Written for first session of 2001 Mind & Language seminar


An Introduction


Content and Its Role in Explanation


Stephen Schiffer



The word ‘content’ as it figures in the title of this seminar is a piece of philosophical jargon; it applies first and foremost to what is called propositional content, which is whatever is ascribed to mental states and speech acts via that-clauses, as in

Jean-Paul believes that existence precedes essence

Martin hopes that there is life after death

In uttering ‘Il pleut’, Odile was telling you that it was raining

"Content" may also be ascribed without a that-clause, as in

He asked me to feed his rodents while he’s in Miami

Betty wants to go to Lespinasse

It’s usually assumed, rightly or wrongly, that even in these latter cases the content ascribed is propositional, or, more cautiously, that, even if it’s not directly propositional, then we’ll at least be a short step from understanding what’s going on in non-that-clause ascriptions of content once we know what’s going in that-clause ascriptions.

The word ‘explanation’ in our title alludes in the first instance to commonsense propositional-attitude explanations. Such explanations are often explanations of propositional-attitude facts, as in

I concluded that you must really loathe New York because I heard you tell Hilda you’d prefer to live in New Jersey

Lester filed for divorce because he hated sleeping on crumbs and couldn’t get his wife to stop eating crackers in bed

There are also propositional-attitude explanations of non-intentional facts, as in

Why is Henrietta stretched out on the floor like that? Because she wants to find her collection of Frank Sinatra’s cigarette butts and thinks it’s probably under the bed

Henry’s face suddenly became red because he just then realized that Jerry Fodor overheard him say that Britney Spears was a better singer than any soprano the MET could produce; the redness of his face was a blush

Propositional-attitude explanations of non-intentional facts will have a special importance for us.

We wouldn’t be giving a seminar on content’s role in explanation if that topic didn’t signify an area of philosophical controversy. The primary source of the controversy can be put succinctly. On the one hand, propositional-attitude properties seem to play a certain causal-explanatory role; but, on the other hand, they seem to many, for one reason or another, to be unable to play the causal-explanatory role they seem to play. Let’s start by looking at some objections to the claim that propositional-attitude properties enter essentially into correct causal explanations of non-intentional facts.[1] After that I’ll take a brief critical look at some of the more salient positions to which a theorist might retreat if she’s skeptical of content’s playing a serious causal-explanatory role, and then I’ll conclude with some all-too-brief remarks about my own take on the questions at issue.

Before getting down to business, I should qualify the advertisement implicit in calling this squib an introduction to this year’s Mind & Language seminar. The qualification is that we shouldn’t expect every "thinker of the week" to speak directly to the issues now to be surveyed. As always, a considerable amount of leeway is permitted in how broadly or narrowly a given philosopher chooses to place on the announced topic so that she may subsume under it whatever she most wants to discuss with us.

I. Skepticism About Content’s Causal-Explanatory Role

Ava is sitting at a table at Le Cirque when of a sudden her right hand rises, arm attached, to a position just above her right ear. What explains the fact that Ava’s hand rose to that position? We appear to have at least two correct explanations. One is a commonsense propositional-attitude explanation: Ava wanted the check; she believed that the best way to get it was to signal the waiter in a way that would be instantly recognized as a request for the check and that raising her hand in the way question .... The other explanation is wholly non-intentional and speaks of electrochemical activity in Ava’s nervous system and the consequent stimulation of effector and receptor cells in her body and of their effects on her bodily movements. No one knows this explanation, but we know it exists. Both explanations appear to be causal explanations; that is, they appear to be telling us what caused the explanandum. If that appearance is correct and the ascribed propositional-attitude properties enter essentially into the proffered explanation that mentions them, then, as any theorist in this area would agree, the intentional explanation must be importantly related to, and somehow dependent on, the more basic non-intentional explanation. It’s at this point, however, that agreement tends to peter out. As already remarked, some philosophers think that what’s required isn’t satisfied and that, therefore, propositional-attitude properties don’t enter essentially into correct causal explanations of non-intentional facts. Philosophers who hold this skeptical view usually have two parts to their story: an objection to the claim that propositional-attitude properties play a causal-explanatory role and an account of them and their use in ‘because’ statements that’s designed to remove the patina of counter-intuitiveness from the skeptical claim. What follows next is a critical look at some of the alleged objections that have helped to move the skeptics to their skepticism about content’s causal-explanatory role (another way of referring to the causal-explanatory role of propositional-attitude properties).

  1. The reasons-aren’t-causes objection. In the fifties, beginning shortly after Wittgenstein’s death, there were several books and articles by American and British philosophers many of whom considered themselves to be developing lines of thought that were either implicit in the later Wittgenstein or at least ones to which one would naturally be led when philosophizing in the style of his Philosophical Investigations.[2] Some of these philosophers—e.g., Gilbert Ryle, Stuart Hampshire, H. L. A. Hart, William Dray, Elizabeth Anscombe, Anthony Kenny, A. I. Melden, R. S. Peters, and P. Winch[3]—held positions which committed them to the following two claims, whose conjunction entails that propositional-attitude explanations aren’t causal explanations, and that therefore content doesn’t play a causal-explanatory role:
    1. Propositional-attitude explanations are causal explanations only if the propositional-attitudes mentioned in them are not merely the reasons for which the agent did what she did, but also the causes of her doing what she did.
    2. But reasons can’t be causes because (i) as Hume taught us, causes must be contingently connected to their effects, whereas (ii) the reasons for which an agent performed an action are conceptually connected to the action in a way that precludes a contingent connection.

    Now (1) is plausible. Just as the explanation one gives in saying

    The match lit because it was scratched

    wouldn’t count as a correct causal explanation unless the match’s lighting was caused by its having been scratched, so the explanation one gives in saying

    Ava moved her hand in a certain way because she wanted to get the check and believed that the best way to get it was to move her hand in that way

    wouldn’t count as a correct causal explanation unless Ava’s belief and desire caused her hand movement. The Humean claim made in (2)(i) in its most plausible formulation is that the causation relation is one that holds contingently: the fact that c caused e is never a logical or conceptual or metaphysical necessity; there’s always a sense in which c might have occurred without causing e. This most plausible formulation is likely to seem pretty plausible, so the crux of the neo-Wittgensteinian line is its claim that the relation expressed by ‘x was the reason for which y did z’ is "conceptual" in a way that precludes reasons from being causes. This is where the neo-Wittgensteinian line falls apart. Two claims were made about the "conceptual connection." First, they said that the reason contains a "reference" to the action, in the way that Henrietta’s wanting to get drunk contains a reference to her getting drunk. Second, they said that when one tries to complete a propositional-attitude explanation sketch, one gets something that looks pretty close to being analytic, such as what you get when you move from

    She went into the kitchen because she wanted a beer

    to what they would call its completion, something like

    If a person desires result R, believes that the best way to get R is to do Y, has no stronger competing desires, knows how to do Y, is not physically or psychologically prevented from doing Y, and doesn’t change her mind, then, ceteris paribus, she will do Y. Mary desired to drink a beer, believed that the best way to accomplish that was to go into the kitchen and get one from the fridge, knew how to get into the kitchen, was in no way prevented from doing that, didn’t change her mind, and cetera were paria. So, Mary went into the kitchen.

