Brendan Lalor. Philosophical Psychology 10, 165-178, 1997.
ABSTRACT. In this paper I argue against the worried view that intentional properties might be epiphenomenal. In naturalizing intentionality we ought to reject both the idea that causal powers of intentional states must supervene on local microstructures, and the idea that local supervenience justifies worries about intentional epiphenomenality since our states could counterfactually lack their intentional properties and yet have the same effects. I contend that what’s wrong with even the good guys (e.g. Dennett, Dretske, Allen) is that they implicitly grant that causal powers supervene locally. Finally, I argue that once we see the truth of an anti-individualism which sees cognition as a fundamentally embedded activity, it becomes clear both that granting local supervenience is granting too much, and that intentional properties do work mere neurological properties could never do. I also suggest how a transcendental argument for intentional potency might go.
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 or meaning, will (indeed must) turn out to be explanatorily irrelevant. Even if some events have a meaning, and even if, as physical events in good standing, 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. (Dretske 1990: 5)
Of course the thrust to naturalize intentionality is good. What I want to address here are two lingering a priori worries about intentional causation which I think may be holding back the naturalization project in some quarters. They loom large and cut deep — and I’ll indicate how I think they’re related. The first is the idea that since action at a distance is an anathema, it must be that causal powers supervene locally; a fortiori, for the intentional to be potent, it must honor local supervenience. The second is that requiring local supervenience of the intentional seems to foster worrisome intuitions about its possible epiphenomenality, since, after all — as one version of the worry goes — it then seems plausible to think our states could counterfactually lack their intentional properties and yet have the same effects. Fodor (1990) christened this nagging worry about intentional impotence epiphobia. The likes of Dennett (1983) and Dretske (1990) respect it. I aim to show here that both theses are mistaken, and that materialism itself needn’t commit us to defending either — although I’ll mostly be repudiating epiphobia.
I’ll start by trying to show how the local supervenience thesis might make the epiphenomenality of the intentional seem plausible. Then I’ll run through some intentionality- naturalizers’ attempts to refute the epiphobics. However, it’ll turn out that though their intentions are good, some of their arguments leave much to be desired, often because they at least implicitly grant individualism. Once we see the truth of an anti-individualist position which sees cognition as a fundamentally embedded activity, it becomes clear that the naturalizers have granted too much in granting local supervenience. And then the rain stops, and clouds part as the sun breaks through. Intentional properties do work mere neurological properties could never do. So I shall argue. (I’ll even suggest how a transcendental argument for intentional potency might go.)
Let me turn to the stuff of which epiphobic nightmares are made.
2. Local supervenience & epiphobia
Lots of materialist philosophers of mind hold that the causal powers of our intentional mental states must supervene on local microstructure. Colin McGinn (1991), for instance, argues that intentional properties must have their effects solely via agents’ motor apparati, which do not need ways of responding to what the concept is distally about (585). So lawlike generalizations of intentional psychology[ 1 ] will be over concepts whose powers supervene headwardly — i.e. on what’s in the head. He thinks
the aspect of content that lies ‘outside the head’, stemming from the environment, functions as a variable parameter with respect to which we can universally generalise without losing nomic thrust; not so, of course, the aspect that resides headward and recurs monotonously throughout … (all the Twin Earths). (580)
In other words, intentional psychology’s laws can quantify specific worldly reference away; i.e. they don’t generalize over earthly objects. Molecule-for-molecule twins have the same mental causal powers despite their contextual variation (581). Same brain states, same behavior. It’s worth registering as an assumption McGinn’s conviction that when we quantify parameters away, what we’re left with are conceptual powers which supervene on local neural structures. Actually, he begs crucial questions against the anti-individualist, as discussion below will suggest.[ 2 ] (His view here, no doubt, traces its spiritual lineage down to Fodor (1980), who complained that if the causal powers of the mental in virtue of which behavior is produced don’t supervene locally, then it’s hard to see how there could be mental causation of behavior at all. Until recently (Fodor 1994) Fodor played a lot on intuitions about such alleged unintelligibility of nonlocal causal powers.[ 3 ] But he still hasn’t fully recovered from methodological solipsism — see footnote 16.)
