

Recently a good deal of literature³⁶ has appeared on this very phenomenon (which is now usually called context sensitivity rather than speaking sensitivity). My "coffee" and "green" examples illustrate how common nouns and adjectives may have very different reference in different contexts compatibly with what they "mean." To determine what is being said by "There is a lot of coffee on the table" or "The tree has green leaves now" in a particular context one needs to know the "meaning of the words," the implicit constraints on what can and cannot be said using those words, and to use good judgment to figure out what is being said in the given context; and, as Kant long ago said (if not in those terms), there isn’t a recursive rule for "good judgment" (not one we can formulate, anyway!). As Cavell argues at length in The Claim of Reason, our "atunement" to another, our shared sense of what is and what is not a natural projection of our previous uses of a word into a new context, is pervasive and fundamental to the very possibility of language—without being something that can be captured by a system of "rules."

There are many other sorts of examples, besides the two I just gave. Whether a surface is "flat" or not depends on what a reasonable standard of flatness is in the particular context. (Against this, Peter Unger once argued³⁷ that only a Euclidean plane is "literally flat." [So whenever I say that a table top is flat I am speaking "figuratively"!] Note that the semantics Unger proposed for "flat" violates the Principle of Charity to the maximum extent!) Whether playing a concerto is "difficult" depends on who is asking whom (imagine a child asking a violin teacher and a professional violinist asking another). Whether a sack of sugar weighs "one pound" depends on whether the question is asked in the supermarket or in the laboratory. But it is time to say what all of this has to do with philosophy.

Consider the following scenario discussed by Wittgenstein in Remarks on the Philosophy of Psychology:

96. A tribe that we want to enslave. The government and the scientists give it out that the people of this tribe have no souls; so they can be used without any scruple for any purpose whatsoever. Naturally we are interested in their language all the same; for of course we need to give them orders and to get reports from them. We want to know too what they say among themselves, as this hangs together with the rest of their behavior. But also we must be interested in what in them corresponds to our "psychological utterances," for we want to keep them capable of work, and so their expressions of pain, of feeling unwell, of pleasure in life, etc., etc., are of importance to us. Indeed, we have also found that these people can be used successfully as experimental objects in physiological and psychological laboratories; since their reactions—including speech-reactions—are altogether those of men endowed with souls [seelenbegabten Menschen]. I assume that it has been found that these automata can be taught our language instead of their own by a method very like our "instruction."³⁸

Here Wittgenstein imagines that, out of the motive of exploiting and enslaving certain people, we picture them to ourselves as being precisely like the "certain people" we hypothesized above. Remember?—

(SOULLESS) Certain people do not have any mental properties, but all of their physical properties are the same as if they did and their physical environments are the same.

When I encounter (SOULLESS) in the context of an argument like Kim’s, I confess that I am utterly at a loss what to make of it. I feel that I cannot conceive of a case in which I would use the description "these people do not have any mental properties, but all of their physical properties are the same as if they did and their physical environments are the same." And precisely because a context, a "speaking," is absent—and if we lack a context in which an utterance U makes sense, we do not provide one by just saying, "Now we are talking philosophy," or by saying, "Consider the possibility that U"!—precisely because a context is absent, I cannot say what it would be for (SOULLESS) to be true. Yet when I encounter 96 above in Wittgenstein’s Remarks on the Philosophy of Psychology, I suddenly realize that there is a case in which, sadly, we might very well use such a description (or the shorter, "soulless automata"). In that context I understand it all too well. I understand the utterances Wittgenstein refers to ("the people of this tribe have no souls"; in 97 he adds, "If anyone among us voices the idea that something must be going on in these beings, something mental, this is laughed at like a stupid superstition") in the sense of being perfectly able to follow what is going on.

Does that mean that I am able to assign a "possible world" to (SOULLESS) in the context Wittgenstein has constructed? to say what it would be for the version given out by "the government and the scientists" to be true? Not at all.³⁹ That a piece of propaganda does not, on reflection, describe a state of affairs we can make sense of does not mean that it does not function effectively as propaganda. (Wittgenstein was well aware of what the Nazis said about the Jews.) One of the morals we should draw from this case is that "understanding" itself is context sensitive; in one sense, I "understand" what it means to say the people of the tribe in question are "soulless automata," I understand what the words are doing, what effect they are intended to have and do have, but that does not mean I understand them in the context of Jaegwon Kim’s argument! Rather, to understand them in that context I would need to understand them independently of Kim’s argument, for that argument presupposes a prior intelligibility of the idea that certain people are "soulless automata." (Or, more precisely, it presupposes that if we reject reductionism, then we must regard (SOULLESS) as making sense—without telling us how to understand the supposed "sense.")

Note that if this is right the problem with Kim’s argument is not that the antecedent couldn’t be used to make an assertion that we understand. If I were to say of a group of bureaucrats "They are quite remarkable—they do not have any mental properties, but all of their physical properties are the same as if they did and their physical environments are the same, but they are really soulless automata," I might be saying something perfectly intelligible (and something that I couldn’t say just as well by saying, "They are absolutely devoid of judgment, sympathy, human understanding"). In that context the words that make up (SOULLESS) might have a perfectly clear content—but not one that is relevant to discussions of "mental causation." The problem with Kim’s antecedent is not with the words themselves;⁴⁰ it is that we do not know what Kim means by them.