    One problem with this second conceptual-connection argument is that even if there were connections of the kind suggested, it would do nothing to show it’s not the case that when you did something for a certain reason, you might have had that reason without doing what you did. As Donald Davidson emphasized in "Actions, Reasons, and Causes," if the cause and effect are described under descriptions that make it analytic that a thing falling under the cause description would be followed by a thing falling under the effect description, this wouldn’t show that it wasn’t a contingent fact that the cause caused the effect. For example, rain clouds would contingently cause rain even if it were analytic that rain clouds cause rain, for it would still be the case that those things that were in fact rain clouds might not have been; it’s merely a contingent truth about those clouds that they have the property of being rain clouds. A second problem with the just stated conceptual-connection argument is that the "ceteris paribus" generalization isn’t analytic unless the "ceteris paribus" qualification simply comes to "unless she doesn’t do Y," which would render it useless as a premise in an explanation of Mary’s going into the kitchen. A third problem is that no such analytic generalization either occurs, or needs to occur, in anything that could properly be called a complete commonsense propositional-attitude explanation. As for the first of two alleged cause-precluding conceptual connections, it’s simply false that reasons refer to the particular actions they appear to be causing: if I want to win the sweepstakes but don’t win it, no one will accuse ‘Schiffer wants to win the sweepstakes’ as suffering from reference failure. The important point, however, is that while it’s true that, say, ‘Henrietta wants to get drunk’ involves the concept of getting drunk, that gives no reason whatever to think that Henrietta might have had that desire and not gotten drunk.

  2. The strong-reduction objection. Let’s stipulate that the property of being F strongly reduces to the property of being G only if the property of being F = the property of being G. Then the present objection to the commonsense view of propositional-attitude explanation may be stated as the following two-part claim:
    1. A propositional-attitude property plays an essential role in a correct causal explanation of a non-intentional fact only if it strongly reduces to a non-intentional property that figures essentially in a correct non-intentional causal explanation of the same fact.
    2. No propositional-attitude property strongly reduces to any non-intentional property.

    The usual argument for (2) is one that is also taken to apply to mental properties that aren’t propositional-attitude properties; this is that such properties are multi-realizable, may, that is, be "realized" by various non-intentional properties without being identical to the disjunction of them. This invites the counter response of David Lewis, Jaegwon Kim and others that what really enters into correct propositional-attitude explanations isn’t, say, the property of believing that such-and-such but rather a contextually relevant species-specific property such as, so to say, the property of human-believing that such-and-such, which property, they further claim, is arguably not subject to multiple realization and can be identified with a causal-explanatory non-intentional property and, ultimately, with a property statable in the language of fundamental physics. "In this way," Kim argues, "multiply realized properties are sundered into their diverse realizers in different species and structures, and in different possible worlds."[4]

    In fact, however, there is a much more compelling case for (2) than any based on multiple realization. For even if multiple realization is the best reason for supposing that non-intentional properties like pain can’t be identified with any non-mental property, there is a much better reason that applies only to propositional-attitude properties. The point is very simple: no property in an underlying non-intentional explanation will have what it takes even to be a candidate for strongly reducing a propositional-attitude property. For consider the property of believing that there are no unicorns. The most plausible account of this property sees it as a composite property composed of the belief relation and the proposition that there are no unicorns, where ‘proposition’ for now is simply a label for whatever things are the references of that-clauses. Given this, any non-intentional property that could strongly reduce the composite belief property would itself have to be a composite property one part of which was identical to the belief relation, the other part identical to the proposition that there are no unicorns.[5] But just try finding the non-intentional relation to the proposition that 12 + 12 = 4 in the physical explanation that subserves the fact that little Johnny wrote ‘4’ because he was trying to answer the question "What’s the sum of 12 + 12?" and he believed that 12 + 12 = 4. If correct propositional-explanations of non-intentional facts are causal explanations that essentially involve propositional-attitude properties, then the non-intentional explanations that underly the propositional-attitude explanations won’t be statable using composite non-intentional predicates that express relations to propositions. We can dismiss out of hand the idea that propositional-attitude properties strongly reduce to properties involved in non-intentional explanations. It further follows that propositional-attitude properties don’t reduce to functional relations that require physical realizations; for such realizations would have to be physical relations to propositions. This leaves open whether propositional-attitude relations can be identified with topic-neutral relations such as the ones implied by Jerry Fodor’s theory of asymmetric-dependence;[6] but it seems clear to me on independent grounds that no such strong reduction can be made to work.[7]

    We’re therefore forced to conclude that if (1) is true, then propositional-attitude properties don’t enter essentially into correct causal explanations of non-intentional facts. With so much riding on (1), what’s to be said in its defense?

    Some theorists accept (1) because they accept the covering-law model of explanation, which holds that explanations are deductive or probabilistic inferences at least one premise of which is a law—a non-probabilistic law in the deductive case, a probabilistic law in the probabilistic case. Explanations that take the form of deductive inferences are called deductive-nomological (DN) explanations. An example of a DN explanation of a singular fact would be:

    Water boils at 212° F at sea level.

    This water was heated to 212° F at sea level.

    Ergo, this water boiled.

    In the case in which what’s being explained is a theory, where that’s taken to be a conjunction of laws, then the explanation takes the form of a deduction of that theory from premises one of which is a more basic theory and the other of which is a conjunction of "bridge laws" linking the theoretical terms of the theory being explained to those of the explaining theory. These bridge laws effect a reduction of the theoretical properties of the derived theory to their counterparts in the theory from which the derived theory is derived. As Patricia Churchland puts it, "intertheoretic reduction is an instance of DN explanation, where what is explained is not a single event but a law or set of laws."[8] Those who hold this view also hold that all true non-basic theories ultimately reduce to fundamental physics.

    But what does "reduction" come to in this context? The dominant answer among covering-law theorists seems to be what I’ve called strong reduction: the reduced properties must be identical to the properties with which they’re correlated in the bridge laws. The rationale for this construal is succinctly put by Jaegwon Kim (where ‘F*’ is the counterpart in the reducing theory of ‘F’ in the reduced theory):

    [Identities of the form property F = property F*] are essential to the ontological simplification that we seek in theory reduction, for they enable us to dispense with facts involving F and G as something in addition to facts involving F* and G*. They also allow us to give simple answers to potentially embarrassing questions of the form, "But why does property F correlate this way [i.e., in the way of the biconditional ‘(x)Fx º (x)F*x’] with property F*" Our answer: "Because F just is F*."[9]

    The sort of rationale Kim offers for why the covering-law theorist needs strong reduction can motivate strong reduction independently of the covering-law model. It’s what lies behind my own claim in Remnants of Meaning[10] that if there are mental properties, then they must be identical to properties that are intrinsically specifiable in non-intentional terms, and it’s what lies behind Kim’s advocacy of what he has called

    "the problem of causal/explanatory exclusion": For any single event, there can be no more than a single sufficient cause, or causal explanation, unless it is a case of causal overdetermination.[11]

