Can mental content explain behavior?
Some representations are mental; others are not. A road-sign, a map and the utterance of a sentence of a natural language are non-mental representations. All representations, whether mental or non-mental, have contents. Some have a conceptual content; others have a non-conceptual content. For example, motor representations, sensory experiences, perceptual representations and mental images have non-conceptual contents. To restrict myself to mental sensory experiences in the visual modality, visual experiences and visual mental images have pictorial non-conceptual content. Among an individual's mental representations, thoughts and propositional attitudes have conceptual contents. And so do an individual's utterances which, as I just said, are non-mental representations.
No doubt, there are many important differences between conceptual and non-conceptual content. In this paper, however, I will disregard the many significant differences between the varieties of mental contents and I will ask a general question about the explanatory role of content: can mental content explain an individual's intentional behavior? Is the content of an individual's mental representation one of the causally efficacious properties of the individual's representation in the process whereby it contributes to the production of the individual's intentional behavior?
Not everything an individual does is produced by the content of some of the individual's mental representations. Only an individual's intentional behavior is. The fact that I snore and vomit - if and when I do - is not something which can be explained by the contents of some of my mental representations. Nor is it intentional behavior. Suppose, however, that I move my right hand to the vicinity of a glass containing water and located a few inches from my chest. Suppose I grasp it, move it to my lips and drink water from it. What I did when I grasped the glass and drank water from it is intentional behavior. And it can be explained by the fact that I believed that the glass in front of me contained water and I wanted a sip of water. Furthermore, I will assume that my belief that the glass in front of me contained water is some state of my nervous system. So, I assume that my brain state has several non-semantic neurophysiological properties. But it also has content or, as I shall say, it has a semantic property. Now, the question I want to raise is the following: can the semantic property of my belief - what I believe - be a causally efficacious property of my brain state? If many of the neurophysiological properties of my brain state are causally efficacious in the process whereby I grasp the glass with the fingers of my right hand, can the semantic property of my brain state be causally efficacious too?
1. The argument against the causal efficacy of content
Many philosophers, whom I shall label intentional irrealists, would respond negatively to this question. On their view, content cannot be a causally efficacious property of an individual's mental representation. I think the argument for their view can be put in the following way:
(1) A decent causal explanation of why some cause c produced some effet e must refer to some causally efficacious property of c in the process whereby c caused e.
(2) Only c's intrinsic properties (or properties which supervene on c's intrinsic properties) are causally efficacious in the production of e.
(3) A representation's content (or meaning) is one of its extrinsic historical properties.
(4) Therefore: the content of an individual's mental representation R is not causally efficacious in the process whereby R contributes to the production of the individual's (intentional) behavior.
I will assume that in the non-psychological sciences, it is widely assumed that causal relations hold between events, as when one says that the short-circuit caused the fire or the collision between a meteorite and the Earth caused the extinction of dinosaurs. c caused e. In the psychological domain, we may say that my belief c caused my bodily motion m. A cause, however, has many different properties, not all of which may equally be causally efficacious. Causal efficacy is not fairly distributed among the properties of a cause. In order to show this last point, I'll borrow a famous example from Fred Dretske (1988).
2. The meaningful sound
Let c be some upper sound produced by a soprano on a particular occasion, e.g., Tebaldi singing a particular air of Traviata at a particular time t. Let it also be the third note of the scale. Suppose c broke a window at t + 1. Call e the breaking of the window. c caused e. Suppose c had a meaning. Call it F. Did c cause e in virtue of being F? Was c's being F causally efficacious in the process whereby c caused e? No, it wasn't. Why not? Because if c had had another meaning or if c had had no meaning at all, it might still have produced e; it might still have broken the window. What was causally efficacious in the process of window breaking was c's having an acoustic property, K, not c's being F. If c had not been K, it might not have broken the window. The reason c's being K was causally efficacious in breaking the window is that the causal relation between c and e can be subsumed under a nomic correlation between the fragility of the window-pane and c's being K. But there is no such correlation between the fragility of the glass and c's being F.
This example shows, I think, two things. First of all, it shows that not all the properties of cause c can be causally efficacious in the process whereby c produces e - something which is presupposed or implied by the first premiss. Furthermore, unlike the acoustic property K of the sound, its meaning F is not one of its intrinsic properties. The acoustic property K is an intrinsic property of the sound in the sense that the sound would not be the sound it is if it lacked property K. Arguably, event c which consists of the producing of a given sound would not be the same event if the sound lacked its meaning. But there is, I think, a sense in which the sound - unlike the event of producing the sound - would be same sound if it had the same acoustic property K and if it lacked semantic property F. But the sound would not be the same sound if it had semantic property F and another acoustic property K*. Unlike the sound's acoustic property K, its meaning F derives from Giuseppe Verdi's intentions. F is therefore a historical extrinsic property of the sound. Granted, the sound itself was chosen by Verdi. But its being K, unlike its being F, did not depend on Verdi's intentions. Verdi gave the sound its meaning; he did not make up the musical scale.
