Wolfgang Schwarz

Review of Tyler Burge: Origins of Objectivity

Earlier this year, I read Tyler Burge's Origins of Objectivity. It's a very long book. Here is an abridged version. A few comments below.

Origins of Objectivity

Representation is a basic explanatory kind in psychology that should be distinguished from mere information-carrying. The most fundamental type of representational state is perception. In perception, an organism attributes properties to objects in its environment. To do this, the organism does not need linguistic capacities, nor does it need to know (or otherwise represent) necessary and sufficient conditions for being the relevant object. Instead, the science of perception reveals that it is sufficient that the organism stands in a suitable causal relation to the object and that its perceptual state involves certain constancies (for shape or colour or distance or whatever) which characterize the object "objectively", abstracting away from contingencies of the present stimulus.

I like the starting point — to think of intentional states as explanatory scientific kinds. Burge doesn't say what exactly he means by this. I would put it as a kind of functionalism: intentional states are characterized (at least in part) by their functional inter-connections and their relationship to environmental causes, behaviour and other psychologically relevant facts.

Burge rightly emphasizes that intentional theories are empirically useful not only when it comes to humans, but also for a wide range of animals. When a jumping spider carefully scans its environment and then goes on a complicated but efficient route towards its prey, it is natural to assume that the spider somehow represents the location of its prey and the chosen route. As Dennett once put it, "one can often predict or explain what an animal will do by simply noticing what it notices and figuring out what it wants". Philosophers are sometimes wary of such ideas because they attach a demanding, hyper-intellectual sense to intentional expressions so that they only figuratively apply to creatures incapable of abstract thought and language.

Sadly, Burge himself falls into that trap. He liberates perception from intellectualist connotations, but all other intentional states remain hyper-intellectual. For example, beliefs are assumed, without argument or discussion, to be "constitutively propositional" [p.538, n.5], meaning that they are directed at effectively linguistic representations and tied to general abilities for abstract logical reasoning [p.542]. Desires or other goal-directed intentional states are not discussed in the book, but presumably Burge believes that they, too, are reserved for humans and maybe a few other "higher" animals.

This hyper-intellectualism about everything except perception creates a lot of trouble for Burge and makes his entire picture much less attractive than it could otherwise have been.

For example, at various points in the book (such as most of chapter 10) Burge insists that representation of such-and-such must be perceptual, even if biological and psychological evidence suggests that it happens on a post-perceptual level. The basic problem can be illustrated with our spider. On its way to the prey, the spider often loses sight of the target. We want to say that it still represents the target's location and/or the chosen path, although this can hardly be part of the spider's perceptual content at times when it can't see the target and the path. Burge is forced to say that it is, because perception is the spider's only representational faculty.

Another problem created by Burge's hyper-intellectualism is that without other representational states, the crucial explanatory/functional role of percetual states remains obscure — as it does throughout the entire book. What are intentional perceptual states good for if they don't affect the organism's belief-like representation of the environment, which then combines with its basic goals or desires to produce adequate behaviour? Burge has to assume that there are interesting and fruitful psychological theories in which perception is the only representational state. It is no coincidence that we are never told even the bare outlines of such a theory.

As a consequence, we don't really know how to adjudicate between different views on the nature of perceptual content. What sorts of properties and individuals are represented in perception, and how? To answer such questions, it would be good if we could look at the wider explanatory and functional role of perceptual representations. But we can't.

Burge instead answer those questions by a priori stipulation. Consider the interesting and substantive question whether perceptual content is generally "qualitative" or whether it instead involves particular objects in the environment. Imagine a frog that sees a fly. The fly is called Fred. Does the frog's visual state merely represent the ("qualitative") fact that some fly (or some black dot) is passing by, or does it represent the ("singular") fact that Fred is passing by? In other words, if we secretly replace Fred by a different but identical-looking fly, does the frog's perceptual content change? There are good reasons to opt for the qualitative answer if our aim is a systematic psychological theory, but the issue isn't straightforward and deserves careful consideration. Burge bypasses all of this and simply stipulates that the singular answer is correct, deriding the qualitative alternative as evidently false and absurd.

Or consider the interesting and substantive question whether perceptual content has "structure" over and above the Boolean structure that automatically comes with accuracy conditions. As came out for example in my discussion with Dave Chalmers on this blog recently, the issue is subtle and turns on the envisaged explanatory/functional connections between perceptions and other attitudes. Again, Burge pays no attention to any subtleties and connections but simply stipulates that perceptual content has structure. (He's not shy about it. "Perceptual representational contents are necessarily and constitutively structures. This claim derives from apriori considerations about perceptual veridicality conditions. All perceptions must have representational content that has both general and singular elements." [p.380])

In fact, Burge usually talks about perceptual content as if it had a basically sentential form, composed of singular terms (including three to four demonstratives [p.451]) and predicates. In successful perception, the singular terms pick out objects in the environment and the predicates properties of these objects: shape, size, location and the like. We are never told why we should think of perceptual content in this peculiar way. When I look at the cloudy sky, how many singular terms figure in my visual content? Is there a single demonstrative for the entire sky, or one for each cloud? Does it matter how my visual system slices up the clouds?

