We have had occasion to address the problem of intentionality in recent posts on Daniel Stoljar
and Jerry Fodor
. You will recall that the intentionality of a mental state is (as contemporary philosophers tend to understand it, anyway) its meaning or “aboutness,” its “pointing” beyond itself in the way that a word or symbol points beyond itself. For example, when I think that the cat is on the mat
, my thought is about or “points to” the cat’s being on the mat in a way that is analogous to the way the sentence “The cat is on the mat” points to or is about the same state of affairs, or the way a drawing of a cat on a mat is about or points to it. The difference is that while words, pictures, and material representations in general lack any inherent intentionality or meaning of their own – they have intentionality only insofar as we impart it to them – thoughts have it intrinsically. So, if a thought is as material as a word token or a picture is (for example, if it is identical to a brain process), what explains the fact that it has intrinsic intentionality while they lack it? That, in a nutshell, is the problem intentionality poses for the materialist.
You will also recall that I have suggested that what makes this problem so difficult for the materialist is the very conception of matter that he has inherited from early modern thinkers like Galileo, Descartes, Hobbes, and Boyle, who abolished final causality or immanent teleology from their conception of the natural world. For the Aristotelian-Scholastic tradition that the early moderns overthrew, a kind of “meaning” pervades the natural order from top to bottom. This is taken to be manifest not only in the usual examples – the functions of biological organs – but even in basic causal regularities. If some cause A reliably generates a certain specific effect or range of effects B – rather than C, D, or no effect at all – then in the view of the Scholastics, this can only be because A inherently “points to” or “aims at” the generation of that effect, as to a final cause. But it is not only that causes “point forward” to their effects; effects also “point backward” to their causes given the Scholastic principle that whatever is in an effect must in some way be in its cause. (I discuss and defend this principle of proportionate causality in Aquinas.)
When immanent final causality was abandoned and the rest of the Scholastic analysis of causation fell away with it, causes and effects came to seem essentially “loose and separate.” This not only opened the way to the famous Humean puzzles about causation and induction, but also made the intentionality of the mental especially difficult to assimilate to a materialist metaphysics. For if the material world, being (so it was held) devoid of any immanent final causes or teleology, is thereby devoid of any inherent “pointing to” or “aiming at,” then it is hard to see how a mental state’s “pointing to” or “aiming at” something beyond itself can be accounted for in material terms.
With this background in place, let us take a look at some of the ideas of Fred Dretske, whose work on intentionality has been particularly influential within contemporary philosophy. What is sometimes called the “crude causal theory” of meaning or intentionality holds that a material symbol will represent whatever causes “tokenings” or instances of it in a law-like way. For example, instances of the word “cat” will represent cats if they are regularly caused by the presence of cats. Like other contemporary proponents of naturalistic theories of intentionality, Dretske emphasizes that a crude causal theory is inadequate, because it cannot account for cases of misrepresentation. For instance, “cat” can represent cats even when cats do not cause tokenings of it – such as when grandma, due to poor eyesight, tells us that the cat is on the mat even though it is really the dog that is on the mat.
In order to deal with this problem, Dretske, again like other contemporary naturalists, appeals to the notion of function. Let’s consider how this goes in his article “Misrepresentation,” which originally appeared in Radu Bogdan, ed., Belief: Form, Content, and Function (and was more or less incorporated later into Dretske’s book Explaining Behavior: Reasons in a World of Causes).
Dretske begins by introducing the notion of a “natural sign,” which is a more or less reliable indicator of the presence of something insofar as it is typically caused by that something in a law-like way. Hence, expanding metal is a natural sign of a rise in temperature; a northerly-flowing river is a natural sign of a downward gradient in that direction; spots on the face are a natural sign of measles; and so forth. Natural signs can be said to have a kind of “natural meaning,” which Dretske abbreviates as meaningn: spots on the face meann that measles are present, expanding metal meansn that the temperature is rising, etc. Dretske takes the notion of meaningn to be a plausible starting point for a naturalistic account of meaning, but it can, in his view, hardly be the whole story insofar as it does not give us what we need in order to account for misrepresentation. For a natural sign cannot misrepresent anything, precisely because it represents whatever causes it. To use Dretske’s example, when the doorbell malfunctions, what its ringing meansn is not that the doorbell button has been pushed (since it hasn’t been), but rather that there is electrical current flowing in the doorbell circuit. Of course, we might interpret the ringing as meaning that the button has been pushed and thus (under the circumstances) take it to count as a misrepresentation, but in that case the “meaning” in question derives from us, and in particular from the purposes to which we put doorbells, and is not there naturally in the material processes themselves.