On my view, there is a difference between the meaning of a sound produced by a singer and the mental content of a human being's brain state, e.g., Verdi's musical intention. The latter is primitive (or original); the former is derivative upon the latter. Generally speaking, the meanings of non-mental representations are derivative on the meanings of mental representations. In spite of this difference though, mental content too is a historical extrinsic property, not an intrinsic property, of an individual's brain state. It depends on correlations between an individual's brain on the one hand and properties instantiated in the individual's environment. This is the substance of the third premiss. To further establish the plausibility of premisses (2) and (3), I'll now borrow an example from Dan Dennett (1987). But I'll exploit it in a Dretskean non-Dennettian way.
3. The vending machine
Let S be a French vending machine emitting soft drinks upon receiving French coins worth 5 French francs. Call m S's output at t, namely the delivery of a drink via the opening of a valve in S at t. Call c a particular 5 francs coin inserted in S's slot at t - 1. The insertion of coin c caused m, the delivery of the drink. Now, c has intrinsic physical properties P: c is a metal disk with a characteristic mass, a characteristic physical and chemical composition, a characteristic diameter, and so on. And c has a monetary value: it's worth 5 French francs. Call c's monetary value V. What conferred V onto c is the fact that c stood in some appropriate historical relation to a person working for the French Treasury Department who imprinted onto it the requisite "autograph". To see how extrinsic the relation which conferred V onto C is, let c* be another metal disk which is physically and chemically indistinguishable from c because it instantiates P too. If c*, however, did not stand in the required historical relation to the hands of the right people working for the French Treasury Department, then in spite of the fact that c* is P, still c* would lack V. To say that V is a historical extrinsic property of c is to say that c's being V does not supervene on c's being P.
No doubt, the insertion of c in S's slot at t - 1 produced m, the delivery of a drink, at t. Had some other physical object lacking P been inserted in S's slot, it would not have produced m. And no doubt c is V. But the question I want to ask is: was V causally efficacious in the process whereby the insertion of c caused m? Was the fact that c was V causally responsible for the opening of a valve in S which in turn freed a drink? I want to say: No, the fact that c was V was not causally efficacious in the process of valve opening which in turn delivered a drink. The vending machine does not really detect monetary value V. What it is sensitive to are not instantiations of V, but instantiations of P. The opening of the valve in S is nomically correlated with property P of c, not with property V. As philosophers say, property V is "epiphenomenal" in the mechanism by which the insertion of c produces m.
So, if c's being V - if the monetary value of a metal disk - is a relevant model of how an individual's brain state instantiates mental content, then what I believe when I believe that the glass in front of me contains water does not causally explain at all what I do when I move my right hand to the vicinity of the glass, grasp the glass with the fingers of my right hand, move it to my lips and drink water from it. This in a nutshell is conclusion (4), i.e., the conclusion drawn by intentional irrealists from premisses (1)-(3). Again, I accept the premisses. Am I then not forced to accept the conclusion? The rest of this paper is an attempt to show why not.
4. Why the conclusion does not follow from the premisses
The vending machine is a physical device which responds to inputs of a specified kind by producing a characteristic output: it emits a drink upon receiving metal disks with intrinsic physical property P and extrinsic property V. The question we asked was: is c's being V causally efficacious in the mechanism whereby a valve opens in S and a drink is emitted? Does the fact that c is V explain m, S's delivery of a drink? Let's say that this is a question about the mechanism of valve opening in S. We were, I claim, forced to admit that only property P of c is causally efficacious in the mechanism of valve opening in S. Property V of c is not.
Notice, however, that strictly speaking, the question about the mechanism is not a question about S's behavior. It is a question about the production of physical motion m. How can this be? Isn't S's behavior simply m, i.e., the delivery of a drink? I want to suggest that no, physical motion m - the delivery of a drink - is not quite the same thing as S's behavior. S's behavior is a process one constituent of which is m - the delivery of a drink - and another constituent of which is the cause of m, namely the insertion of coin c. Indeed, on what I shall call the componential view of behavior (championed by Dretske 1988 and which I accept), it is a conceptual mistake to confuse the output of behavior with behavior itself. As much so as to confuse a constituent of something (e.g., an event) with what it is a constituent of (a process or a sequence of events). S's behavior is the process whereby the insertion of coin c in S's slot at t - 1 produces m, the delivery of a drink at t. The very fact that m cannot happen at the very instant at which c is being inserted shows that there are two distinct events: the insertion of c in S's slot and then the production of m. Thus S's behavior is a sequence of events which are related to one another as a cause to its effect. Hence, S's behavior could be symbolized thus:
[c ---> m].