Given the local supervenience constraint on accounts of the mental causation of behavior, it’s not that hard to see how someone might begin to worry that since mental events’ neurological (/syntactic[ 4 ]) properties are spatially and temporally closest to the production of behavior, their intentional properties might be disqualified from being the causally potent ones to figure into accounts of the causation of behavior. One flavor of this concern goes like this: It seems that any intentional mental event m which causes another event n would have caused n even if it lacked its intentional properties, so long as m‘s neurological properties are intact. The epiphobic worry can thus be put in terms of this counterfactual wonder: What if there is a nomologically possible world in which the same neurological states control behavior just as they do in the actual world, but lack their intentional properties?[ 5 ]
Notice that this isn’t just the more aged worry that beings functionally isomorphic to us might lack some of our mental properties. It’s that maybe the intentional properties of our mental states are causally irrelevant to our behavior — since when we imagine what the actual production of our behavior must look like, we think of it as generated by, say, processes defined over representations whose nonrelational (hence, nonintentional) features are computationally salient. This is what really makes us go. This sort of consideration gives rise to another flavor of the concern: If that’s so, the question is, How could it be that human action is governed by intentional properties in any satisfying sense?[ 6 ] As Dretske (1990) puts it: ‘Though (my beliefs and desires) cause me to do A, their being beliefs and desires … is as … irrelevant to why I do A as is the fact that a brick was made in Hoboken to why it broke the window’ (6).
Nowadays, the distinction between causally salient and semantic or intentional properties is illustrated with some currency among philosophers by comparison to the distinction between genuine and counterfeit money.[ 7 ] My $20 bill is accepted because of its look (which is causally potent), not because of its genuineness — counterfeit bills by and large have the same effects as the rest. Likewise, my belief that P has its behavioral effects not because of what it’s about, but because of its power to bring about the tokening of other representations.
3. The cure
I want to show that this worry loses its force when we consider the powerful arguments that have been put forward by anti-individualistic philosophers such as Tyler Burge (1979) and Ron McClamrock (1991) to refute the thesis that the powers of intentional mental properties honor local supervenience. They have convincingly argued that the causal powers of mental states supervene in part on environmental and social mechanisms. I won’t repeat their arguments to scale here. But I’ll have to explain the view as set-up for points I want to make below.
Burge calls our attention to cases in which obliquely occurring terms in content clauses which give intentional characterizations of a subject depend on social conventions for their meanings. For instance, suppose we rightly characterize Thrifty Theo as wanting to buy briquets in bulk to burn in his grill. The causal powers of his mental state depend counterfactually on his community’s usage. He needs neither to be able to recognize briquets, nor to know that they are compressed coal dust, in order for his state to be about them, since he can rely both on his ability to find experts who do know such things, and on their knowledge. As he acts to satisfy his want, he turns to the yellow pages in which he finds the local briquet dealer listed, and the dealer sells him what the dealer certifies as briquets. Counterfactually, in a community in which things named by the word- form ‘briquets’ are compressed of a different substance, TT would be in a different intentional state, due to the difference in the substance to which he’s tied, and the environmental and social mechanisms on which the causal powers of his counterfactual intentional state depend.
Given that the causal powers of intentional states depend intimately on the world they are featured in, it should be no surprise if intentional behavioral systematicities emerge at the level of organism-object relations (see McClamrock 1991). After all, if TT’s having briquet-attitudes depends on his ability to exploit all kinds of social mechanisms (which serve as hooks, so to speak, by which he can ‘pull himself’ to the object of his wants), then it shouldn’t be surprising that the systematicity in his behavior emerges with respect to that thing in the world, briquet — rather than, say, with respect to his fine-grained bodily movements.[ 8 ]
In general, ‘sameness of causal powers is relative to a way of taxonomizing a system‘ (McClamrock 1991: 348). So if the taxonomy is that of, say, neuroscience, the causal powers in question are those of neural structures of brains qua neural structures. Presumably, these supervene locally ( — i.e. the laws of neuroscience are implemented by processes involving local structures). However, when it comes to the laws of scientific intentional psychology, you might often have to taxonomize over objects and properties in the world to get at the right behavioral generalizations. I want to draw attention to two implications of thusly taxonomizing. First, the variety of possible skills and social and environmental mechanisms which might implement a given type of behavior form a multiply realizable class over which intentional laws generalize. Second, being a certain type of intentional state is context- dependent — that is, while the characterization of some state a as neurological type N will hold across contexts, characterizing it as intentional state type I may presuppose certain contextual facts and external structures. So there will be many contexts in which Na is true while Ia is not.[ 9 ] When both Na and Ia are true it’s important to bear in mind that while a may be an intentional state, its satisfying N isn’t what makes it the intentional state it is. Neurological properties aren’t projectible in intentional laws.
