Burge on Perception
Jerry Fodor, Rutgers University
Tyler Burge has written a long,(fn) disputatious, and sometimes difficult book about cognition,fn
------
fn The Origin of Objectivity is approximately 20 times longer than the reference copy of Winnie
The Pooh that I keep on my desk to use in emergencies; and it contains many, many fewer jokes.
------
with special emphasis on the relation between perception and conceptualization. Among its other virtues, it is a serious work of scholarship. People like me, who can only rarely manage to get their references to stay put, will be awed by its fifty-odd pages of bibliography, which cites not just the usual standards in the philosophy of mind and the philosophy of language, but also such exotica as, for example,: `development in young infants reasoning about the occlusion of objects’, `exoskeletal sensors for walking’ and a lot of other papers drawn from the professional journals of psychology and related sciences. It is a contention of Burge’s that an embarrassing number of the theses that philosophers of mind have held to be true a priori, turn out, in light of empirical findings, not to be true at all. (Burge’s favorite examples are theses about what, in light of their lack of language, animals and prelinguistic infants can and can’t perceive.) Burge is surely right to consider that a scandal. In the old days, philosophers thought that they could arrive at conceptual truths by performing conceptual analyses; so it is perhaps understandable that, when they got around to the philosophy of mind, they ignored scientific results and did their stuff from armchairs. But then Quine pointed out that successful conceptual analyses have been remarkably thin on the ground; indeed, that nobody seems to be able to say just what a conceptual analysis is. The prospects for an analytic theory of perception now seem about as promising as the prospects for an analytic theory of continental drift; and much the same can be said of the prospects for analytic theories of believing, thinking, remembering and other cognitive states and processes .
I think that Burge’s methodological observations about how philosophical research on perception and cognition should be conducted are patently correct; this paper will take them for granted. My primary concern will be the question: `Given that theories of perception/cognition (fn) must be
------
(fn) To conserve back-slashes, I’ll use `theory of perception’ as short for `theory of perception or other aspects of cognition’ except when I think it matters which of the disjuncts is being discussed. I assume (untendentiously, I hope) that percepton is a branch of cognition, hence that the science of perception is a branch of cognitive science. Much of this paper is about which branch of cognitive science it is and how that part differs from some of the others.
------
responsible both to empirical findings and to reasonable demands for conceptual coherence, what theory of percepption ---or, at a minimum, what kind of theory of perception--- should we endorse?’ Burge’s book repeatedly asks philosophers, psychologists, linguists, ethologists and anybody else who is prepared to lend a hand, what is required of a creature, and of its environment, such that the one should be able to perceive the latter ‘objectively’fn I think this is indeed a sort of question that we should all be asking and that those are the sorts of people we should ask it of. But I’ll argue that Burge gets the answer wrong; interestingly wrong, but wrong all the same .
------
Fn, To see a thing `objectively’ is, at very least, not to see it as having properties that it hasn’t got. In particular, it involves distinguishing properties of the percept from artifacts of the perceptual process (like the apparent di minution of distant objects.) On Burge’s view, a number of features of perception are explained by their being conducive to objectivity, the paradigms are the perceptual constancies.
------
So what, according to Burge, does one discover if one addresses theories of perception with due attention to the available empirical results? Primarily that philosophers have over-intellectualized their accounts of how perception works (and also, perhaps, their accounts of how thought does). In particular, Burge thinks that philosophers have done so by holding:
i. that perception typically requires the conceptualization of the percept, so what you can perceive depends on which concepts you have,
and
II. that which concepts you have depends on which beliefs you hold. (fn)
------
Fn This sort of view of the semantic content of concepts is often called `Internalism,' or ‘Individualism'. Burge uses both terms.
------
This, however Is where exegesis gets sticky. As far as I can tell, Burge thinks that he thinks that (i) and (ii) are both false. I think, however, that his position with respect to (II) is, in fact, equivocal; and that the equivocation matters a lot to what he says about related issues of importance. By contrast, his position with respect to (ii) is entirely clear: it Is the core of what Burge calls `anti-Individualism’. I’ll start with (i).
I want to remind you of the traditional distinction between seeing a thing that isan F and seeing a thing (whether of not it is an F) asan F. Both seeing and seeing as can, of course, be instances of objective perception. Burge is perfectly aware of the `see’/’see’ as distinction, but he seems not to think it bears much weight. In fact, he sometimes uses the two expressions more or less interchangeably. That is, in my view, a Very Bad Mistake; I think that Burge’s treatment of the relation between perception and conceptualization depends on his making it.