From now on, I will assume that the componential conception of behavior is on the right track and that S's behavior is distinct from its output, m, the delivery of a drink. If so, I can now say why I think conclusion (4) does not follow from premisses (1)-(3). What reflection on the vending machine example shows is that extrinsic property V of c is not causally efficacious in producing physical motion m in S. The economic value of coin c is not causally efficacious in the emission of a drink via the mechanism of valve opening in S. Nothing has been said yet about explaining S's behavior. By parity of reasoning, what follows from premisses (1)-(3) is not that the content of an individual's representation R is not causally efficacious in explaining the individual's behavior, which could be symbolized thus [R ---> m]. Rather, what follows from premisses (1)-(3) is that the content of an individual's representation R is not causally efficacious in the production of some of the individual's physical motion, m (where again m is a constituent of the individual's behavior).
Since I assume that behavior is not just physical motion, I will now argue that there is a sense in which explaining S's behavior is not the same thing as explaining the production of S's physical motion m. What I earlier called the question about the mechanism was the question: what property of inserted coin c is causally efficacious in the mechanism of valve opening which yields a drink? Consider now a different question: why is it that every time a metal disk c with property P is inserted in S's slot, S does m, i.e., S emits a drink? Why do instances of the former regularly produce instances of the latter? This is a question about the structure of S's behavior in the following obvious sense: what we want to know now is why S's behavior is a process of a certain type. Why does S's behavior consist of the insertion of metal disks of type c followed by events of type m, emissions of a drink? Why does S's behavior consist in the production of m by c? Why is S's behavior constituted by the coordination between c and m rather than between c and something else?
Notice that, unlike the question about the mechanism, the question about the structure of S's behavior is a historical question. The question is: how did it come about that insertions of coins of type c regularly produce events of type m? And the expected answer to this historical question about the structure of S's behavior is going to refer to a selective process. Furthermore, I want to suggest that perhaps c's having extrinsic property V does explain after all the structure of S's behavior: perhaps, c's being V explains the coordination between c and m.
Let us suppose first that, thanks to the efforts of the French Treasury Department, instantiations of property V are reliably correlated with instantiations of property P. The general answer to the historical question about the structure of S's behavior will go like this: c has been coordinated with m because metal disks of type c with property P were chosen (or selected) by a human engineer as causes of m, the opening of a valve in S emitting drinks. The correlation is not perfect: occasionally, the insertion of a counterfeit - a coin which is not really worth 5 French francs - gets the machine to deliver a drink. But it is good enough. Presumably, the cost of designing a more sensitive vending machine is not worth the effort. Notice that a counterfeit must have intrinsic property P, even though it does not instantiate extrinsic property V.
Now, suppose that for some reason at time t, the Head of the French Treasury Department decides to change the intrinsic properties of coins worth 5 French francs. After t then, there is no reliable correlation between instantiations of intrinsic property P and instantiations of extrinsic property V: metal disks with property P do no more instantiate property V. After t, property V is to be instantiated by metal disks with intrinsic property P*. If this were so, then insertion of metal disk c having property P in S's slot would not cause m any more. After t - assuming that the price of drinks remains fixed over the change of decision -, French engineers must design new vending machines such that insertion of metal disks c* with intrinsic property P* (not P) in their slots will cause m, the delivery of a drink.
If so, then it's fair to say that some metal disks with property P* have been selected in the process of production of m because they instantiate extrinsic property V. Likewise, before the decision to change the intrinsic properties of coins, metal disks c having property P were selected as causes of m because they then instantiated extrinsic property V. Had they not instantiated property V, they would not have been selected as causes of m in the first place. So I do want to claim that extrinsic property V does after all explain the structure of S's behavior. It does explain why coins with intrinsic property P were selected as causes of m.