All of this fits nicely with the breed of materialist ontology John Haugeland (1982) labeled ‘vapid materialism,’ which is just the doctrine that when you fix the universe’s lower-level, physical properties, all its higher-level properties get fixed thereby. (It’s far less encumbering than usual compositionality doctrines.) What’s important to note here is that it’s noncommittal about what happens when you fix the lower-level properties of certain structures in the universe — such as people and their brains. It allows that those structures might have all kinds of interesting, causally relevant relational properties, which make a difference to what higher-level type they are.[ 10 ] Vapid materialism thus sanctions properties which might be crucial to intentional psychology, such as the property wanting briquets (as I’ve explained its manner of supervenience). And on my account of the powers of intentional properties in particular, there’s no occult ‘causation at a distance’ to worry about: All the mechanisms involved in the implementation of behavior are somehow built of matter — the agent, the phone book, the briquet expert or dealer — and they act on one another in the usual ways. The physical mechanisms on which TT’s property wanting briquets supervenes are his brain and bodily skills and the features of his physical and social environments which he might (actually or counterfactually) exploit in tracking down briquets. So my view clearly honors the intuitive materialist principle: No causal powers without some mediating physical structures or other. It’s just that those structures don’t all reside headward.[ 11 ]
The union of philosophy of mind with the a priori prejudice in favor of local supervenience has produced two offspring which are regrettable from the current perspective. The first is the general hampering of progress in the project to naturalize intentionality. The second is the dysfunction under discussion here, epiphobia. Only such a prejudice can keep intentional psychology from helping itself to the sorts of relational properties I’m defending. Recognizing this is the first step on the road to recovery for epiphobics.
If you’re still unmoved, grant my points for the sake of argument, just to see what they might suggest about epiphobia.[ 12 ] Now let’s turn to some anti-epiphobic arguments, and evaluate them in light of morals drawn from the anti-individualist view outlined here.
4. Anti-epiphobes with good intentions
Numerous intentionality naturalizers have attempted to put epiphobia to rest. But most of them have something regrettable in common: They are satisfied with too little — which allows epiphobia to keep its foothold to an extent we needn’t tolerate. In particular, they are content to defend the potency of intentional properties with respect to our general behavioral patterns, and let them fall by the wayside with respect to specific token behavioral acts. Let me register that I’m in no way disparaging their insights, and I recognize most of them as genuine. I’m just claiming that certain of their implicit presuppositions have kept them from seeing how far the effort to defend intentional potency can and should be taken. And I intend to help remedy this condition.
But first I want to banish from this discussion a view of some popularity in recent history: The pre-’90s Fodorian pre-established harmony view[ 13 ] of the relation between syntax and semantics. It runs thus: Mother Nature has seen to it that lots of syntactic (/neurological) properties of brains correspond to their semantic properties, which are so dear to folk psychologists. While the view concedes that semantic properties are epiphenomenal with respect to explanation (since causation depends sheerly on syntax), it claims a kind of potency with respect to behavioral prediction. But as is, it’s a miserable position. To naturalize intentionality you have to find real ways of establishing harmony between the syntactic level and the semantic or intentional.[ 14 ] So let’s move on to something with more naturalistic potential.
Suppose event A, which has the meaning M and the intrinsic properties C, causes B. Just as the genuineness of a $20 bill supervenes on one set of facts (namely, its history and relations), and its actual effects another (namely, its ‘intrinsic’ look), so too with M and C respectively. Dretske’s (1990) suggestion is roughly that even if meaning doesn’t figure into the brute facts of individual cases of causation, it may help explain patterns of causation. So the bad news is that the fact that A means M can’t be what explains A‘s causing B: it’s screened off from explaining it because ‘in similar circumstances, an event lacking this meaning, but otherwise the same, will have exactly the same effects’ (9). Dretske’s gospel, however, is that meaning M nevertheless matters in another way: It helps explain ‘the fact that events of type A, when they occur, cause events of type B.’
A deep understanding of an organism’s systematic behavior is tied up with the answer to the question about why one sort of event in it causes another sort — how things got set up that way rather than otherwise.[ 15 ] What Dretske has in mind here are evolutionary and individual learning stories. Whatever doubts you have about the sufficiency of this sort of explanation aside, I take it you’ve heard how a process of selection (across generations, or within a life) might favor a reliable indicator state (which indicates, say, an entity-to-be-feared) hooked up so as to cause an appropriate behavioral response (fleeing) when it’s activated. There may be different answers to this question about why things are set up the way they are, even for molecularly identical organisms. In seeking the answers here, we derive a sense in which even identical patterns of bodily movements aren’t sufficient for same behavioral characterization. One woman may be fleeing a charging rhino, while another is trying to win the Olympic hundred-yard- dash. The point is that the selectional processes behind this sense of behavior are sensitive to semantic and intentional properties, not intrinsic formal ones (13). Verily, ‘to explain an animal’s behavior — not A, not B, but A‘s causing B — one can, indeed one may have to, advert to precisely those historical and relational facts about A on which its meaning supervenes’ (12). If this is the right way of thinking about behavior, then we can see how M might be relevant to behavioral explanation. Hence, no epiphobia.
