An Introduction to the Science and Philosophy of Mental Imagery
Nigel J.T. Thomas
Page 2
Source: http://www.imagery-imagination.com/mipia.htm
2. Imagery in cognitive science
§2.1 The imagery revival
A revival of research on imagery was an important element of the cognitive revolution of the 1960s and 70s, contributing greatly to the rising scientific interest in mental representations. Seemingly, this revival initially stemmed largely from applied psychology research on sensory deprivation and on hallucinogenic drugs (Holt, 1964). Another important catalyst was Yates' (1966) seminal historical work on the significance of imagery mnemonics in Ancient through Renaissance thought. Once the powerful mnemonic properties of imagery were experimentally confirmed (Paivio, 1971), imagery could no longer be dismissed by psychologists. Interest was only heightened, during the 1970s, by the stunning "mental rotation" experiments of Shepard and his students (Shepard and Cooper, 1982), and experiments by Kosslyn (1980) demonstrating "mental scanning" and related effects. This work was taken to show that imagery is involved in visuo-spatial reasoning, and has inherently spatial properties.
§2.2 The "analog-propositional" debate
But how could these findings on imagery be reconciled with the functionalist, computational symbol manipulation approach to cognition that was emerging during the same period? The standard philosophical interpretation of this approach depicts cognition as the computational manipulation of representations expressed in "mentalese" (the "Language of Thought") a hypothetical, essentially language-like, representational system supposedly built into the brain (Fodor, 1975). Two rival approaches arose toward integrating the empirical findings about imagery into computational cognitive science.
Pylyshyn, in a series of influential papers (e.g. 1973, 1981), argued (in effect) that all the genuine phenomena associated with imagery (indeed, all truly mental phenomena) can and must be explained entirely in terms of mentalese representations. For Pylyshyn and his allies, the computational paradigm of cognitive science demands that the underlying representational reality of imagery (and of actual perceptual experience) is not picture-like, but rather a detailed mentalese description of a scene.
Other cognitive scientists, however, notably Shepard and Kosslyn, argued that the evidence implies that imagery must be a distinct, non-language-like form of representation. Kosslyn, in particular, developed a "quasi-pictorial" computational theory of visual imagery, based on an analogy with computer graphics (Kosslyn, 1980). Computer graphics files store information in a compressed, non-pictorial form, but when they are displayed they are translated into a mathematical map (bitmap) of the computer monitor screen, that specifies the color at each pixel (tiny dot) on the screen itself. Likewise, suggests Kosslyn, visual information may be stored in the brain as compact descriptions, but we experience an image only when this information is used to create a two dimensional map of visual space in a special, functionally defined memory area he calls the "visual buffer". The picture in Kosslyn's theory is merely "quasi", because there is no equivalent to the monitor screen to display it. What we experience as imagery, and what is available to the cognitive processes that use imagery, is the functional picture, the mathematical map, in the visual buffer. In later work, Kosslyn (1994) identifies this "visual buffer" with the several retinotopically mapped visual areas of the brain.
"Description" and "quasi-pictorial" theorists disagreed sharply over what sorts of computational symbols, or data structures, are acceptable within cognitive theory, and which best capture the empirical properties of imagery. During the 1970s, in particular, this led to a lively and high-profile controversy, commonly, if somewhat misleadingly, known as the "analog-propositional" debate. ("Picture-description" debate would have been better. "Proposition" is jargon borrowed from philosophy, where it signifies the underlying meaning of a sentence, not, as is intended here, a descriptive "sentence" of mentalese. Furthermore, the force of "analog" in this context is hardly clear: Kosslyn, after all, models his quasi-pictures on digitized bitmaps.)
Although the leading combatants in this dispute were psychologists, and experimental evidence was frequently cited, many of the issues raised were conceptual or meta-theoretical in nature - Anderson (1978) questioned whether it was even possible to resolve the debate experimentally - and philosophers soon became involved. The very concept of mental representation seemed to be at stake. Many of the most influential articles from the heyday of this debate, by both psychologists and philosophers, are collected in two volumes edited by Block (1981a, 1981b). Description theory still finds philosophical defenders (e.g. Slezak, 1995), but Tye (1991; see also Rollins, 1989) has undermined much of its appeal with a persuasive defense of the conceptual legitimacy of quasi-pictorial arrays as a distinct form of computational representation. Furthermore, many descriptionist explanations of empirical findings seem worryingly ad hoc (Kosslyn and Pomerantz, 1977).