    In other words, the problem for the strong reductionist who supposes that something can have both a physical and a mental cause at a given time is that the physical cause "threatens to exclude, and preempt, the mental cause."[12] In Remnants, I argued that even if one assumes that propositional-attitude state tokens are identical to physical state tokens, one should also hold that propositional-attitude properties are identical to physical properties, given that propositional-attitude properties really do play an indispensable causal-explanatory role in correct causal explanations of non-intentional facts.[13] The problem that’s apt to make one a strong reductionist is the problem of explaining the relation between an apparently causally relevant propositional-attitude property and the concurrent underlying causally essential neurophysical property. When one surveys the explanatory options, one may be tempted to think one has an argument by elimination for the strong reduction of propositional-attitude properties to properties intrinsically expressible in naturalistically kosher non-intentional terms. I won’t go through the moves of such an argument by elimination, except, with some chagrin, to tell you the reasons I gave for rejecting the idea that we could explain the connection by saying that, while intentional properties weren’t identical to properties intrinsically specifiable in naturalistic terms, the former properties nevertheless supervene on the latter. Here, before someone quotes it back to me, is what I wrote in Remnants of Meaning:

    G. E. Moore was a non-naturalist in ethics: he held that moral properties could not be identified with natural properties; and he held, on the positive side, that they were simple, irreducible, unanalyzable, non-natural properties, and that, being non-natural, they were discerned through a special faculty of moral intuition. Tough-minded physicalist types (including many Logical Positivists) agreed that moral properties could not be reduced to natural properties..., but had no sympathy at all with Moore’s positive thesis, which postulated a realm of non-natural properties and facts. These properties, it was felt, could not be made sense of within a scientific world view; they were obscurantist and produced more problems than they solved. At the same time, philosophers who abhorred Moore’s irreducibly non-natural properties knew that he also held this thesis about them: that it was not possible for two things or events to be alike in all physical respects while differing in some moral property. No one thought that Moore’s positive theory of moral properties was in any way mitigated by this further supervenience thesis. How could being told that non-natural moral properties stood in the supervenience relation to physical properties make them any more palatable? On the contrary, invoking a special primitive metaphysical relation of supervenience to explain how non-natural moral properties were related to physical properties was just to add mystery to mystery, to cover one obscurantist move with another. I therefore find it more than a little ironic, and puzzling, that supervenience is nowadays being heralded as a way of making non-pleonastic, irreducibly non-natural mental properties cohere with an acceptably naturalist solution to the mind-body problem.... Supervenience is just epiphenomenalism without causation.[14]

    I now find this unsatisfactory. There’s a weak and a strong response to the suggestion that we have in the supervenience relation between intentional and non-intentional properties a necessitation relation holding between disparate properties whose holding can be discovered only a posteriori. The weak "objection" is merely that the suggestion owes something more to be acceptable—namely, some sort of explanation or demystification of how there could be that kind of supervenience state of affairs. That point seems right to me, notwithstanding the vagueness of "explanation" or "demystification" in this context. The strong objection is that no such explanation or demystification can be given. The trouble with this strong objection is that no reason has yet been given to think it’s true, other than that no such explanation or demystification has so far been given. Related to this is that the quoted passage puzzles over a non-pleonastic property supervening on a distinct non-pleonastic property, but what if all properties are "pleonastic," entities whose existence is secured by conceptual truths such as the proposition that if Fido is a dog, then Fido has the property of being a dog, entities whose nature is determined by the linguistic and conceptual practices by which they’re introduced into our ontology? Perhaps then it won’t be so difficult adequately to account for the way the intentional depends on the non-intentional. There will be a bit more on this later.

    A final motivation for strong reduction used to be urged on me by Hartry. In Remnants §6.5, after rejecting strong reduction, I appealed to the fact that the conceptual roles of our propositional-attitude concepts meshed with our underlying physical workings to secure the reliability of our propositional-attitude predictions and explanations. To this Hartry responded that strong reduction, or something pretty close to it, was needed precisely to explain the mesh. I responded to this objection in "Physicalism,"[15] but my argument there was too bound up with a nominalism I no longer accept. In any case, I don’t accept Hartry’s objection. It’s enough for now that we see that reduction isn’t needed to explain the reliability of our propositional-attitude-based predictions, the predictions we make about what others will do based on the propositional-attitudes we correctly ascribe to them. When we correctly predict what someone will do we rely on our beliefs about how people come to have their propositional-attitudes and about how a person’s propositional attitudes affect what she will do. What Hartry’s challenge primarily demands is an account of what determines the content of our propositional-attitude states that’s consistent with what’s required non-intentionally in order to account for how we form our propositional-attitudes and for how our propositional attitudes lead to action. There are several coherent stories to tell about the determination of content all of which cohere with underlying mechanisms that account for how certain sorts of predictions based on correct ascriptions of propositional attitudes tend to be reliable. I see no reason to suppose that anything even approximating reduction is needed to achieve this coherence, and (not that it’s relevant, but) Hartry’s present account of propositional attitudes forces him to reject his former mesh objection.

  3. The wide-content objection. This familiar objection goes like this:
    1. If propositional-attitude properties played a causal-explanatory role in correct explanations of non-intentional facts, then they would supervene on what’s in the head—i.e., on the non-intentional features of one’s neural states which would be shared by any of one’s Doppelgängers, regardless of the Doppelgänger’s external environment. As Stephen Stich put it:
    2. … any state or property properly invoked in a psychological explanation should supervene on the current, internal, physical state of the organism. Thus, a pair of Putnamian doppelgangers, being molecule for molecule replicas of one another, must share all the same explanatory psychological states and properties.[16]

      Jerry Fodor expressed the same view, only wrapped in a dire warning:

      Causal powers supervene on local microstructure. In the psychological case, they supervene on local neural structure. We abandon this principle at our peril; mind/brain supervenience ... is our only plausible account of how mental states could have the causal powers that they do have.[17]

    3. Propositional-attitude properties typically don’t supervene on what’s in the head.

    Fred Dretske’s gloss of the problem is representative of those who accept (1) and (2):

    The prevailing wisdom among materialists is that even if we can give an otherwise creditable account of Intentionality, the properties that give a structure its intentional identity, the facts that underlie its content ... will ... turn out to be explanatorily irrelevant. Even if some events have a meaning, and even if ... they have an impact on their material surroundings, the fact that they mean what they do won’t help explain why they do what they do.

    This doctrine about what Dennett calls the impotence of meaning should not be taken to imply that the objects having meaning are causally inert. It only means that it is not their intentional properties, their content or meaning, from which they derive their causal powers. Though a brick was made in Hoboken, it gets its power to break windows from its velocity and mass, not from its having been made in Hoboken. By the same token, although events in the brain, those we might want to identify with a particular thought about Hoboken, are about Hoboken, their power to stimulate glands and regulate muscle tension—and thus to control behavior—derive, not from what they mean, not from the fact that they are about Hoboken, but from their electrical and chemical properties....

