Visual Imagery and Consciousness
Nigel J.T. Thomas
Page 5
http://www.imagery-imagination.com/viac.htm
Quasi-pictorialists seem to have two comebacks to the homunculus objection. One is to point out (as Kosslyn has often done) that a computer simulation of quasi-pictorial theory has been implemented, and that there can be no homunculus (at least of any objectionable sort) within a computer. Unfortunately, however, it turns out that the computer program in question makes no attempt whatsoever to model the conscious nature of imagery. Like most of Kosslyn's work, it is concerned to explore how certain sorts of spatial thinking might be achieved through imagery, via such operations as scanning linearly across a quasi picture, or rotating it into a new orientation. The program performs such operations on internal data structures that are supposed to model, at a functional level, the brain quasi pictures posited by the theory, and it displays its results in the form of actual pictures that are rotated or shifted across the computer's display screen in relevant ways. No attempt is made to simulate the hypothesized "mind's eye function," and consciousness, either real or simulated, does not enter into the matter at all unless and until some human operator looks at the screen. It is true that there is no homunculus in the computer, but, inasmuch as the system models consciousness at all, it models it by co-opting a full sized conscious human being to play the homunculus role.
Not coincidentally, the program also omits to model the intentionality of imagery. The fact that the two-dimensional patterns it produces and manipulates may look like pictorial representations to human onlookers (and were designed to look that way by the programmers) is quite irrelevant to the program's functioning. For the computer, they represent nothing.
The second comeback is to suggest that perhaps cognitive science can render the homunculus and its "mind's eye" innocuous by showing that it can be reduced to a set of computational or neural processes. Quasi-pictorial theory was originally developed in the context of the "information processing" paradigm which dominated perceptual theory at the time (the 1970s), and which still deeply influences the way many cognitive scientists think about perception. In essence, "information processing" theory regards vision as a one way flow of visual information, in through the eyes and then through a series of processing stages in the brain until it is eventually transformed into a representation, or a set of representations, suitable for guiding behavior. (This is an oversimplification, but not, in this context, a misleading one. Developed information processing theories generally call for a significant degree of top-down modulation of the bottom-up flow of information from the sense organs. Nevertheless, the bottom-up flow dominates and drives perception.) A common assumption (usually implicit) is that this final set of representations, this ultimate product of the visual information processing system, is the immediate cause of conscious visual experience.
(Some information processing theorists may prefer to think of visual consciousness not as something attaching to representations, but, rather, as something arising from the workings of the processing system as a whole, perhaps even including the sense organs and the muscles that support behavioral response. However, this position is not open to the pictorial imagery theorist, who necessarily holds that we have conscious experience of inner representations.)
According to quasi-pictorial imagery theory, one of the earlier stages of this visual information processing is the creation of quasi-pictorial representations in the brain. These may be derived from actual present visual input from the eyes (when we are actually seeing), or they may be created from stored data in memory (when we are remembering or imagining), but in either case they must be passed through several more stages of processing in order to extract useful information from them. It is these further stages that constitute the "mind's eye function." Thus, it can be argued, the mind's eye has a principled and independently motivated role within the broader theory of vision. As scientists are actively investigating and seem to be making progress in understanding the computational and neural mechanisms of visual information processing, this "mind's eye" function is not so much a non-explanatory homunculus as a promissory note drawn against the expected success of an ongoing research program.
But even if information processing theory does provide the right framework for understanding visual perception (and not everyone thinks it does), the output of all the processing is just more representations, instantiated as patterns of neural activity. Once again, it seems that we must say either that these representations are conscious in and of themselves, which brings us smack up against the "hard problem" once again, or else we need another homunculus to read them and be conscious of what they represent. Perhaps this homunculus, also, might be reduced to a further series of stages of information processing, but this would only lead us to the same place yet again, and so on in unending regress. We still do not begin to understand how quasi pictures could be, or could produce, conscious imaginal experiences.
Description theory
Description theory, whose most important advocate has been Zenon Pylyshyn, began as an attempt to understand how the phenomena of imagery could be fitted into a computational theory of the mind. Pylyshyn holds that the way computers (and, by extension, brains) represent information is necessarily more like language than like pictures. The syntax, and, indeed, the vocabulary, of the hypothetical internal brain language (sometimes called "mentalese") might be very different from that of any language that anyone speaks, but it is language-like in that it ultimately consists of symbolic tokens that represent things in the world in much the same way as the words of a language do, not through resemblance but through some essentially arbitrary correspondence relation.
In the case of "natural" languages, like English or Chinese, this correspondence (which determines, for instance, that the word "dog" refers to dogs) is established by social convention. In the case of humanly written computer programs the external reference of symbolic tokens (when they have one) is set by the programmer. Matters are much less clear when we come to hypothetical symbolic tokens in the brain, and there is great controversy over how, or whether, the "symbol grounding problem" (the problem of understanding how the "words" of mentalese might be able to refer to things outside the brain) can be solved. Nevertheless, much work in cognitive science proceeds on the assumption that a solution is possible.
We Make it Easy to Succeed
Successwaves, Intl.
Brain Based Accelerated Success Audios
 |