What is the point of attempting to make a case for cognitive impenetrability of visual perception?

Boris Crassini,a Jack Broerse,b R. H Day,c Christopher J. Best,a and W. A. Sparrow,a
a School of Psychology, School of Human Movement, Deakin University, Geelong, VIC 3217, Australia; b School of Psychology University of Queensland, St. Lucia, Queensland 4072, Australia; c School of Psychological Science, La Trobe University, Bundoora, VIC 3083, Australia
buc or cjb or sparrow@deakin.edu.au broerse@psy.uq.edu.au r.day@latrobe.edu.au www.hbs.deakin.edu.au/psychology
www.uq.edu.au www.latrobe.edu.au

Abstract: We question the usefulness of Pylyshyn's dichotomy between cognitively penetrable and cognitively impenetrable mechanisms as the basis for his distinction between cognition and early vision. This dichotomy is comparable to others that have been proposed in psychology prompting disputes that by their very nature could not be resolved. This fate is inevitable for Pylyshyn's thesis because of its reliance on internal representations and their interpretation. What is more fruitful in relation to this issue is not a difficult dichotomy, but a different look at perception such as proposed by Gibson (1979).

When one begins with the assumption that the function of vision (early or otherwise) is to provide the perceiver with a "structured representation of the 3-D surfaces of objects sufficient to serve as an index into memory," a fundamental problem arises. The assumption of early vision as a representation-producing process involves the further assumption of a representation-interpreting process. As we know, however; this admits the further and thornier problems of infinite regress and solipsism (see Katz 1983). Pylyshyn seems to allow as much in his discussion of what he calls "Bruner's influential theory": If "poor folk" actually see (not metaphorically "see") a coin as larger than the same coin seen by "rich folk" then the misperceptions and consequent maladaptive behaviours that may occur in folk who are in various other deprived states do not bear contemplation. Given this unsatisfactory state of affairs it is appealing to be told that in the visual brain there are mechanisms that eliminate the more fanciful or even dangerous interpretations of representations "provided" by early vision. Even more appealing is the guarantee that these "smart" mechanisms are unable to be influenced by the perceiver's motivations, beliefs, and knowledge; that is, the mechanisms are cognitively impenetrable. However, it is our view that the solution to the problems inherent in theories of perception based on the generation and interpretation of representations provided by Pylyshyn's notion of cognitive impenetrability is problematic on several grounds which we explore in this commentary. We conclude that a better solution to the problem is provided by Gibson's approach to perception.

The origins of Pylyshyn's current version of cognitive impenetrability can be seen in his 1973 publication "What the mind's eye tells the mind's brain: A critique of mental imagery." In this he argued that "the picture metaphor underlying recent theoretical discussions of...[the relationship between perception and imagery]...is seriously misleading - especially as it suggests that the image is an entity to be perceived" (1973, p. 1). As an alternative Pylyshyn (1973) proposed that mental representations of the world were abstract rather than picture-like, and involved, for example, propositions in the form of pre-compiled subroutines, or rules, which were the interpretation. Part of Pylyshyn's argument rested on the claim that there was no necessary connection between a person's introspections and the nature of mental representations. An extension of this disconnection between conscious awareness and mental representation is at the heart of the earlier (Pylyshyn 1980; 1981) and present versions of cognitive impenetrability; that is, not only are mental representations opaque to inspection by the mind's eye, they are immune from influence by the mind.

Pylyshyn's choice of mental imagery as the vehicle to present his notion was not surprising given that imagination can properly be regarded as a quintessentially cognitive activity. More surprising, however, is the conspicuous absence of consideration of imagery in the present elaboration of cognitive impenetrability. This omission cannot be explained on the grounds that the problems at the heart of the imagery debate have been solved (see Tye 1991). In the 1980s psychophysical data initially interpreted as revealing functional equivalence between imagery and perception were "explained away" in terms of, for example, tacit knowledge (Pylyshyn 1980; Richman et al. 1979), or on methodological grounds (Broerse & Crassini 1981; 1984). More recently, claims of functional equivalence have been made based not only on psychophysical data, but also on brain-imaging data (e.g., Kosslyn et al. 1995). The problems, discussed over many years, inherent in interpreting psychophysical data taken to demonstrate equivalence between imagery and perception still remain. Furthermore, as Sarter et al. (1996) point out, interpretations of brain-imaging data are also problematic: although these data are typically taken to reveal something about the role of neural activity in "causing" mental states, Sarter et al. (1996) argue persuasively that such conclusions cannot be drawn, given the nature of brain-imaging data.