    This ... problem about the causal role of meaning ... arises from the fact that meaning ... supervenes on a set of facts that are different from the facts that explain why a structure (with that meaning) has the effects it has. Let M be the set of properties and relations in virtue of which event E means what it does. Let C be the set of properties and relations in virtue of which E causes what it does. M is not identical to C.... What gives sounds the meaning they have is not what confers on them the power to shatter glass and rattle eardrums; what makes ... a picture of ... my Uncle Harold is not what gives it the power to reflect light in the way it does.... The same can be said abut those physical events, processes and structures in the brain that are supposed to be a person’s thoughts, hopes, and desires. The fact that something in the head, a thought for instance, has truth conditions doesn’t help explain the thought’s effect on motor output. Thoughts, the things with content, make a difference ... but the fact that they are thoughts ... is not relevant to the difference they make.[18]

    No one should question (2), the claim that content doesn’t supervene on what’s in the head, and hardly anyone nowadays would. For suppose that the things to which your Twin Earth Doppelgänger applies the word ‘dog’ look and behave exactly like dogs but are actually of a different zoological species and thus aren’t dogs. Nevertheless, it’s not plausible that only one of you expresses a true belief when she says ‘There’s a big dog growling at me’. The belief your Doppelgänger expresses has as much right to be true as the one you express, thus showing that in this case propositional-attitude properties don’t supervene on what’s in the head.

    But why accept (1), why take this to show that propositional-attitude properties don’t play a causal-explanatory role?

    One answer is more or less explicit in Dretske. He’s saying two things: first, that propositional-attitude properties like believing that a car is coming aren’t "causal powers" of the states that have them, and, second, that it follows from their not being causal powers that they’re also "explanatorily irrelevant," which I take to mean that though we may cite such properties in giving psychological explanations, the reason we cite them isn’t that they really occur in the explanations we’re giving. Now, ‘causal power’ is a term of art and we’re entitled to ask what it means. It’s hard to say what it means; it’s not defined, so we have to garner its meaning by figuring out when those who use it are willing or unwilling to apply it. Still, it does seem that the following is taken to be sufficient for a property F’s not being a "causal power" of c with respect to its causing e:

    Very well, let’s agree to use ‘causal power’ in such a way that satisfaction of the foregoing conditions is sufficient for not being a causal power. Why should a propositional-attitude property’s not being a causal power in that sense be taken to show that it’s explanatorily irrelevant? An argument is needed to get us from "Such-and-such isn’t a causal power" to "Such-and-such is explanatorily irrelevant." What is it? While Dretske’s trying to answer that question, we should also demand that he answer the following question:

    It’s clear that by virtue of its meaning "Oops, I did it again," the sequence of sounds produced by the soprano was neither a causal power with respect to its causing the glass to break nor relevant to the explanation of why the glass broke. This is clear because we would without any hesitation reject the explanation that the glass broke because the emitted sounds meant "Oops, I did it again." Even without any philosophical theorizing, that’s just not an explanation anyone would countenance. The fact that the sounds had that meaning is as explanatorily irrelevant as the fact that the soprano who emitted them was wearing purple nail polish. Yet the situation is radically different as regards the explanation of Ava’s stepping back to the curb. Without any philosophical theorizing, we would without any hesitation accept the claim that she stepped because she believed that a car was coming, even though, as it’s now revealed, the property of being a belief that a car is coming isn’t a "causal power" of the neural state that at the relevant time caused her stepping back. What explains this difference between the property of meaning "Oops, I did it again" and the property of believing that a car is coming?

    Maybe a good strategy would be to assume as a provisional hypothesis that the reason we’re happy to say that Ava stepped back because she believed that a car was coming but not that the glass broke because the sounds emitted by the soprano meant "Oops, I did it again," even though neither meaning "Oops, I did it again" nor being a belief that a car is coming is a causal power, is that the second property, but not the first, is explanatorily relevant. If the account arrived at in this way coheres well with our firm intuitions about when something is essential to a correct causal explanation, then we ought to conclude that the inference from "Not a causal power" to "explanatorily irrelevant" is a non sequiter. In any event, we’ve yet to see a good reason for thinking the inference is a sequiter.

    There are two further problems for the line of thought under consideration. One is that we’ve so far been proceeding as though Ava’s belief state token were identical to the neural state token that was a cause of her bodily movement; but this assumption is controversial. As I see it, it’s at best indeterminate whether propositional-attitude state tokens are identical to physical or topic-neutral state tokens. In order for an identity statement to be determinately true, there must be a non-trivial account of what makes it true, as for example the identity of heat with molecular motion is accounted for by the fact that we fix the reference of ‘heat’ as whatever causes in us the sensation of heat and then discover that its molecular motion which plays that causal role. Such an account will typically, if not always, determine an epistemic route implicit in the concepts of the thing or things in question by virtue of which the truth or falsity of the identity claim could be discovered by relevantly placed agents. No such account or route is available for a claim that such-and-such propositional attitude = such-and-such neuro-physical state token. In no way, however, should this make one conclude that propositional attitudes can’t be causes of non-intentional states—unless one’s also prepared to conclude that propositional attitudes can’t be causes or effects of non-intentional states, as when light reflected from an oncoming rock enters your eyes and causes you to have the belief that a rock is coming towards your head, which belief in turn causes you to duck. Yet the present line of thought supporting (1) would imply that Ava’s belief that a car was coming played no role either in causing her bodily movement or in the explanation of her stepping back to the curb, unless that state token were identical to a neural state token. That should strike one as highly counterintuitive.

    The second further problem is that although those sympathetic to the Dretskian line say that content would be explanatorily relevant if only it supervened on what’s in the head—precisely what’s suggested by the wording of (1) of the wide content objection—that’s in fact not a view to which they’re entitled. What’s really implied by the Dretskian argument for (1) is that it’s entirely irrelevant whether or not content supervenes on what’s in the head. For if the argument were sound, it would show that, whatever the supervenience situation was, the only explanatorily relevant property would be that neurophysical property that was the causal power. This effectively cuts the legs out of those who accept the present objection to "wide" content’s playing a causal-explanatory role and go on to claim that what’s needed is a notion of "narrow" content. What they ought to conclude is that no kind of content can play a causal-explanatory role.

    A second reason some have found for accepting (1) is motivated by the following sort of thought experiment, this one by Brian Loar:

    … suppose that I do not know whether in Bert’s linguistic community "arthritis" means arthritis or tharthritis, but that I know all the relevant individualist facts about Bert [i.e., intentional facts that supervene on what’s in Bert’s head]. I read in his diary: "I fear I have arthritis, and so today I have made an appointment with a specialist." It is difficult to accept that we do not fully understand the psychological explanation given here, despite our not being in a position to produce the correct that-clause.[19]

    Loar’s point is that since you can know the explanation of Bert’s going to a specialist without knowing the propositional contents of the beliefs and desires involved in it, content is irrelevant to propositional-attitude explanation and hence doesn’t play a genuinely causal-explanatory role. But we should be suspicious of Loar’s example when we realize that he might just well have said the same thing about our reaction to Bert’s diary entry ‘I fear I have aldosteronism, and so today I have made an appointment with a specialist’, when we have no idea what medical condition aldosteronism might be. A better response to such examples is that commonsense explanations are answers to contextually implicit why questions, and its easy to imagine contexts in which the diary entry would tell us all we’d want to know about Bert’s behavior. You might as well conclude that a computer’s hardware properties don’t play a causal-explanatory role because you could be satisfied by being told that your computer isn’t working properly because of a defect in its motherboard, even though you have only the foggiest idea of what a motherboard is.

    A third reason for accepting (1) is offered by Daniel Dennett, but one can find it repeated by Fodor and others:

    If psychology is going to be science, it had better not posit mysterious action-at-a-distance [my emphasis], so a principle of the "supervenience" of the psychological on the physiological must be honored: the brains of organism differ whenever their minds differ.[20]

    The main problem with this (there are others) is that action-at-a-distance requires distant but unconnected causes, and an absence of action-at-a-distance is entirely consistent with causes having properties that relate them to distant things.

