Are sensory properties represented in perceptual experience?

Abstract

Philosophers of perception widely agree that sensory properties, like color, are represented in perceptual experience. Arguments are usually needed to establish that something other than sensory properties, for example three-dimensional objects or kind properties, are part of perceptual content. Call the idea that sensory properties are represented in perceptual experience the Sensation View (SV). Given its widespread acceptance, we may expect to find strong reasons for holding SV. In this paper, I argue that we lack such reasons: SV is largely unjustified. We have surprisingly poor justification for thinking that sensory properties are represented in perception. By focusing on the case of vision, I show that an embedded understanding of visual perception, and empirical evidence in cognitive psychology, indicate that SV is far from warranted.

Keywords:

• Colors
• Content
• Perception

Acknowledgements

I am indebted to my colleague Casey O’Callaghan for useful comments on earlier drafts of this paper. I thank my former classmate Dylan Sabo for pointing me to the relevant empirical evidence in developmental psychology.

Notes

[1] There is considerable debate concerning what to include as part of the visual system. I am sympathetic to the view that the visual system includes a moving head and body (Noë, 2004; Wilson, 2004). But, for present purposes, we can assume the less controversial picture of the system as including only the eye with its retina and optic nerve, and the visual cortex. I don’t think that the main argument in this paper turns on this relatively narrow conception of what the visual system includes.

[2] I think that these two conditions on representation are in line with how the notion is used in cognitive science. To just give a couple of examples, William James remarks that we impute mentality to beings only when what they do is explained by their “idea” of what the conditions are, rather than by the conditions themselves (James, 1892/1950, chapter 1). A similar approach is found in the more contemporary work of Rodney Brooks. Brooks remarks that there is no need to talk of representation when “to a large extent the state of the world determines the action of the creature” (1991, p. 150).

[3] J.J. Gibson (1979) is sometimes interpreted as denying that perception is representational, but contemporary Gibsonians, like Alva Noë (2004) do not follow Gibson on this point.

[4] I take this to be part of the debate concerning the distinction between observation and theory in philosophy of science. Those who argue against the distinction (Churchland, 1989) maintain that any representation obtained from sensory states is informed by one's conceptual repertoire, and is, therefore, theory-laden. Defenders of the distinction, on the other hand, argue that perceptual processing is insulated from the influence of one's conception of the world, and the representations that it produces are thus genuinely perceptual, or theory-neutral (Fodor, 1984, 1988).

[5] See my Seeing in practice: Putting vision in its place (unpublished manuscript), available upon request.

[6] I am open to the idea of there being a sense of “computation” under which the embedded perceptual systems I describe count as performing computations. See Wilson (2004) and Haugeland (1998) for arguments in favor of extending the notion of computation beyond what traditional digital computers do. At the very least, embedded perceptual systems are computational in the sense that they can be described as performing computations, even if they do not literally do so.

[7] I am indebted for this idea to, among others, Zenon Pylyshyn (1999) and Robert Wilson (2004). Pylyshyn talks about the visual system embodying the assumptions rather than representing them; Wilson talks about exploiting the assumptions or using exploitative representations of them. I prefer the jargon of relying on regularities in order to make clear that the assumptions are not encoded anywhere in the visual system. See also Sabo (2006) for a similar model of our mechanism of concept acquisition.

[8] This reflects a distinction between explanatory reasons and justificatory reasons that is spelled out by Fred Dretske (2006). Explanatory reasons are facts that explain, or help explain, why something happens: they are the reasons why something happens. Justificatory reasons, on the other hand, are given by the way in which facts are represented to be: they are reasons for something to happen. Justificatory reasons are also explanatory, but they explain by making reference to the way things are represented to be. In the fire alarm case, the fact that smoke particles deflect the direction of light is one of the explanatory, not justificatory, reasons for what the alarm does, because the fire alarm does not represent that fact.

[9] I leave it open whether the visual system is not only embedded in the environment, but it also forms a coupled system with the environment, and can be said to extend beyond the skin's boundaries (Clark, 1997; Wilson, 2004). There is considerable debate concerning the plausibility of the extend view (Adams & Aizawa, 2008; Rupert, 2004) and I do not have space to approach the debate here.

[10] Questions arise concerning the similarities between sense data and the sensory states posited by contemporary vision scientists. The latter are typically taken to be representational states, and thus states that can misrepresent. Perception can go wrong both in representing an apple when there is no apple and in representing something red when there is nothing red. By contrast, the perception of sense data was typically assumed to be not only direct but also infallible. One couldn’t be wrong about the sensory appearance of an object: one could only be wrong about the presence of the object. In other words, one could be wrong about the presence of an apple, but one could not be wrong about the presence of something red (the sense datum). This and the introspective accessibility of sense data are two important differences between traditional sense data theory and contemporary inferential accounts of visual processing.

[11] Defenders of inferential/computational accounts sometimes appeal to perceptual shifts rather than misperception in order to defend their view (Churchland, 1989; Rock, 1983). Inferential accounts are said to be better able to explain how we can have different experiences of the same figure, like when we shift from seeing a duck to seeing a rabbit in the famous duck-rabbit figure. In such cases, the visual system seems to interpret the same visual stimulation in different ways suggesting the presence of an inference. Elsewhere, I provide an embedded account of perceptual shifts, and explain how the embedded visual system can produce different experiences without performing inferences. See my Embedded seeing-as: Multi-stable perception without interpretation (unpublished manuscript) and Searching for the duck: A situated account of aspect shifts (unpublished manuscript), available upon request.

[12] This is part of the reason why David Marr (1982) thought that vision proper ended with the representation of shapes rather than with the representation of meaningful objects. Obtaining a representation of objects, according to Marr, requires matching the visual percept with stored, and more complex, representations.

[13] There are certainly other ways of acquiring concepts. One can acquire the concept of a unicorn, for example, without ever perceptually encountering one. But it is undeniable that perceptual experience is one important way in which we get our concepts, and this is all that I am assuming in this paper.