It may be that Pylyshyn's failure to engage in further discussion of the relationship between imagery and perception reflects an appreciation of the fact that the current imagery debate, like the debate about imageless thought at the start of this century (see Boring 1950), and the debate about the role of representations in perception, is more about faith than about fact. That is, given their nature, mental states do not afford objective measurement enabling the carrying out of an experimentis crucis. In addition, the imagery debate and indeed the debate about the role of representations in perception involve the positing of dichotomies of the type that Hinde (1974) describes as having "bedevilled" the history of behavioural science. In this context he makes special men-Hon of the particularly problematic nature of dichotomies that in-valve sources of information, or types of behaviour, or underlying processes defined solely in negative terms. Consider, for example, the disputes that have occurred regarding the distinction between motives that are learned versus motives that are not learned (i.e., instincts); and the distinction between development that is based on interaction with the environment versus development in the absence of interaction (i.e., maturation). Hinde's (1974) view is that the disputes which have "plagued" research in these areas are fundamentally insoluble because those involved in the disputes have "adopted different theoretical approaches and did not see that they differed primarily because they were interested in different questions" (p.37, our emphasis). To the list of such dichotomies can be added another: top-down interpretation of representations that is based on perceivers' beliefs, attitudes, and knowledge (i.e., top-down processes that are cognitively penetrable) versus top-down interpretation of representations that is not based on these mental phenomena (i.e., top-down processes that are cognitively impenetrable).

Our concern is that Hinde's pessimistic analysis of the earlier dichotomies can be generalised to this one. This is despite the evidence adduced by Pylyshyn to support his claim for "perceptual intelligence" in which "the visual system appears to 'choose' one interpretation over other possible ones, and the choice appears remarkably 'rational'... [but it does not involve] cognitive penetration." Pylyshyn's evidence relies heavily on the operation of "smart" mechanisms which through their operation produce "natural constraints on vision" that provide better guarantees of unique interpretation than do, for example, other cognitively-impenetrable top-down constraints such as the Gestalt principles of perceptual organisation. We are somewhat puzzled at Pylyshyn's implication that natural constraints such as Ullman's rigidity constraint are different from (better than?) earlier Gestalt principles. What is Ullman's rendition of a rigid object under rotation if it is not a generalised form of shape constancy (Broerse et al. 1994), and in turn, what is shape constancy if it is not a generalised form of size constancy and size-distance invariance (Koffka 1935)? Furthermore, as Pylyshyn indicates in the qualification he sets out in Note 12, "additional constraints must be brought to bear" to account for more complex cases of motion such as non-rigid biological motion. This qualification hints at the criticism of post-hoc-ness and tautology often aimed at the use of Gestalt principles, and their like, as explanatory devices. But more importantly, Pylyshyn's dichotomy between cognitively-penetrable top-down processes and cognitively-impenetrable top-down processes begs the question of what experiment could be designed to differentiate between these processes? Our answer is that any such attempt would inevitably lead to the kinds of dispute that characterise the imagery debate, and is ultimately futile.

The solution is to ask a different question. Rather than beginning with the assumption of a dualism between the environment to be perceived and the perceiver, adding to this a further dualism between the environment and a representation of the environment "inside" the perceiver, and then adding yet another dualism between top-down processes that are or are not cognitively penetrable, a more fruitful approach is to eschew all three dualisms. It is not enough, as Katz (1983) pointed out, to replace pictures in the mind's eye with propositions in the mind's mind; both pictorial and abstract forms of representation are faced with the problem begging an assumed interpreter. In Katz's (1983) words: "The regress can only be avoided if the whole organism is made the interpreter, and representations are given their appropriate place; in the external world, not inside heads" (p. 269). This, of course, is Gibson's approach in which perception is considered an achievement of action rather than a process of interpreting internal representations. The difference between Pylyshyn's and Gibson's approaches to perception is made clear by Gibson's distinction between direct perception in an environment, and indirect perception of pictures (i.e., representations). According to Gibson (1979) the latter "requires two kinds of apprehension" (p. 291); a direct perceiving of the surfaces comprising the picture, and an indirect (interpreted) awareness of what these surfaces depict. But Gibson's distinction is another story.