Dretske’s rejection of the bodily-movement conception of behavior in favor of an intentionally- directed-process conception is surely a move in the right direction. He rightly recognizes the context dependence of bodily movements. But I want to make two related points: First, the discussion in section 2 of how an agent’s behavior might be thought to supervene on its skills for exploiting social and environmental features points up a way in which even Dretske’s conception of behavior is too limited. He doesn’t seem to realize something we spelled out there — that having a certain intentional property is a context-dependent feature of an organism which may presuppose that certain mechanisms outside the organism’s body are in place. I’ll let that go for now and come back to the idea below. Second point (which, I think, leads indirectly back to the first): Dretske seems to think individuation of contentful or intentional states is a matter separable in principle from their powers for causation. Remember, he says it’s intrinsic properties of the brain that implement behavior, even if relational factors individuate content. The anti-individualistic embedded account of intentional states from section 2 highlights the contestability of this assumption. There it was suggested that the elements that constitute a given intentional state’s being the kind of intentional state it is are mechanisms on which the intentional causal powers supervene.[ 16 ] It’s just that not all those mechanisms are intrinsic properties of the brain. If Dretske recognized this, the temptation to identify behavior with bodily movement would be far gone. The extreme multiple realizability of classes of behavior precludes such an identification.
It looks as though subtle individualistic assumptions have been the blinders keeping Dretske from seeing how intentionality might be saved for causation. Let’s look at another anti-epiphobe.
Dennett (1990) sees no reason to be down, since ‘even if meaning cannot be locally potent, it can be put into reliable pre-established harmony with locally potent structures’ via Dretskean channels (26). But, given what’s come before, we can see why we might take issue with his cheerful announcement that ‘meanings cannot directly cause things to happen — and hence they cannot directly cause themselves to correspond to any causal regularities in the world’ (21). He too overlooks the sense I’ve been highlighting in which meanings can have here- now potency: Construe contentful states as properties an agent has in virtue of the way her skills tie her to her world, and there emerges a way in which her meaning or intending what she does actually screens off the activity of her ‘locally potent structures’ from explaining her actions, rather than the other way around. See: Individualism is at work among even the finest of naturalists!
5. A transcendental argument for intentional potency?
On a bit of a different note, Colin Allen (1995) tries out what he calls a ‘transcendental defense’ of the causal potency of at least some intentional properties. ‘(E)ven if intentional properties of brain states fail to be causally relevant to their owner’s behavior … they may nonetheless be causally relevant to behavior in a different way’ (117). His argument goes: Although it might be true that (i) a brain state type may sometimes lack its intentional properties and yet still have the same effects as when it has those intentional properties, it does not follow that (ii) even if no brain states of that type had those intentional properties any of them could still have those same effects; (iii) hence a sufficient proportion of states having the intentional properties may be (and, he additionally argues, is) ‘transcendentally’ relevant to the others’ having those effects.[ 17 ] After all, although a counterfeit $20 bill will buy you just as much as a genuine one, if it weren’t for the cops keeping most counterfeiters in check, it’s arguable that nobody’s bills would do the work they do now. Hence, the genuineness of most $20 bills is an empirically established transcendental precondition for the potency of any.
While I am somewhat sympathetic to the spirit of Allen’s point, and to certain pieces of his argumentation, I have doubts about the cogency of his defense as applied to most intentional properties. If anti-individualism is right then his assumption that local properties of brain states might be sufficient for bringing about certain effects of a behavioral kind is tendentious. His allowance that ‘any particular brain state might cause the same behavior despite lacking the intentional properties in question’ (118 f.) allows too much for cases in which the lacked intentional properties are at all paradigmatic. This possibility depends on an impoverished notion of ‘behavior’. In the case of most intentional properties it is just false that their effects can also be brought about in virtue of physical-brain-state properties.[ 18 ] If the intentional properties Allen intends to transcendentally save for causation are just the ones whose behavioral effects are immediate, then perhaps it is true after all that salvation is only for the elect.
On the doctrine I’m preaching, the causal powers of an intentional property depend on all sorts of relational facts about agent and environment. For instance, the property intending to see the Pacific supervenes on the complex set of skills an agent has for getting to the Pacific, and on features of the environment which the agent might exploit to get there. Thanks to these context dependent features of the agent’s state, the intentional property has causal powers neurological properties could never have — such as being that in virtue of which the agent maneuvers through a maze of roads, pays tolls at booths, fills and re-fills the gas tank, and eventually sinks her toes into the sands of a Pacific beach.