  4. The no-law objection. This can be put in the following way.
    1. Propositional-attitude properties play a causal-explanatory role only if they figure in true causal laws.
    2. Propositional-attitude properties figure in no such laws.
    3. So, propositional-attitude properties don’t play a causal-explanatory role.

    Many who want to resist the conclusion—e.g., Jerry Fodor when he’s not accepting the conclusion—deny (2) by claiming that while there are no true strict, or exceptionless, propositional-attitude laws, there are true ceteris paribus propositional-attitude causal laws. I doubt that there are such laws, and I refer you to my article "Ceteris Paribus Laws," where I try to show why.[21] The same issue of Mind contains Jerry Fodor’s "You Can Fool Some of the People All the Time, Everything Else Being Equal: Hedged Laws and Psychological Explanations,"[22] which replies to my paper. When the issue containing our papers appeared, Fodor said to me that the editors of Mind forgot to say who won. I’m going to assume that I did, so that the question now is what can be said for premise (1). Why, that is, think that causal explanations require causal laws?

    One who thinks causal explanations require laws is mostly likely an adherent of the covering-law model of explanation and holds tenaciously to the view that if there were no ceteris paribus laws there would be no special-science laws, hence no special science explanations, and hence no special sciences. But when we look at typical explanations in special sciences such as biology or geology, one’s apt to wonder whether the covering-law model is the best model for the explanations taken to be correct in those sciences. I challenge anyone to find geological laws in the preferred explanation in plate tectonics of how the Himalayas were formed, or to find biological laws in the explanation of how the HIV penetrates host cells. What one finds are law-free stories about the mechanisms by which certain things happen. To be sure, laws may well be needed to account for how the mechanisms work, but these laws will be laws of more basic sciences, not laws of the special sciences offering the geological or biological explanations. And many correct commonsense explanations seem far removed from a nomological component, as in the following example which Terry Horgan attributes to Barry Loewer:

    … there are examples of causal transactions in which the cause and the effect have properties which evidently are not connected by even a [ceteris paribus] generalization, but which seem explanatorily relevant anyway. Suppose, for instance, that Barry’s noticing a flower shop causes him to remember that tomorrow is his wife’s birthday. The properties being a noticing that there is a flower shop yonder and being a remembering that tomorrow is one’s wife’s birthday, certainly appear explanatorily relevant to the causal transaction; yet generalizations like

    Ceteris paribus, a (married) man who notices that there is a flower shop yonder will remember that the following day is his wife’s birthday

    seem just false.[23]

  5. The eliminativist objection. This objection enjoys a simple statement: propositional-attitude properties can’t play a causal-explanatory role because there are no propositional attitudes. There are at least two reasons philosophers have claimed for believing that there are no propositional attitudes (its their oxymoron, not mine). The first is associated with Paul and Patricia Churchland, though it has precursors in Paul Feyerabend and Richard Rorty.[24] As stated by the Churchlands, the idea is that the notion of a propositional attitude is implicitly defined by its role in commonsense psychology and, as so defined, it requires that our brains enjoy a certain functional organization. Since, they further claim, there’s very good reason to suppose that our brains don’t have that functional organization, there’s very good reason to suppose there are no propositional attitudes and that, therefore, the theory that defines the notion of them is false. Propositional attitudes and propositional-attitude theory must go the way of phlogiston and phlogiston theory. The other reason for denying that propositional attitudes exist is that the truth of propositional-attitude reports—e.g., the claim that Ralph believes that the earth is flat—require abstract entities called "propositions" to be the references of that-clauses, and propositions don’t exist. This is reminiscent of Hartry’s claim that arithmetical statements such as ‘1 + 1 = 2’ can’t be true since there are no numbers for numerical singular terms like ‘1’ and ‘2’ to refer to. Mark Crimmins, who will occupy the hot seat later this semester, takes this line, and, just as Hartry tries to lighten the counter-intuitiveness of his theory by claiming that arithmetic statements are at least true in the "fiction of arithmetic," so Crimmins holds that propositional-attitude statements are true in the fiction of, I guess, the story of propositional attitudes. And just as Hartry needn’t reject physical explanation that try to refer to numbers, so Crimmins needn’t deny that correct explanations are given by propositional-attitude explanations. It’s just that those propositional attitudes don’t enter essentially into those correct explanations.
  6. To the Churchlands I say that, while I can’t speak for them, I know that nothing could convince me that I don’t have any beliefs. If I were to learn that, like a chocolate Easter bunny, I was hollow inside, that would just tell me that you don’t need a stuffed head to have beliefs. To Crimmins and other fictionalists I say that their position is not well motivated because they don’t have the good reasons they think they have for doubting that propositions exist. If their worry is how to account for knowledge or reliable beliefs about propositions, then they should appreciate that given the right conception of propositions, it’s a conceptual truth which we know a priori that propositions exist.[25]

  7. The psychology-doesn’t-need-them objection. In his well-known and influential paper "Mental Representation,"[26] Hartry wrote:
  8. If the task of psychology is to state (i) the laws by which an organism’s beliefs and desires evolve as he is subjected to sensory stimulations, and (ii) the laws by which those beliefs and desires affect his bodily movements, then semantic characterizations of beliefs and desires are irrelevant to psychology: one can state the laws without saying anything at all about what the believed or desired sentences mean, or what their truth-conditions are or what their subject matter is.[27]

    This was then taken by Field and others to show that content doesn’t play a causal-explanatory role, doesn’t, that is, occur essentially in whatever correct explanations are implied by our commonsense propositional-attitude ‘because’ statements. I have a major and a minor problem with this objection.