A revealing passage on p.244 is germane. Burge is arguing (rightly I think) that, in the paradigm cases, the object of sensation is the same as the object of perception: What we sense, like what we see, is a cat lurking in the bushes (not a glimpse, or a `sense datum`, or an array of shapes. colors and textures.... and so forth.) Burge remarks: “...one can immediately and fully apprehend something as a body... It is not evident that it takes extra time, beyond the moment of viewing, to see a body or to see something as a body.” This sounds to me like a confusion of two claims: On the one hand, there’s the (as Wittgenstein might have said ) `grammatical’ observation that, if there really is a cat in the bushes, then seeing it and sensing it may be the very same experience. (They come apart, however, if there isn’t a cat.) On the other hand, there’s the question whether seeing a cat as a cat ”takes longer” than (merely) seeing a cat. Only the second question bears on the issue whether inferential (hence conceptual) process mediate seeing..
This is an empirical issue, as Burge would surely agree; and, to the best of my knowledge, Burge is wrong about the empirical facts. There is lots of evidence that perception of the `sensory` properties of a stimulus typically takes less time than perception of its `object’ properties. You’re faster at seeing a red pencil as red than at seeing it as a pencil (though evaluation of such findings is complicated because the registration of sensory properties is quite often subconscious.) Unsurprisingly, we typically see things as things; but that doesn’t settle whether we typically see them as something else first.
Just what I think has gone wrong is a longish story, but a rough approximation fits in a nutshell: On the one hand, there is no perceiving without perceiving as; and, on the other, there is no perceiving as without conceptualization; and, on the third hand, there is no conceptualization without conceptual content.. That is what’s right about (i) and the tradition in the philosophy of mind and the psychology of perception that endorses it. What’s wrong with that tradition is its forever wanting to infer from `perception requires conceptualization’ to, `perception must be subject to epistemic constraints'. The link between premise (which is true) and the conclusion (which isn’t) is the idea that there are typically epistemic constraints on concept possession; an idea that I take to be false root and branch. Accordingly, the path of virtue is to concede that perceiving requires conceptualization (because perceiving requires perceiving as and perceiving as requires conceptualization ), hence that ( contrary to Burge) a viable theory of perception presupposes a viable theory of concept possession. What should, however, be denied to the last ditch is that a theory of concept possession requires epistemological foundations. Burge accuses Conceptualists (fn) of having over-
Fn. By stipulation, a Conceptualist is someone who holds that perception requires the conceptualization of the percept. On the Conceptualist view, seeing a thing is rather like thinking about the thing: thought and perception both rerquire the application of conceps to their objects. Conceptualists hold that `no concepts no thoughts’ and `no concepts no percepts’ are both true.
------
intellectualized the psychology of perception. But that’s the wrong charge and the wrong target. Conceptualism is OK; seeing something at all requires seeing it as something or other, and seeing something as F requires having the concept F. What’s wrong with the tradition Burge opposes isn’t that it intellectualizes perceiving; what’s wrong is that it epistemologizes concept possession. I’m going to argue that the conditions for concept possessionare, in general, notepistemic; so Burge is right to say, as he repeatedly does, that having the concept F does not require knowing ` what it is’ for something to be an F;or ‘how to tell’ whether something is an F’; or `what it’s like’ to see an F (still less `what it’s like` to be an F’); or what the criteria for F-ness are; or what belongs to `the F stereotype’... and so on all the other epistemological theses about conditions for concept possession that philosophers and psychologists have championed from time to time. In particular, you can be a Conceptualist without holding that concept possession is subject to epistemic constraints; you might, for example, hold that it supervenes on mind- world causal connections. So (i) and ii can both be true; Conceptualism, and Anti-individualism, can both be true. That is, after all, unsurprising; Individualism is about the metaphysics of conceptual content, not about the role of concepts in perceptual processes. Of course there is a robust sense in which you can think about water without thinking about H20 (there is also a robust sense in which you can't). Theories of mind can thus contrive to have the best of both worlds. But, only at a price: They are required to treat the perceive/perceive as distinction with due respect; which is what I think Burge fails to do. That's the short form; details to follow.
Seeing something as an F requires seeing it as an F; just seeing an F does not since you can perfectly well see an F by seeing it as a G. For some reason not clear to me, philosophers like to say this sort of thing in the formal mode: `perceives an F that is G` is transparent at the G position, but `perceives an F as G’ is opaque at the G position. If you see a dog, and the dog you see is brown, then it follows that you see a brown dog. That’s so whether or not you notice that it is brown; indeed, whether or not you notice that it is a dog. By contrast, to see a thing as a brown dog is to see it as brown and as a dog; that is so whether or not it is, in fact, either. To see a red geranium as a brown dog is to see a red geranium as brown and as a dog, even though it is neither the one nor the other. This really is puzzling; there is no rug under which to sweep it. Inthe case imagined, all there is in the world is a you and a red geranium, so how did a brown dog get in?
