Burge on Perception and Representation

Saturday I attended a workshop on the predicative structure of experience sponsored by the New York Consciousness Project, which is in turn sponsored by the New York Institute of Philosophy . Speaking was Tyler Burge and Mark Johnston with commentary by Alex Byrne and Adam Pautz respectively. I may write a separate post on Johnston’s talk but here I want to say something about Burge’s talk.

The first thing that Burge wants to do is to clarify the notion of representation in the claim that perception is representational. For a state to be representational is for veridicality conditions to be an ineliminable part of the scientific explanation of the formulation of the state. Thus representational states, in this sense, are not states that merely co-vary with some thing in the world. For instance, the level of mercury in a simple thermometer causal co-varies with the temperature but Burge wants to deny that the mercury level in the thermometer represents the temperature in his preferred sense. This is because the scientific explanation of how the mercury level came to be such-and-such proceeds “from the inside” so to speak, and does not need to bring in notions like true or false. He admitted that we could, if we want, adopt a certain stance towards the state and call it representational. But there is still something unique about the kind of states that psychologists are interested in. The central task of perceptual theories, for Burge, is that of discovering the conditions under which we correctly, or accurately represent the world and when we fall into mistakes, i.e. illusions. The idea of “getting it right” does not enter into the explanation of why the mercury is at a certain level in the thermometer. The explanation for the level of mercury being thus-and-so as opposed to such-and-such would be the same whether or not we just started with the proximal stimulation or not.  Here is how I understood this point. When we calibrate a mercury thermometer we take the contraption and put it in ice as it is melting (i.e. the just melted ice-water) and wait for the mercury to stabilize. We then do the same for boiling water and then assign 0 to the first mercury level and 100 to the second and divide the space between these two marks into 100 equal parts. Does it really make sense to say that the thermometer got the temperature right in the first step? Can we make sense of the notion that it got it wrong? Wherever it settles we call 0 so how could it be mistaken? Or to put it slightly differently, how could we make sense of the notion of it being under some illusion? These kinds fo considerations don’t even seem to apply. Now as already said we can adopt this sort of talk if we want to, but if it is really the case that nothing is lost when we stop talking that way then it is just a stylistic thing. When we talk about representational states in perception we are immediately confronted with truth-value talk. And Burge wagered that psychology as a science would not give up this notion.

For my own part I find this distinction quite plausible but I don’t see why it then follows that no causal or teleological theory can work. The special category isn’t representation, it is mental representation. Mental representations just are representation for the subject in precisely the sense that matters. It doesn’t seem far fetched to think that we might have evolved systems that are in the business of getting it right and so it seems to me that someone who accepts a causal-teleological account could still accept this distinction.

But back to Burge. The second thing that he wanted to clarify was the notion of perception. He first distinguished between mere sensory registration, which is just statistical co-variation, and perception. For a state to be a perceptual state it must be in the business of objectifying the world. That is, it is in the business of offering a solution to the underdetermination problem. The classic example is the construction/recovery of a 3-d image from the 2-d image projected onto the retina. There are an infinite number of ways that the brain could generate a 3-d image from that information but of course the perceptual systems are in the business of ignoring most of those. This fact is then used in the explanation of various visual illusions. The mark of perceptual systems are perceptual constancies. So, consider color. We Human Beings are pretty good at telling the actual color of a thing in a variety of lighting conditions. That is, our perceptual systems somehow take a range if inputs and treat them as the same. The same is true for length, etc. He distinguishes perception from any kind of mere sensory registration and seemed to think that olfaction and taste were non-perceptual senses. The reason that he gave was that there were no smell constancies. We don’t seem equipped to be able to track the same smell under a bunch of different environmental conditions. But he did admit that this is an empirical question. 

Finally, and perhaps the most interesting part of the discussion, he defended his claim that talk about perceptions as representational and perceptual processes as computational does not commit us to a purely syntactic view of how it is implemented. So by way of illustration consider the way that we talk about the logical category of being a predicate. We can give a purely syntactic description but the level of explanation that matters is the level at which semantic information plays a role in individuating the state. This is somewhat like a point that has always bothered me. In logic we pretend that we are dealing with purely syntactic rules, but they are useless unless they are individuated semantically. But anyway, Burge’s point was that he thought that we would not even be able to get to the point where we could individuate perceptual states in a purely syntactical way, unlike the logic case where we can (so he thought). He speculated that the reason so many people think that you must do computational psychology purely syntactically is because of an antecedent position on the mind/body problem. It is because people start with the assumption that this stuff must ultimately be physical and so we must only apply to physical properties (viz. syntax). But Burge objected that we should do psychology autonomously and the see what the mind-body problem looks like afterwards. There is nothing in computational theory that would force us to opt for a purely syntactic theory of computation and so, Burge claimed, no reason that people who accept the language of thought hypothesis, even for perception, were committed to these computations being done on the basis of purely syntactical properties of the computata (if that is a world).

Now all of this is only by way of clarifying what perceptual representation is (!) and after this is done he goes on to talk about the structure of perceptual representations. This is getting long so I will make it short now and perhaps come back to it later for a fuller account. The basic idea seemed to be that perceptions must be composed of a general attributive part and a part that indicates some particular thing (singular reference part). So, for example, to perceive that one object is to the left of another object is to have a state that represents the general relation of ‘to the left of’ as being instantiated by these two particular objects. In both its general, attributive aspect, as well as the singular aspect perception is always trying to demonstrate some particular. It can fail to do that but it is always trying to do so. He seemed to want to model this on demonstratives. As I said there is much more to be talked about, including his view that the difference between conscious and unconscious perceptions might lie in some aspect of the states mode of presentation, but this is already too long!

[cross-posted @ Philosophy Sucks!]