However, Kosslyn's (1994) declaration of victory in "the imagery debate" may be premature, even though he has certainly developed the venerable picture theory to an unprecedented level of empirical and conceptual sophistication. His (1994) recasting of quasi-pictorialism in neurological terms does little to resolve the significant problems it still faces. Pylyshyn (2002) has now launched a major counter-attack, not only restating his empirically and conceptually based objections to quasi-pictorialism, but arguing forcefully that (despite many claims and some superficial appearances to the contrary) there is no firm evidence whatsoever to support it. Even the (much disputed) results suggesting that visual imagery experience is correlated with activity in retinotopically mapped visual cortex (e.g. Kosslyn et al, 1995) are quite consistent with non-pictorial theories (Pylyshyn, 2002; Thomas, 1999). Furthermore, quasi-pictorial theory does not parsimoniously account for a range of experimental results showing that people have difficulty, in many circumstances, in finding new representational meanings in their imagery, meanings that are relatively easily found in an actual picture (Slezak, 1995; Thomas, 1999). Also, it is unequipped to explain the fact that the experimental effects that it most directly addresses (mental rotation, mental scanning, mnemonic effects, etc.) have since been demonstrated to occur in congenitally blind as well as in sighted subjects. The blind experimental subjects apparently employ haptic (touch) imagery, but any haptic analogue of a quasi-picture would be quite unsuitable to play an equivalent explanatory role (Thomas, 1999). In addition to these empirical shortcomings, quasi-pictorialism fails to address two cardinal characteristics of imagery: its intentionality and its consciousness (see §3.2 and§3.3 below).
§2.3 Beyond pictures and propositions?
Outside the theoretical context of symbolic computationalism, in which the "analog-propositional debate" first arose, other accounts of imagery, neither pictorial nor descriptional, may become conceivable. Despite a handful of connectionist simulations of versions of quasi-pictorialism, the connectionist movement has had surprisingly little impact on imagery theory. However, more recent alternative approaches to cognition, such as "dynamical systems theory", and "situated" or "embodied" cognition, call into question the basic assumption that mental contents are to be identified with computational representations (in the sense of data structures, embodied as brain states and manipulated in the cerebral computer). Related work on both robotic and human vision (and perception in general) is converging on the idea that perception is notbest understood as the processing of sensory input into a detailed inner representation - a description or depiction, of the scene before us - but rather as ongoing, directed exploratory activity (e.g. Landy et al, 1996; O'Regan and Noë, 2001). On this view, as our brains direct all the most minute details of our ongoing behavior (including the exploratory perceptual activity itself) they require a constant stream of answers to specific questions about the detailed disposition of the environment, and, instead of seeking this information in a pre-established inner representation, they deploy the sense organs, like measuring instruments, in order to obtain the answers from the environment as and when needed. In order to explain how brains coordinate this activity, we may well need to invoke data structures in the brain. However, they would not encode descriptions or depictions of the environment, but, rather, the procedures that most appropriately direct its exploration. O'Regan and Noë (2001) suggest that perceptual experience is not the experience of having a representation (a percept) in the brain, but, rather, of being engaged in ongoing perceptual exploration. In similar vein, Thomas (1999) proposes that imagery experience (and the empirical data on imagery) is best explained as arising from a sort of abortive, and largely covert,perceptual activity: a truncated "going-through-the-motions" of exploring objects or situations that are not actually there to be explored, under the control of an appropriate procedural representation.
Meanwhile, certain philosophers broadly sympathetic to this view of imagery, and with doubts about the "standard" computational-functionalist view of the mind, have begun to revive the traditional conception of imagery as the vehicle of conscious thought, and the fundamental bearer of mental content, or intentionality (Heil, 1998, chapter 6; Ellis, 1995; Thomas, 1997a). Other recently elaborated approaches to cognition are built around closely related conceptions such as "image-schemata" (Lakoff and Johnson, 1999) and "perceptual symbols" (Barsalou, 1999).
.
We Make it Easy to Succeed
Successwaves, Intl.
Brain Based Accelerated Success Audios
 |