A theory that appealed only to meaningn would therefore be no advance on a crude causal theory. What we need to add, Dretske says, is the notion of “functional meaning” or meaningf. As indicated, we ordinarily count even the malfunctioning doorbell’s ringing as meaning that the button has been pushed because it is the function of the doorbell to tell us when someone is pushing it. In this case the function derives, again, from us, but Dretske suggests that there are also natural functions that might provide a naturalistic ground for meaningf and, in turn, for the possibility of misrepresentation. Such functions most plausibly derive from biological need. Dretske gives the example of marine bacteria possessing a sensory mechanism called magnetotaxis, by which they are able to propel themselves along the earth’s magnetic field. For example, such bacteria living in the northern hemisphere are able to propel themselves toward magnetic north. Now the function of this mechanism may be to allow the bacteria, who thrive only in oxygen-free environments, to avoid oxygen-rich surface water. We might say, then, that the meaningf of such a sensory state in one of these bacteria is that oxygen-free water is present in this direction. And this is what the state will meanf even in the case where no such water is present, because (say) we have placed a bar magnet near the bacterium and thereby disoriented it. Hence we seem to have a naturalistic conception of meaning which can account for misrepresentation.
But as Dretske himself acknowledges, even this will not quite do. One problem is that it is not obvious how it accounts for meaning where biological need is not in question, though Dretske thinks the account might be extendable in a way that does account for such cases. What he thinks is a more serious problem is what he describes as the indeterminacy of biological function. If we say that the function of the bacterium’s sensory states is to indicate where oxygen-free environments are, then it seems we have an account of misrepresentation of the sort the naturalist is seeking. But why describe the function of the sensory states this way? Why not say instead that their function is to indicate the direction of geomagnetic north, or even just to indicate the direction of magnetic north? Any of these would be a plausible candidate for the function of the sensory states in question. But if we take the last of these options, then the bacterium’s sensory state does not misrepresent the environment when we place the bar magnet near it. For it really does detect (the bar’s) magnetic north in that case.
That what the bacterium ultimately needs is oxygen-free water rather than to propel itself to magnetic north per se does not help to eliminate the indeterminacy. To borrow another example of Dretske’s, if what I need is vitamin C, it doesn’t follow that any state of my perceptual-cognitive apparatus has to meanf or represent something as containing vitamin C; representing it as a lemon or orange will suffice, since being a lemon or orange is correlated with having vitamin C.
To find a naturalistic ground for the possibility of misrepresentation, then, we need in Dretske’s view to add some further element to the story; and he proposes that this further element has to do with an organism’s having (unlike the bacterium, which is sensitive only to magnetic north as an indicator of oxygen-free environments) more than one way to detect the presence of something that it needs either to seek or to avoid. So, suppose a creature needs to avoid a certain kind of tree that is poisonous to it, and that it can identify the tree either by its leaf pattern or by the texture of its bark. When it has either an internal sensory state I1 which meansn that the leaf pattern is present, or an internal state I2 which meansn that the bark is present, the creature will go into a further state R that leads it to run away. Now R itself in this case does not meann either that the leaf pattern is present or that the bark is present, because there is no regular correlation between either one of those, specifically, and R; either one could cause R. But R does meann that the tree is present, because whether it is via the leaf pattern or via the bark that the tree causes R, it will reliably cause R. And since it is its need to avoid the tree that causes the creature to go into state R, what R functionally means, meansf, is specifically that a tree of the sort in question is present, rather than that the leaf pattern is present or that the bark is present; the indeterminacy that characterized the bacteria example has been eliminated. Moroever, R will have this meaningf even if we present the creature with a fake tree with the same leaf pattern or the same bark texture. Hence we will have, in that circumstance, a case of misrepresentation, and one that can be accounted for in naturalistic terms.
To Dretske’s credit, he acknowledges that even this supplemented account is not free of difficulty. For even if R does not meann that the leaf pattern is present, specifically, or meann that the bark is present, specifically, why could we not say that R has a disjunctive meaningn, namely that R meansn that either the leaf pattern is present or the bark is present? But if we say that, then indeterminacy enters the picture yet again: R could meanf, that is to say, functionally mean, either that a tree of such-and-such a kind is present or it could meanf that either the leaf pattern or the bark is present, and there will be, in this case as in the bacteria case, no non-arbitrary way to favor one over the other as the “true” function or meaningf of R. And if we say that the latter, disjunctive meaningf is the true one, then once again we do not really have a case of misrepresentation at all: When we present the creature with a fake tree, since at least the leaf pattern or bark texture will in that case be present, the creature’s sensory state represents things accurately. Misrepresentation has, therefore, still not been explained naturalistically.