A simple Twin Earth example will show this is not a counterfactual property of any neurologically taxonomized state. Suppose our Thrifty Theo lives in Missouri, and one day decides to get in his car and fulfill his dream to see the Pacific. That same day, neurologically type- identical Twin TT (who lives in the corresponding place called ‘Missouri’ on Twin Earth) likewise decides that he must see what Twin Earthers call ‘the Pacific’. In fact, the nonintentional characterization of his life is so far exactly the same as TT’s. But not for long: As it happens, the roads and towns on their routes are laid out somewhat differently — although, of course, as luck would have it, the differences haven’t tainted our protagonists’ neurologies at all yet. TT, because he’s skilled at map reading and car driving, successfully gets on Route 66, passes through Oklahoma, New Mexico, and southern California, on to the Pacific. Twin TT, though similarly skilled, has a different fate: His ‘Route 66’ meanders up into ‘Colorado’ and by-passes ‘New Mexico’ all together. Their intentional properties differ, since the one is specified by reference to that Earthly object, the Pacific, the other by reference to what Twin Earthers call ‘the Pacific’. Where twin roads diverge, you just can’t preserve the neurological type-identity of twins if you want to preserve their respective intentional properties.[ 19 ] So, here, only the intentional properties predict behavior; the same neurological properties (to start with) even eventuate in different behaviors. Thus, neurological properties (qua neurological) can’t have the same counterfactual macro-behavioral effects as intentional ones. So being a neurological property isn’t causally relevant for the explanation of (intentional) behavior at all. That’s why it doesn’t figure in any laws of intentional psychology.
As Fodor (e.g. 1990: 153) rightly spells it out, to explain a particular event e you have to subsume it under a law. But if e is subsumable under a law of intentional psychology, it is so in virtue of its characterization as a kind of intentional state. Since it’s not subsumed in virtue of its neurological properties, those properties are irrelevant to the explanation of behavior — although, of course, they’re not irrelevant to its implementation. (All this is especially clear if anti-individualism is true.)
Hence, contra Allen, something’s possession of a neurological property without the intentional property is insufficient for its subsumption — hence insufficient for its causal responsibility for that behavioral effect. A brain state qua brain state isn’t responsible for any behavioral effect qua behavioral effect. Based on the Dretskean motivations mentioned above, we could even develop a retort to the effect that same bodily movement doesn’t mean same behavior anyway. Another possible point against Allen: Since the implementation of prototypical intentional behavior is elaborate, it seems implausible to think one might manifest the elaborateness without having the intentionality. My intuition suggests that if something engages in activity of sufficient complexity — reacting appropriately to phone book listings, dialing the phone, asking the right people the right questions, and finally bringing home a bulk load of briquets, for instance — barring astounding evidence to the contrary, we have every reason to attribute intentionality and explain the activity by subsuming it under intentional laws. What could possibly motivate thinking otherwise?[ 20 ]
The above reflections suggest the possibility of a stronger transcendental argument than Allen’s: For a wide range of intentional properties (if not all), the preconditions on having them preclude their being impotent, since having such a property presupposes a state of affairs in which it is precisely what screens off fine-grained bodily movements, neurological properties, and the rest of the flurry of lower level activity, from explaining behavior. It’s not just that a sufficient proportion of behavior-causing states must have the right intentional properties for any of them to have their behavioral effects. It’s that each token behavior-causing state must have intentional properties for it to have behavioral effects.
6. Embedding epiphobism (six feet under)
Let me pull a few strands together at this point. The epiphobic worry was that maybe the aboutness of states isn’t that in virtue of which behavior is produced. However, as we’ve seen, anti-individualists needn’t sweat over this possibility, since they hold that intentional laws often generalize over objects (like the Pacific) and properties (such as briquethood) themselves — in which case (to overstate the point a little), it’s hard to see how the behavior could fail to be about, directed toward, the objects or properties in question. The embeddedness perspective needn’t limit the potency of intentionality to explaining coarse-grained processes of selection whose upshot is something like the harmonization of syntax and semantics across behavioral acts. It holds out a philosophically respectable, scientifically plausible way of understanding fine-grained directedness as potent within acts.
The defense of intentional potency from this perspective is so natural for two general reasons that have so far remained rather penumbral. First, as we’ve seen in a few cases, the view allows lower level detail to be screened-off from causal and explanatory relevance in favor of the higher, intentional level. Second, it permits agents to be sensitive to contingent features of the environment — to devise strategies for action in light of feedback from the world, so to speak. I’ll say a little about each of these facts.