It is also strange to think of structures made of predicates and singular terms as contents. Aren't they rather vehicles or bearers of content? The actual content could then be understood in terms of truth-conditions, or structures of denotations. Burge often talks that way, too, speaking of perceptual states as attributing properties to individuals in the environments. But that isn't quite right, as we learn in chapter 9. Here the true content of perception is identified not in terms of external individuals and properties, but in terms of modes of presentation: "the representational content [...] consists in modes of presentation as of particulars, and modes of presentation as of attributes that are perceptually attributed." [p.385] The "as of" construction is meant to allow for cases of illusion and hallucination where the relevant individual or attribute is absent.

Burge gives two reasons for this complication: the perspectivalness of perceptual content, and hallucinations. As he points out, disjunctivism is alien to the science of perception. The typing of perceptual states most relevant to psychological explanation classifies veridical perceptions of the fly Fred with veridical perceptions of other flies and with halluzinations brought about by direct stimulation the retina. In general, if we hold fixed the rest of an animal's psychological state, then the content of its perception is actually determined by its proximal sensory stimulations [p.385].

At this point, one can't help but wonder why Burge spent hundreds of pages insisting that individuals in the environment directly figure in perceptual content, if his own view actually renders that false. The puzzle isn't resolved.

Another oddity about Burge's account concerns the importance of constancies. For Burge it is "a constitutive fact" that perceptual content involves constancies such as colour or brightness that don't simply track physical features of the incoming stimuli. (I suppose what actually figures in perceptual content are modes of presentation of such constancies, but let's ignore that.) This abstraction from the proximate stimulus Burge calls objectification. "A perceptual system achieves objectification by — and I am inclined to believe only by — exercising perceptual constancies." [p.408]

As it turns out, not all perceptual systems or perceptual states make use of constancies. According to Burge, states that do not aren't perceptual states at all (by stipulation). For example, since there are few constancies in smell and taste, Burge concludes that these "seem largely to be non-perceptual sensory systems" [p.415]. Similarly for sensations of heat or cold or pain [p.415, p.421]. Burge admits that when we feel heat or pain in our limbs, there is a "feeling of location", but he claims that there is nevertheless "no explanatory need to invoke veridicality conditions or representational content" [p.421]. The poor honey bee fares even worse. Since its visual states generally lack the required constancies, they have no representational content: bees "do not rely on perceptual capacities when they are flying fast in route" [p.508]; "Nothing in the explanation of [the bee's] state needs to take it as representing, as having veridicality conditions at all" [p.504].

These are striking claims. If my sensations of smell or temperature have no content whatsoever, what's the systematic story about how they affect my beliefs? Doesn't my sense of temperature sometimes misrepresent an object's temperature? Can't we explain the bee's action by noticing what it notices and figuring out what it wants? Why should all this be impossible just because the relevant sensory states don't involve suitable constancies? Why should that be a necessary condition for having content at all?

For Burge, the answer is: by stipulation. Here we suffer the consequences of knowing so little about the target. Is there "explanatory need to invoke veridicality conditions" for sensations of smell? No, says Burge — but not because he has considered any explanatory projects in which one might want to appeal to such veridicality conditions. He simply notes that his demanding sense of "veridicality conditions" and "content" doesn't cover sensations of smell, and that's the end of the story.

Burge does register that biologist do attribute content to the bee's visual states. Without considering why, he simply declares that as a mistake. This is rather ironic for a book that advertises itself as following the science and criticizes most other philosophers for not doing the same.

Burge certainly does cite a lot of articles from biology and psychology, but these citations are used mostly as decorative ornaments for unsubstantiated a priori declarations. There is little real engagement with developments in the science of perception. No word on Bayesian models of perception. These very successful models rely on a notion of perceptual content that goes beyond statistical correlation but isn't tied to constancies. Burge leaves no room for such a notion.

In sum, I found this a rather frustrating read. It's made worse by a repetitive and obscure style of writing that often borders on the continental. Here, for example, is Burge's characterization of his main enemy, individual representationalism. It is the view that "an individual cannot empirically and objectively represent an ordinary macro-physical subject matter unless the individual has resources that can represent some constitutive conditions for such representation" [p.14]. What could that possibly mean?