    The major problem is that I don’t see how the truth of the quotation is supposed to show anything about whether content occurs essentially in correct propositional-attitude causal explanations. The proposition expressed by the quotation in no way entails that content doesn’t play a causal-explanatory role. An additional premise is needed. What is it? What’s the necessary condition for content’s playing such a role which the truth of the quotation shows can’t be satisfied? It can’t be a point about "explanatory exclusion" per se. If, as many would claim, the result of replacing ‘psychology’ with ‘neuroscience’ in the quotation would again yield a truth, nobody would take that to show content doesn’t occur essentially in correct causal explanations. So why should putting ‘psychology’ back make a difference? A first thought (not Field’s) might be that the real point is that the propositional-attitude explanations we give always implicitly offer a correct "syntactic" explanation of the kind to which Field is alluding. This, however, has no credibility. Forget that it ignores propositional-attitude explanations of propositional-attitude facts; it also seems to forget that our propositional-attitude explanations are almost never trying to explain a mere bodily movement, and that when we have in mind the sorts of non-intentional facts we actually use content to explain, it will become painfully clear that anything that might be called a "syntactic," or non-intentional computational, explanation of the same fact would take about a bizillion years to state and isn’t even remotely available to anyone but God, and She isn’t about to clue us in. For example, you ask why Harold is in Akron, and I explain that he’s being pursued by his ex-wife and thinks it won’t occur to her to look in Akron. What would be the correct "syntactic" explanation of the fact that Harold is in Akron on the assumption that it exists? Well, it would require a very complete computational theory of Harold, one that could be used to explain each of his intentional bodily movements, a theory that is very, very far from being known by anyone, layman or scientist. And that would be just the beginning. We would also need both a virtually complete state description of the physical world at some designated "initial position" (e.g. the time when Harold first learned that his ex-wife was after him) plus a complete physics. While it’s not unreasonable to think that such an explanation must in some way underlie the explanation I actually give in citing Harold’s beliefs and desires, it’s insane to think I was "implicitly" giving any such non-intentional explanation. Of course, Hartry wouldn’t deny this. But if he thinks we’re doing something right other what one would pretheoretically think we’re doing right, what is it? Presumably, it would be something to the effect that when I tell you that Harold’s in Akron because he wants not to be found by his ex-wife and believes she won’t think of looking for him in there, what I’m really offering is something like an "explanation sketch" of an incomplete computational-cum-physical explanation. Yet even this seems very implausible. I would hazard that the most one could be doing is implicitly saying that there is a correct such explanation. But, first, ordinary people aren’t in a position to do that, implicitly or explicitly, and, second, even if one were in a position to make that extremely boring and trivial existential generalization, why on earth would we bother to say anything? Surely, anyone to whom I was trying to tell that there was a correct computational-cum-physical explanation of Harold’s being in Akron would already believe that there was and would have no interest in being told it if she didn’t already believe it. Besides, none of this even begins to explain why we use intentional idioms to give these non-intentional explanation sketches. Hartry, as you might have suspected, has something to say in response to these questions, but since his readings for his session on February 20th will include that response, I won’t pursue this stuff any further now. I should, however, remind you that my point isn’t that Hartry doesn’t have a good reason to deny that content plays a serious causal-explanatory role; my point has been that I don’t see how the quoted passage itself gives us any reason to deny it.

    My minor problem with the psychology-doesn’t-need-them objection, stated here without elaboration or defense, is that I think it’s quite problematic that there are correct "syntactic" explanations of the kind Hartry had in mind.

II. Withdrawal Positions

Suppose a theorist takes one or more of the foregoing objections to be valid. To where in conceptual space should she withdrawal now that she denies that propositional-attitude properties don’t occur essentially in any correct causal explanation given by a commonsense propositional-attitude explanation?

  1. Extreme withdrawal. This is the position of the Churchlands and perhaps of no one else. It holds that we should react in the same way scientists who tried to explain combustion in terms of phlogiston reacted when they discovered there was no such thing as phlogiston: they simply realized that they were wrong in the explanations they gave and stopped giving them, except perhaps in some unserious instrumental vein, since, they probably allow, our commonsense notions at least have some predictive value. No comment.

  2. Withdrawal to science. Embarrassed by the stridency and counter-intuitiveness of their earliest pronouncements about content, which seemed to imply that our commonsense propositional-attitude explanations were on a par with appeals to phlogiston in explanations of combustion, theorists like Stich and Fodor retreated to claims about what’s needed for scientific, as opposed to commonsense, psychological explanations of behavior. OK, but where does that leave the idea that content plays an essential role in correct commonsense causal explanations? If that’s still denied, what’s the relevance of the claim about scientific explanation? If it’s no longer being denied, then this big change of mind means we’ve no business discussing this view here.

  3. Withdrawal to fictionalism. This response begins with the claim that while commonsense propositional-attitude explanations are never true (since there are no propositions), they are "true in the fiction of propositional-attitude theory" (I haven’t yet seen Crimmins’s contribution to this seminar, but for all I know his response might begin with this fictionalist response). Yeah, and where, pray tell, might I buy that novel, which I haven’t yet had the pleasure of reading? Also, this theorist must of course explain why we should care about truth in this particular fiction. One partial answer that might be given is one presently to be considered on its own—namely, in giving the false fictional explanations we’re implicitly giving correct non-intentional explanations or explanation sketches. Alternatively, or in conjunction with the true-in-the-false partial answer, she might say the reason we give explanations that are true only in the intentionality fiction is that the fictional explanations have a certain instrumental value; they allow us to influence and to predict one another’s behavior in ways that are very useful to us. But then we must be told how the fiction can do this, and one worries that any serious attempt to do this will invite the question "Do you really think that your reasons for accepting that (propositions don’t exist and propositional-attitude reports can’t be true unless they do exist) outweigh all the reasons you have for thinking that there are true propositional-attitude causal explanations?"

  4. Withdrawal to "narrow content." There was a time, about fifteen to ten years ago, when most of those who advanced the "wide content" objection went on to claim that what psychology needed was a notion of narrow content, a notion of content which, unlike the wide content ascribed by that-clauses, did supervene on what’s in the head. This view seems no longer to be in favor, and for good reason, I would say. Some of the things being called narrow content, such as computational roles, had no business being called content, so that the claim of these theorists was just the claim, in disguise, that the only correct psychological explanations of behavior were non-intentional, much like Hartry’s suggestion in "Mental Representation." Fodor suggested an intentional notion but it was hard to make both precise and plausible, and it wasn’t clear that it really could be narrow. Other characterizations of narrow content were quickly seen to be flawed. The main objection I would make, however, is that the usual arguments advanced to show that wide content didn’t play a causal-explanatory role would, if sound, also show that no notion of narrow content could play a causal-explanatory role. For what those arguments really imply is that the only properties of a cause which genuinely play a causal-explanatory role are those of its properties that are the causal powers by virtue of which the cause caused its effect.[28] Any other property of the cause would be epiphenomenal if it wasn’t a causal power, regardless of whether it happened to supervene on a causal power.

  5. Withdrawal to another kind of explanatory role. Some have claimed that although content doesn’t play a causal-explanatory role in correct causal explanations, it does play an explanatory role. For a long time, mostly in the seventies, the dominant view was that mental properties were functional properties and that, therefore, psychological explanations were functional explanations. Suppose you ask why a certain boiler exploded and someone explains that it was caused by an event in the boiler that had some unknown physical property F that was related to such-and-such inputs to the boiler and such-and-such other boiler states and conditions in such a way that an event’s having F would be a necessary part of a causally sufficient condition for causing an explosion. You would have been given a kind of causal explanation, but the second-order functional property used in the explanation—the property of having some physical property F such that ...—seems intuitively not to be playing anything worth calling a causal-explanatory role. According to the functionalist, the same goes for the functional properties expressed or referred to by our propositional-attitude predicates in propositional-attitude explanations. The problem with functionalism about propositional attitudes, it’s now generally appreciated, is that there are evidently decisive reasons for denying that propositional-attitude properties reduce to functional properties.[29]

  6. Several of these philosophers—e.g., Fred Dretske, Colin McGinn, and Ruth Millikan—hold that content plays some sort of teleological role, although they don’t all opt for quite the same teleological role. Fred Dretske, for example, has claimed that propositional-attitude properties explain not why propositional attitudes cause the actions they cause, but rather why we have those internal causes in the first place. This seems wrong; among several other things, it’s hard to see how anything like this can be going on when we explain that Johnny wrote ‘4’on his exam paper because he believed that 12 + 12 = 4, or that Ralph was looking for a virgin so that he could catch a unicorn.