To sum up, the vending machine shows three things. First, it shows that the delivery of a drink, m, is not the same thing as S's behavior. Second, it shows that we can distinguish two kinds of causal explanations: the ahistorical explanation of the mechanism whereby c causes m and the historical explanation of why c was recruited as a cause of m. Third, it shows that c's being P is involved in the mechanism whereby c causes m and c's being V explains why c was recruited as a cause of m. Of course, coins with property V were selected as causes of m by human engineers with propositional attitudes (intentions, beliefs and desires). So the selection process was an intentional process. I have simplemindedly (I confess) assumed that the causal efficacy of c's being V in the explanation of the structure of S's behavior is a relevant model of the causal efficacy of mental content in the explanation of an individual's intentional behavior. It is incumbent upon anybody who makes such an assumption to show two things. The first is to argue that the content of an individual's mental representation R can indeed explain the recruitment of R as a cause of some physical motion m of some of the individual's bodily parts. The second is to argue that the selection process whereby R was selected as a cause of m is a non-intentional process.
I will merely ask you to imagine a prey which I'll call D (for deer) executing some flight motion m at t. I grant that the prey's behavior is not voluntary behavior because it is instinctive. However, I will assume that it is intentional because the flight motion is caused by some internal representation. I will now argue that we can at least ask three different questions about D's flight motion m.
First, we can ask: why does D produce m at t? Answer: because at t - 1, D detected predator P in the vicinity. Call c D's internal state whereby D dectected its predator. The presence of predator P prompted D to enter c at t - 1 which in turn caused m at t.
Secondly, we can ask the ahistorical question: what is the internal mechanism whereby c in D causes m? General answer: motor neurons in D's motor cortex are wired to fire upon receiving inputs from sensory neurons in D's visual (or olfactory) system. And the former in turn govern muscle contraction in D's limbs. Presumably, the fact that c represents or reliably indicates the presence of predator P is not part of the answer to the question of the mechanism whereby why c causes m.
Finally, consider the historical question: how did it come about that every time D is in state c, D does m? Why does it do m rather than something else? Why was c recruited as a cause of m, flight movement, rather than fight movement m' or no movement at all? The answer, I think, to this last question, would come in two steps. First, c was recruited as a cause of m (flight movement) in D by natural selection. I assume that natural selection is a non-intentional selection process. In the competition for survival between ancestors of preys of type D and other animals which were preys of ancestors of predator P, natural selection favored the former because they had a coordination between c and m. Preys of ancestors of P which lacked a coordination between c and m did not survive and went extinct. Secondly, natural selection would not have recruited c as a cause of m in ancestors of D if c had not been correlated with the presence of predator P. In other words, I claim, c would not have been turned into a cause of m by natural selection if it had not been a reliable indicator of the presence of a predator.
Like property V of coins, the content of a prey's perceptual state c - its being a reliable indicator of predators - is a historical extrinsic property . It too, I suggest, plays a causal role in the historical explanation of the structure of the prey's behavior: it contributes to explaining why c was coordinated with flight motions m in the presence of predator P. Unlike the monetary value of coins, however, the content of the prey's brain state turns on natural selection, which is a non-intentional selection process.
Churchland, Paul M. (1984) Matter and Consciousness, Cambridge, Mass., MIT Press.
Davidson, Donald (1980) Essays on Actions and Events, Oxford: Clarendon Press.
Dennett, Daniel C. (1987) "Evolution, Error, and Intentionality." In Dennett, Daniel C. (1987) The Intentional Stance, Cambridge, Mass., MIT Press.
Dretske, Fred (1988) Explaining Behavior, Cambridge, Mass., MIT Press.
Fodor, Jerry A. (1987) Psychosemantics, Cambridge, Mass.: MIT Press.
Jacob, Pierre (1997) What minds can do, Cambridge: Cambridge University Press.
Putnam, H. (1974) "The Meaning of 'Meaning'", in Putnam, H. (1975) Philosophical Paper, II, Cambridge: Cambridge University Press.
Quine, Willard Van Orman (1960) Word and Object, Cambridge, Mass.: MIT Press.
Stich, Steve (1983) From Folk Psychology to Cognitive Science, Cambridge, Mass.: MIT Press.
Unlike many philosophers, I think content can explain some of the things an individual does. It is an open question whether Davidson's (1980) anomalous monism allows content to explain intentional behavior. Eliminative materialists such as Quine (1960), Churchland (1984) and Stich (1983) explicitly deny it. So does Dennett (1987) according to whom there cannot be semantic engines. And so does Fodor (1987) who thinks that syntactic, not semantic, properties of mental symbols can be causally efficacious.
This argument makes use of the contrast between a thing x's intrinsic and extrinsic properties. Although the contrast is notoriously hard to capture, it is, I believe, useful. Intuitively, a property is intrinsic just in case its being instantiated by x at time t does not depend on anything other than x at t. It is extrinsic if it is not intrinsic.
 The extrinsicness of meaning has been emphasized notably by Putnam (1974).
 See Jacob (1997, ch. 8).