Dretske further acknowledges that this indeterminacy problem will reappear for any even more complex system as long as we can identify for it some corresponding more complex disjunctive property the detection of which might be characterized as its function. His response to the problem is to propose one final wrinkle to the theory. Suppose now that we have a creature capable, through conditioning, of continually adding to the number of properties of the tree to which it is sensitive. Hence while at one point it is sensitive only to the leaf pattern and to the texture of the bark, it might come later to be able to detect also the tree’s visible root structure, still later its average size, and so forth. If, as in our previous example, we think of the meaningn of R in this case as some disjunctive property, then since the disjunctive property in question will change as the creature adds to the properties to which it is sensitive, the meaningn of R will also change over time. And that would entail that, if we thought of the function of R as the detection of this disjunctive property, then that function, and thus the meaningf of R, would also change over time.
Nonetheless, R will still be a reliable indicator of the tree over time, and thus meann that a tree of such-and-such a sort is present over time. Hence, if we are to regard R as having some stable meaningf or functional meaning over time, Dretske says, the only such meaning available for it to have – since the disjunctive property is not stable over time – is that a tree of such-and-such a sort is present. And since it will have that meaningf even when triggered by something other than the tree (e.g. a fake tree having the same leaf pattern, or whatever), we have a naturalistic basis for explaining misrepresentation.
What should we think of Dretske’s account? There are, I believe, several problems with it. To begin at the end, if Dretske’s final revision really shows anything at all, the most it shows is that if R has some meaningf that is stable over time, then that meaningf must have to do with the detection of the tree. But that’s a big “if.” For why not say instead that R has no stable meaningf over time? What is there in the physical facts that determines that R meansf, stably, that a tree of such-and-such a sort is present, as opposed to having an ever-changing series of disjunctive meaningsf? Dretske offers us no answer. But if what R really has is nothing more than an ever-changing series of disjunctive meaningsf, then at any particular time t R will not be misrepresenting that which triggers it. Dretske’s final account is really no less subject to indeterminacy problems than are the accounts he acknowledges to be inadequate.
Of course, that might seem counterintuitive. Surely it is at least extremely plausible to say that R has the function of getting the creature to avoid trees, and does not plausibly have the ever-shifting alternative functions in question? I agree, but the problem is that we need to know how a materialist (like Dretske) can account for this, consistent with the conception of matter we have seen he is committed to, on which there are no inherent ends or purposes in nature. For if a materialist tries to solve the problem in question by postulating such ends or purposes as a way of explaining how R can have a determinate function, then he has thereby ceased to be a materialist and returned to a more or less Aristotelian or Scholastic conception of nature.
Some of the other things Dretske says point in the same direction. Recall that he claims that what he calls “natural signs” have at least a kind of “natural meaning,” even if not the kind of meaning that might explain misrepresentation. But how can a materialist justify even that modest claim? What licenses a description of expanding metal (say) as a “sign” of a rise in temperature, or as “meaning” that the temperature is rising? True, when we observe the expansion, we can (given our background knowledge) infer that the temperature is rising. But how does that show that the expanding metal has, all by itself and apart from our knowledge of it and its properties, a semantic property like “meaning”? Did it have “meaning,” or count as a “sign,” before any human beings were around? “Meaning” to whom, in that case? A “sign” to whom? In fact there would seem to be no “meaning” or “sign” literally present at all until intelligent creatures like us come along and, discovering that a rise in temperature causes metal to expand, come to take expanded metal to “mean” or be a “sign” of a rise in temperature.
Now something like built-in meaning might plausibly be attributed to expanding metal, spots on the face, and other examples of the sort Dretske gives if we were to adopt an Aristotelian-Scholastic conception of nature, on which final causality or directedness-to-an-end is inherent to patterns of what Aristotelians call efficient causality (that is, to what moderns mean when they speak of causality). If, as I put it earlier, causes inherently “point forward” to their typical effects and effects inherently “point backward” to their causes, then Dretske’s examples might do the sort of work he needs them to do. But the whole point of the early moderns’ chucking out of Scholasticism was to get rid of this sort of immanent teleology.
Yet as I have noted many times before (e.g. in this post, and in The Last Superstition and Aquinas) this sort of move is extremely common in contemporary materialism. When proponents of “naturalistic” theories of mind help themselves to notions like “natural meaning,” “biological function,” “algorithms,” “programs,” “information,” and the like, they are either using these terms in a figurative way – in which case they cannot do the work they are intended to do – or they mean them seriously, in which case those who use them are tacitly acknowledging that the Aristotelians were right after all and that modern materialism rests on a mistake. What I like to say of the moderns in general can be said of the work of contemporary naturalists in particular: What’s new in it isn’t true, and what’s true isn’t new.