    More recently, Frank Jackson and Philip Petit have argued that propositional-attitude explanations are what they call "program explanations."[30] For an exposition of their proposal, as well as plausible objections to it, see Kim’s Mind in a Physical World, pp. 72-77.

  7. Withdrawal to an undercover explanation. "Why," you ask, "did Lester get so violently ill?" "Because," someone explains, "he ate the stuff in Mary’s lunch box." In context, this may be an adequate explanation, but should we want to say that the property of having eaten the stuff in Mary’s lunch box is playing in this explanation a causal-explanatory role? In this context, we don’t want to say that the property of being the stuff in Mary’s lunch box was gratuitous, had no pointed related to explanatory concerns, but we intuitively want to say that the real causal property alluded to was the property of having eaten rat poison (well, we might want to say that until we realize that we can make the same point about the property of being rat poison viz-à-viz the property of having such-and-such chemical composition). In any event, it might be, and has been, suggested that propositional-attitude properties are playing the same sort of role, alluding, so to say, to properties that might more legitimately be said to be playing a causal-explanatory role. In this way, so-called propositional-attitude explanations would turn out covertly to be non-intentional explanations of this kind. The main problem confronting this line is to make sure that what one claims is the undercover explanation doesn’t force one also to conclude, by parity of reasoning, that the only explanations are those forthcoming from a correct version of fundamental physics. Since Hartry will be advancing an undercover line when it’s his turn at bat, I’ll defer further discussion until then.

III. My 2¢[31]

So far I’ve been using expressions like ‘cause’, ‘causal explanation’, and ‘causal-explanatory role’ as though in each case it determinately applied or determinately failed to apply, it being our job to discover which determinate fact obtained. Ha, ha. The truth of course is that these expressions are even vaguer than most vague expressions (virtually all expressions are vague to some degree). We want to do our best to make sure that a dispute on these matters isn’t about where to draw an arbitrary line in some expression’s penumbra. I’ll try not to say anything that will degenerate into a squabble with no determinate resolution about which views get to own the honorific labels.

An explanation is what is typically conveyed by a ‘because’ statement, as when we explain that Ava stepped back to the curb because she believed that stepping back was the best way to avoid getting run over and preferred not to get run over. This is a paradigmatic propositional-attitude explanation, and in this typical use of propositional-attitude properties lies their explanatory role. The explanatory role of propositional-attitude properties just is their ability to occur in true and contextually apt propositional-attitude ‘because’ statements, and I operate on the defeasible assumption that some of those statements are both true and contextually apt. I’ll try to say as much as can be said up to indeterminacy about what are and what are not the truth and aptness conditions for propositional-attitude ‘because’ statements, and thereby about the explanatory role of propositional attitude properties. If enough can be said, it should throw considerable light on the point of our having propositional-attitude concepts. I won’t now attempt to motivate my ex cathedra pronouncements any more than they’ve already been motivated. I begin with some negative claims about utterances of the form

[*] x Fed because x PAed that ...

N1. The truth of a [*] utterance requires neither strict nor ceteris paribus propositional-attitude causal laws of any relevance to our discussion. This isn’t to deny that a ceteris paribus generalization might express a truth; it’s only to deny that such a truth would be anything that could reasonably be called a causal law (or indeed any other kind of law). It’s also the case that paradigmatic special-science causal explanations don’t require strict or ceteris paribus laws. And while we can speak of completions of the partial explanations we give in uttering propositional-attitude ‘because’ statements, no such completion contains anything that can even be called a serious theory, forget about laws.

N2. Propositional-attitude properties typically doesn’t supervene on what’s in the head, and thus can’t be identified with any relevant "causal powers" that are in the head. In this connection we should note that it’s plausible that many special-science explanatory notions are analogously "wide." For example, the property of being a gene doesn’t supervene on the mircrostructural properties of DNA sequences, for, as Ron McClamrock points out, in order to be a gene something must contribute to the phenotype in ways that will typically depend on relations to other genetic materials and to the nature of the coding mechanisms that act on the DNA sequences. "So the properties of a particular gene considered as that kind of DNA sequence are supervenient on its local microstructure, but its properties considered as that kind of gene are not."[32]

N3. As I’ve already suggested, it’s at best indeterminate whether propositional-attitude state tokens are identical to physical state tokens, and the reason this is so also shows that it’s at best indeterminate whether propositional attitudes are identical to topic-neutral state tokens.

N4. Our use of ‘cause’ is pretty relaxed. Pace Davidson, causes are not just event or state tokens. They can also be conditions ("What caused the accident was the icy condition of the road") or facts ("The fact that the vase was cracked caused it to break when he touched it"). Evidently, something may in a suitable context correctly be called a cause merely if the effect would not have occurred had it not obtained and it can be cited in a correct explanation of the fact that the effect occurred.

On the positive side I would claim:

P1. [*] is true only if, absent the sort of overdetermination that rarely occurs, x would not have Fed if x hadn’t PAed that ....

P2. The fact that, e.g., Ava wouldn’t have stepped back had she not believed that a car was coming is due to the fact that her believing that a car was coming was in some sense non-causally necessitated by—in some sense "supervened" on—a non-intentional fact that included causally crucial individualistic facts. What makes it true that Ava wouldn’t have stepped back had she not believed that a car was coming is that neither the causally crucial individualistic fact (the fact that a certain neural state had such-and-such neural property) nor any other such fact would have obtained if Ava hadn’t believed that a car was coming. What’s not so clear is, first, what sort of supervenience is needed here (especially in view of the fact that some beliefs involve being in certain experiential states), and, second, what explains such supervenience between such disparate properties.[33]

P3. Let’s call the necessary condition stated in P1 the counterfactual core of a propositional-attitude explanation (and let’s ignore that fact that not every propositional-attitude explanation is of form [*]). The counterfactual core isn’t a sufficient condition for the truth of a [*] utterance. Suppose, to recall one of the examples with which I began, Henry’s face suddenly became read because he realized that Jerry Fodor overheard him say that Britney Spears was a better singer than any soprano the MET could produce. We wouldn’t suppose this to be explained by the statement that

Henry’s face suddenly became red because he believed that he wasn’t the only person alive

even though that satisfies the counterfactual core. What more do we need? To a rough first approximation, we can say:

In light of the foregoing, let’s now return to the vague questions put on hold: Are propositional attitudes causes of bodily movements? If so, then do propositional-attitude properties enter essentially into the correct causal explanations afforded by propositional-attitude ‘because’ statements?

There can be no serious doubt that as we use ‘cause’ in ordinary language, the propositional-attitude facts expressed in correct propositional-attitude explanations can be said to be causes and to determine propositional attitudes that can be said to be causes. To be sure, this is consistent with wanting a more refined notion of cause for this, that or the other purpose.