Three factors help account for why the lower level is often screened off: The variety of sources of information in the environment, the variety of means to ends, and the robustness of agents’ skills to exploit these. This trio makes for the multiple realizability of successful behavior. Agents can by and large be counted on to — one way or another — skillfully exploit the environments in which they’re embedded so as to fulfill their intentions. There are many ways to fry an egg. To get to the Pacific, agents might rely on any number of sources of information: Maps, flight schedules, travel agents, directions from friendly pedestrians, or just the position of the sun. Even fixing the source of information — as, say, a map — agents might take one or another possible routes depending on expected utilities (or whatever). But the details of these possibilities can be abstracted away from since intentional laws apply to agents who are equipped to navigate their environments so that, ceteris paribus, their behavior correlates suitably with what their states are distally about. McClamrock (1994) puts it concisely: ‘In cases where there is no class of more proximal properties in virtue of which a particular class of effects takes place (even if each case has some particular cause at a lower level of organization), the distal cause may screen off the more proximal ones…’ (159).
Second, the embeddedness perspective permits an intentional system to be sensitive to contingent features of its environment via feedback loops which (perhaps continuously) send information from the environment back into the system about the on-going effects of its actions. Such information enables the system to ‘adjust its settings’ accordingly, so its strategies for action are appropriate to the way the world is. This allows the mechanisms in virtue of which mental states have their causal powers to run outside the head.[ 21 ] The use of feedback in the driving-to-the-Pacific case, for instance, is ubiquitous: Theo’s knowing how hard to turn the steering wheel when making a right turn, or how much to let up on the clutch, depends on tactile feedback; when he’s wondering whether he got off the highway at the right exit, signs serve as feedback from the world to either confirm his success, or let him know he’d better turn around and get back on the highway. In these sorts of cases, while the overall intention — to get to the Pacific, for instance — is explicit and fixed, feedback loops allow the fine-grained details of the strategy for implementation of the behavior to be sensitive to contingencies encountered along the way. The intentional system’s constant and skillful reliance on how the world is allows Pacific-directed behavior to be ‘put under the control,’ so to speak, of the Pacific. So it’s ultimately the intentional and semantic rather than neurological properties of neural structures that matter to the explanation and prediction of behavior.
The epiphobe might argue at this point as follows.[ 22 ] ‘Imagine that we find in TT a Neuralese item of the shape, ‘Go to the Pacific.’ Since neural structures can be sensitive to feedback from the environment, they can be just as context-sensitive and predictive of behavior as the intentional properties they happen to token.’ The problem with this argument is that neuroscientific taxonomy doesn’t take account of context-dependent properties of neurological structures, and so the properties that fall out of such a taxonomy can’t be used to predict TT’s behavior with respect to that extra-neural thing in the world, the Pacific. Suppose, albeit wildly (see footnote 19), that a Neuralese item of the shape ‘Pacific’ is discovered and found to be featured in a representation of a goal TT expresses by saying, ‘I want to go to the Pacific.’ While the gray matter in question might indeed be a token of the state, intending to see the Pacific, TT’s Pacific-directed behavior isn’t predictable by an inventory of the gray matter’s neurological properties. Such prediction requires taxonomizing the gray matter according to its context-dependent features, in particular, its intentional properties. This requires discovering what the Neuralese ‘word’ ‘Pacific’ means in such contexts, which in turn, the anti-individualist’s argument tells us, forces us out of the head and into the world (see section 2 above). So while those neural structures may well be sensitive to environmental input, their neurological properties don’t predict behavior.
‘Shoo fly!’ Dretske says as he swats a pesky fly away. As he has it, if a just-then- miraculously-materialized biological Twin-Fred were to similarly move his arm, this Twin-Fred wouldn’t be shooing (Dretske 1990: 13 f.). Shooing acts require the right history; they can’t be shooing acts until intentionality ripens in their perpetrators — until their executors build-up robust connections with the world. You might be inclined to agree, and I think there may be something to this etiological fretting. (How could Twin-Fred’s states be such that they have fly-aboutness, given that he’s never before so much as seen, felt, heard, or even heard about, a fly?) But consider that as soon as Twin-Fred’s own visual motion-detection system locks onto the fly, his hand-waving actions vary counterfactually with its trajectory because of feedback from his perceptual system. Now, I suggest, his state is about the fly;[ 23 ] and the possible variations in his fine-grained actions should be explained in terms of his relatedness to the fly. Fodor’s (1994) right in his claim that in cases like these, ‘it’s the counterfactuals that count for content’ (118). And I need only add that, here, Twin-Fred’s dispositions (/skills) are such that, given the environmental situation in which he’s embedded, and the feedback loops that tie him to it so intimately, his hand would follow the fly withersoever it goest, and his fly-utterances would be appropriate. Mutatis mutandis, his fly-thoughts hereafter would be about flies. (Maybe even herebefore.)
So I don’t think our account is impugned by cases in which strange beings, such as Twin- Freds and Swampmen, miraculously materialize and allegedly produce behavior without intentionality.