The immensely plausible answer is that that it got in via the concepts deployed in the seeing. A cat may look at a queen and, in favorable conditions, a cat may even see one. But, though I hold them in great esteem, I doubt that a cat can see a queen as a queen. The reason it can’t is, surely, that doing so requires having the concept QUEEN; which cats do not. At one point (p. 347) Burge remarks that a science of vision will need to “take perceptual states as representational [sic] in a psychologically distinctive sense”, Quite so; but in what psychologically distinctive sense? Answer: it requires a distinction between what is seen and what it is seen as; and `see as ...’ is an opaque context; and seeing something as F requires deploying the concept F. Compare zoology; that a thing is a cat does not raise the question `what is it a cat as?’
So then: On the one hand, nothing that lacks the concept QUEEN can see a queen as such. But, on the other, a cat can’t see a queen without seeing her as something-or-other. All that being so, a cat can’t see anything at all if it hasn’t no concepts at all. Cats without concepts are blind. (Didn’t Kant say something of the sort some while ago?)
I don’t know what Burge would say about that; as I remarked, he seems not to make much of the `see’/’see as’ distinction. It is no help in sorting this out that (as far as I can tell) Burge regularly uses versions of the locution `perception as of an F’ both for perceiving and for perceiving as, thereby eliding the distinction between them. So, for example, when Burge is discussing the way that the constancies militate in favor of the objectivity of perception, he speaks of the perception of a round plate as being `as of something round’ even when the plate is tilted. (p. 383; I’ve slightly altered Burge’s example, but not in a way that affects the present issues). But (in a passage about the mental mechanisms that mediate the constancies) he says that “the transformational story begins with two-dimensional retinal sensory representation and ends with visual perception as of attributes in the physical environment.” 358). This way of talking raises a crucial question that Burge doesn’t answer: Whether you can have a perception `as of an attribute in the physical environment` if there is, in fact, nothing in the physical environment that has that attribute; this is , in effect, the question whether perception `as of an F` can have an opaque reading. That question is crucial; in a moment’ we’ll come to why it is. First, what is almost certainly a digression:
There’s some rough going c. page 152-168 that I think traces to Burge’s not having a robust notion of seeing as . Consider a case where a distal orange orange is displayed in a green light. A subject duly reports seeing a green orange. I would think the natural way to describe this situation is that an orange orange being seen as green, hence as a misperception of an orange orange. This is just the sort of work that `see as’ is good at. But (to repeat) Burge doesn’t have a notion of `seeing as’; so what he says instead is that the representational content of the subject’s perceptual state should be specified relative to the proximal stimulus, not relative to the distal stimulus. Since, as we may assume, the proximal stimulus you get when an orange orange is seen in green light is the same as the proximal stimulus you get when a green orange is seen in a white light, Burge says that the content of perceptual state that subject is in when seeing an-orange orange in green light is `as of’ seeing a green orange. So, to see an orange orange in green light as green isn’t, after all, to misperceive the orange. It counts as veridical perception according to Burge’s way of counting.
But this is surely counterintuitive; and, it’s a strange thing for Burge, of all people, to say. Burge is a committed externalist whatever else he is; and an externalist should specify perceptual content relative to its distal cause. To see something that looks just like water as water is to misperceive it unless the distal stimulus is water. And, anyhow , if seeing an orange orange in green-light as green is veridical, what on earth is it veridical of? Not the distal stimulus since, by assumption, the distal stimulus is orange. Not the proximal retinal array because (unless you are very oddly wired), you can’t see your retina without a mirror. Not your brain, because brains are gray (more less). Not an intentional object that is green; or a green sense datum, or anything of that sort; Burge has no truck with any of those (rightly, I think. Whatever seeing as is, it’s not a relation to a queer object. So what then?
Burge says that his way of understanding this sort of case is not a matter of philosophical interpretation but “of scientiific fact.” (388). I suspect, however, that it’s an artifact of his failure to appreciate the extent to which psychology needs some notion of seeing aseven to describe the phenomena that it is trying to explain. If `see as’ makes problems for semantics, or epistemology, or ontology (or all three at once), they will just have to be faced and coped with.
The digression ends here.
As previously remarked , `see’, `perceive’ and the like are typically `transparent` in the sense that, if you see/percieve an x, there is an x that you see/perceive. By contrast, ‘see as’, ‘perceive as’ , and the like are typically opaque; you can see something as an F even though no F is in view. (Indeed, you can see/perceive something as an F if even if there isn’t anything that you perceive as an F; as when, in a hallucination, you see/perceive the surround as populated by elves. In such a case `the surround’ functions a dummy; it pretends to be a referring expression but really isn’t. ) No news so far.
Suppose, then, that you can have a perception of a plate `as of something round’ even though the plate is square. Then, to put it crudely, a Conceptualist will want to know where the roundness comes from. Well, where does it come from? Not from the plate since, by assumption, the plate is square; but. if Burge is right, not from the plate seen (as one says) `under the concept round , because, according to Burge, perception doesn’t involve conceptualization. This is essential to his case; indeed, Burge’s argument that