OK, so far as ordinary language goes, propositional-attitude explanations are causal explanations, no matter what they do or don’t manage to supervene on. (This isn’t to deny that we wouldn’t know what to say if we couldn’t explain the truth of the propositional-attitude counterfactuals.) Can we also say that propositional-attitude properties are essential to the causal explanations they’re used to give? That wouldn’t follow from what we have so far. We can see this with a non-propositional-attitude explanation. Suppose Sally is walking down the street and suddenly lands on her keester. One person explains Sally’s fall by saying that she stepped on a banana peel, and someone else explains it by saying that she stepped on a slippery surface. It’s by no means clear that the first explanation is a more complete explanation of the fall than the second, notwithstanding that it satisfies the counterfactual core and the aptness conditions specified above. Nor does it seem right to say that the two statements give different but equally correct explanations. But if both statements give the same explanation, then, by virtue of the asymmetry between the property of being a banana peel and the property of being slippery, we should conclude that, though there are good pragmatic reasons for citing the former property, it’s not an essential part of the explanation the ‘because’ statement offers. Things get interesting, however, when we next ask whether the property of being slippery is essential to its explanation. For suppose we now have a third explanation, offered by one physicist to another, which says that Sally fell because she stepped on a surface having blah-blah microphysical structure. There is an asymmetry between the property of being slippery and the property of having blah-blah microphysical surface structure, but if we say that slipperiness isn’t essential to the explanation it’s used to give, then we’ll for sure end up saying that the only explanatorily essential properties are those expressible in quantum physics. And that simply isn’t how common sense individuates commonsense explanations.[34]

Is there a difference between the banana peel/slipperiness asymmetry and the slipperiness/blah-blah microphysical surface structure asymmetry that would account for our intuition that while being a banana peel isn’t essential to the explanation its used to give, there’s a sense in which slipperiness is? I think so. The first asymmetry is that stepping on a banana peel is liable to cause a fall only because banana peels tend to be slippery, but stepping on something slippery isn’t liable to cause a fall because some slippery things are banana peels. The second asymmetry is quite different: stepping on something slippery is liable to cause a fall regardless of the surface’s microphysical structure. The asymmetry here is a matter of supervenience: being slippery supervenes on having blah-blah microphysical structure (or on a property that includes that microphysical property), but not vice versa. Being a banana peel does not supervene on being slippery (nor, of course, vice versa).

When we turn back to a typical propositional-attitude property mentioned in a true ‘because’ statement (e.g., Ava’s belief that a car is coming), we see that there’s no other property in the offing that stands to it in the way being slippery stands to being a banana peel, although, of course, lots of properties stand to it the way having a certain microphysical surface structure stands to being slippery. In this respect there’s nothing any more special or explanatorily relevant about an underlying computational-cum-physical explanation of behavior, even if it can be called ‘psychological’, than there is about an underlying purely physical explanation of the same behavior.

I draw two conclusions. First, that what matters most is that we can see what it is about our propositional-attitude concepts that accounts both for our having them and for our using them in ‘because’ statements. Second, that as far as common sense and ordinary language and our vague notions of causation and causal explanation go, propositional attitudes are causes of the actions for which they’re also the reasons why the actor performed those actions, and propositional-attitude properties are essential to those explanations. If you disagree with me, please be sure your disagreement isn’t verbal.


[1] Versions of some of these objections may also be used to show that propositional-attitude explanations don't even enter essentially into correct causal explanations of propositional-attitude facts; but simplicity of exposition recommends ignoring this for now.

[2] [cite]

[3] For specific references, see Donald Davidson, "Actions, Reasons, and Causes," reprinted in his Essays on Actions & Events (OUP, 1980). Davidson's famous paper, which was first published in 1963, was widely thought to have refuted the neo-Wittgensteinian line and to have shown that reasons can be causes, though it raised problems of its own about the causal-explanatory role of propositional-attitude properties.

[4] Jaegwon Kim, Mind in a Physical World: An Essay on the Mind-Body Problem and Mental Causation (MIT Press, 1998); p. 111. For Lewis, see, e.g., "Mad Pain and Martian Pain," ....

[5] Being a relation to a proposition doesn't per se make a relation intentional, any more than being a relation to person makes a relation intentional just by virtue of the fact that thinking about that person involves an intentional relation. There is no more pressure on a physicalist to reduce propositions to physical entities than there is to reduce numbers to physical entities.

[6] A Theory of Content and Other Essays (MIT Press, 1990).

[7] See ch. 4, "Intentionality and the Language of Thought," Remnants of Meaning, and the essays by Lynne Rudder Baker, Paul Boghossian, and Brian Loar in B. Loewer and G. Rey, eds., Meaning in Mind: Essays on the Work of Jerry Fodor, (Blackwell, 1991).

[8] Neurophilosophy: Toward a Unified Science of the Mind/Brain (MIT Press, 1986); p. 294.

[9] Philosophy of Mind (Westview Press, 1996); p. 215.

[10] MIT Press, 1987.

[11] Philosophy of Mind, p. 150.

[12] Mind in a Physical World, p. 36.

[13] But I didn't end up concluding that propositional-attitude properties strongly reduced to properties that were intrinsically specifiable in non-intentional terms. Instead, I ended up concluding that properties-not just propositional-attitude properties but also physical and all other properties-didn't really exist. They existed only as a manner of speaking, a façon de parler, as "pleonastic" properties. My present construal of 'pleonastic' isn't nominalistic, although it does offer a deflationary account of the existence of abstract entities such as properties and propositions. See my "Meanings," in J. Campbell, M. O'Rourke & D. Shier (eds.), Essays on Meaning & Truth (Seven Bridges Press, 2001).

[14] Remnants, pp. 153-4.

[15] Philosophical Perspectives, 3 (1990): 153-185.

[16] …; p. 239.

[17] Psychosemantics: The Problem of Meaning in the Philosophy of Mind (MIT Press,1987).

[18] "Does Meaning Matter?" ...; pp. 5-7.

[19] Brian Loar, "Social Content and Psychological Content," in …; p. 572.

[20] "…," Journal of Philosophy ... (1988):000-000; p. 385.

[21] Mind, Vol. C, 1 (January 1991): 1-17.

[22] Op. cit.: 19-34.

[23] [Cite]

[24] [Cites]

[25] I argue for this most recently in "Meanings," in J. Campbell, M. O'Rourke & D. Shier (eds.), Essays on Meaning & Truth (Seven Bridges Press, 2001).

[26] Reprinted in N. Block, ed., Readings in Philosophy of Psychology, vol. 2 (1981).

[27] Ibid., p. 000.

[28] This probably isn't true of Loar's argument for narrow content, which is primarily based on the kinds of examples considered earlier in the text and not so much concerned with wide content's failure to supervene on what's in the head.

[29] For some of these reasons, see Ned Block, "Troubles with Functionalism," Philosophical Review ..., and ch. 2 of my Remnants of Meaning.

[30] "Program Explanation: A General Perspective," Analysis 50 (1990): 107-117.

[31] Roughly the same 2¢ appears with equal superficiality in my "Ceteris Paribus Laws."

[32] Ron McClamrock, Existential Cognition: Computational Minds in the World (…).

[33] Paul Boghossian puts the point well: "in the absence of further comment, a relation of supervenience between sets of distinct and highly disparate properties is puzzling. How could there be a set of necessary connections between such properties as being a certain configuration of molecules and believing that Lully was a better composer than Purcell, given the admittedly highly divergent characters of the properties involved? We are entitled to be mystified" ("Naturalizing Content," in B. Loewer and G. Rey, eds., op. cit.; p. 66). My book-in-progress, The Things We Mean (to be published by Oxford University Press), offers a demystification.

[34] But mightn't the physicist say that he's not giving a different explanation of the fall than the one that cites slipperiness? Yes, she might say that, but then she's best interpreted as saying she's not offering a different competing explanation.