7. A final stab
Now that we have a thumb-nail sketch of what sorts of features of an agent might make for potent intentional directedness on the world, let’s see whether we can conjure any epiphobic intuitions about this form of anti-individualism by trying to restate the epiphobic worry. Imagine Thrifty Theo, holding his attitude, intending to see the Pacific. Now, try to imagine a counterfactual in which TT is merely disposed to act just as he actually does, but in which his intentional properties lack their potency. Fix all his dispositions to exploit social and environmental features of his world. Imagine that all the chunks of his world that the anti- individualist intentional state ascriptions presuppose ‘travel’ to the counterfactual. Only try to imagine that the intentional properties are causally (and thereby explanatorily) irrelevant. That is, imagine what the regularities in Thrifty Theo’s environment and the skills which implement the production of his behavior look like; and try telling yourself that if those are the mechanisms on which his behavior supervenes, then it seems the causal powers of intentional properties might be screened off by them, since they’re the more proximal cause of his behavior — they’re what really make him go.
I, for one, can’t conjure the intuition. The fine-grain of intentional directedness suggested by the embeddedness perspective dissolves the phobia. TT’s skills seem irreducibly employed to get to the Pacific. Thrifty Theo may navigate his way Pacificward via different routes given various contingencies and counterfactual discrepancies; but through it all — actual and counterfactual — he still maintains his Pacificwardness. It’s why he does what he does.
This doesn’t seem to me like one of those cases, like John Searle’s notorious Chinese Room or pipe-brain, where our inability to imagine the requisite hypothetical complexity leaves intuition ill-equipped to render a verdict. On the contrary, intentional potency actually seems intuitive here. The moral is that when we start to see how intentionality might be naturalized (a la embedded conception of cognition presented here) we simultaneously begin to dissolve anxiety about intentional impotence. Conceive of intentionality the right way, and the gloomy dread goes away. Embrace embeddedness and the weather takes a turn for the better. As promised: The rain has stopped, the clouds have parted, and the sun is out. Intentional properties do work mere neurological properties could never do. It is what you think that makes you go.
Burge, Tyler (1979) Individualism and the mental, in Midwest studies in philosophy, vol. IV. Peter French, Theodore Uehling, Jr., and Howard Wettstein, eds. U of Minnesota P.
? — (1986) Individualism and psychology, Philosophical review 95: 3-45.
Churchland, Paul & Patricia (1983) Stalking the wild epistemic engine, No?s 17: 5- 18. (Pagination to reprint in William Lycan, ed. (1990) Mind and cognition. Basil Blackwell.)
Davidson, Donald (1970) Mental events, in Experience and theory, Lawrence Foster, and J.W. Swanson, eds. U of Mass P.
Dennett (1983) Intentional systems in cognitive ethology: The ‘Panglossian paradigm’ defended, in Behavioral and brain sciences. 6(3): 343-90.
? — (1990) Ways of establishing harmony, in Information, semantics, and epistemology, ed. Enrique Villanueva. Basil Blackwell.
Dretske, Frederick (1990) Does meaning matter?, in Information, semantics, and epistemology, ed. Enrique Villanueva. Basil Blackwell.
Fodor, Jerry (1980) Methodological solipsism considered as a research strategy in cognitive psychology. Behavioral and brain sciences. 3(1): 63-110.
? — (1987) Psychosemantics. MIT P.
? — (1990) Making meaning matter more, in A theory of content and other essays. MIT P.
? — (1994) The elm and the expert. MIT P.
Haugeland, John (1982) Weak supervenience, in American philosophical quarterly. 19(1): 93-103.
McClamrock, Ron (1991) Methodological individualism considered as a constitutive principle of scientific inquiry, in Philosophical psychology 4(3): 343-354.
? — (1994) Existential cognition. U of Chicago P.
McGinn, Colin (1991) Conceptual causation: Some elementary reflections, in Mind 100(4): 573-586.
[ 1 ] In this paper I take for granted that there are intentional psychological laws which subsume psychological states by reference to their contents. Dissent on this point is okay, since what’s really at issue is the causal potency of intentional properties, not whether they’re featured in laws.
[ 2 ] I don’t find fault with the power-parameter distinction per se, but only with his assumption about what it buys him. After much ado about just what the discriminatory device is, his defense of local supervenience is that it is what we naturally get when we ‘carve off the parameters for later conversion into nomological variables’ (581). This is precisely what the anti-individualist argues is not true — which needn’t keep even him from helping himself to the distinction.
[ 5 ] It’s not too far-fetched to think reflection on the classic Davidsonian (Davidson 1970) view that mental events are subsumed by laws under their physical descriptions might lead one to this fret. That is, one might ask, ‘Wouldn’t such an event have had the same causal properties even if it lacked its mental ones?’
[ 6 ] I suspect that there might be good arguments against epiphobia available even to those who accept local supervenience. Although epiphenomenality intuitively seems more imaginable here, it’s not clear that the truth of the local supervenience thesis about causal powers of intentional properties would show that there are any nearby nomologically possible worlds in which our intentional states lack their intentional properties. But this isn’t my burden here — nor do I think the local supervenience view works in the end anyway. So I’ll proceed to dispose of it, and move on to bigger and better things.
[ 9 ] Prima facie, you may think that in citing these implications I’ve only helped condemn intentional potency by showing how intentional psychology’s in stark contrast with real sciences, like physics. Please. These implications have correlates in lots of higher-level sciences. For a good discussion of some of these points see McClamrock (1991), especially section 3.
[ 10 ] I, for one, am glad that some materialist philosophers will allow this, since, in point of fact, the laws of lots of sciences show that it’s true. A great standard example here is from genetics: Physically type-identical DNA strands can play different hereditary roles — be different genes — depending on their contexts.
[ 12 ] I refer the skeptic to McClamrock (1994), which is, as far as I know, the best philosophical sales pitch around for seeing intelligent behavior and cognition generally as fundamentally embedded phenomena.
[ 14 ] As McClamrock (1994) suggests, there need not even be a language of thought in which there’s a one-to-one mapping between meanings and inner syntactic objects: Meaning can emerge at the level of natural language and organism-object relations — which we might get by allowing that maybe the mechanisms in virtue of which mental states have their causal powers essentially run outside the head (160).
[ 15 ] See Burge (1986) for a discussion in this vein regarding human visual perception. Paul and Patricia Churchland (1983) suggestively dub the kind of content Dretske has in mind here calibrational (‘i.e. something which, by random mutation and on random selection, is tuned to measure, via the excitable cells, certain features of the environment’) (307).
[ 17 ] This is, of course, not completely unlike the thrust of Dretske (1990). Also, Fodor (1994) gives a remarkably similar argument in trying to show how intrinsic properties might be reliably kept in sync with extrinsic ones — only he leaves out the part about it being transcendental in nature (18 ff.). Stories like Fodor’s and Allen’s, which try to show how computation and content might be kept in correlation, are misguided as a result of their disregard for the implications of the context dependence of implementations — as I’ll show below for Allen.
[ 18 ] He shows signs which might be taken as awareness of this when he says ‘… neurological features of those states are sufficient to account for their immediate effects …’ (121; my emphasis). But he clearly doesn’t recognize the robust notion of the intentional I’m defending.
[ 19 ] A referee for this Journal noted that I can’t rule out the possibility that some of the Theos’ neurological properties might remain in sync, ones which might be supposed to keep the Theos looking out for certain types of cues in the environment. However, even if, as we may suppose for ease of imagining, in both Theos is tokened a goal-instantiating neural structure of the shape, ‘Go to the Pacific’, this would be a coincidence — a by-product of the already freaky fact that their histories have been qualitatively identical thus far. The causal powers of their neural structures would vary with counterfactual differences on their respective planets. And the less similar they and their planets become (and the less freaky the hypothetical example becomes), the less we would be tempted to think any neurologically type-identical structures they do have have similar intentional causal powers, and are thereby predictors of similar behaviors. So ‘Pacific’ as it occurs in these two Neuralese tokens is neither synonymous nor paronymous, but homonymous, so to speak.
This talk about the shapes of neurological structures harmlessly over-simplifies things to make a point. In fact, I think the details of the story about how cognitive architecture is implemented are quite a bit messier than my example lets on. Particularly, to switch to the lingo of computational psychology, my hunch is that tokens of syntactic types are not only context-dependent, but also multiply realizable with respect to neurological properties.
[ 20 ] Granted, in cases like these it’s not always clear where epistemological questions about our evidence for content ascription leave off, and ones about the metaphysics of content begin. But I do think we can say this: A determination of content will make reference to the system’s counterfactual properties. If the right counterfactuals are true of it, its states are ipso facto intentional.
[ 21 ] Some AI design theorists are welcoming the idea that intentionality needn’t involve computationally expensive explicit prior internal representation of fine-grained-action- implementation plans (especially in light of how planning for contingencies can lead to combinatorial explosion). The best way to conserve system resources is to adopt a wait-and-see-how-the-world-is strategy whenever possible — which is, in effect, what I’m suggesting.
[ 23 ] Dennett (1990: 25) in effect recognizes this, which is why I find it a little odd that he should resist the ‘direct’ influence of meanings on behavior, as he does four pages earlier, as noted in discussion above.