Brains, Persons, and Society *** ABSTRACTS    Cervelli, Persone e Società ***ABSTRACTS

Giorgio Lando
Univ. of Eastern Piedmont – Vercelli

A Taxonomy for Objects of Vision: Wittgenstein and the cognitive approaches

 
Some antinaturalist philosophers - e.g. Putnam - mention Wittgenstein as a point of

reference in their criticisms of cognitive theories of vision. I maintain that these references to

W. are misleading. The issue grows clearer when we turn our attention away from W.'s

methodological remarks opposing sharply philosophical descriptions and scientific

explanations and we draw a detailed comparison between W.'s taxonomy of objects of vision

and analogous taxonomies typical of computational approaches to vision, inspired by the work

of Marr.

This comparison unveils many surprising analogies.

First of all, I reconstruct the Wittgensteinian taxonomy, analysing an heterogeneous

bunch of remarks and ignoring W.'s pretense not to produce taxonomies at all. Different kinds

of objects of vision can be distinguished on the following criteria, explicitly mentioned in W.'s

later works on the philosophy of psychology:

a) can the object of vision be "enlightened" (brought to attention) by willingness or by

circumstances?

b) is this object of vision seen per se or in virtue of its internal relations with other

objects of vision?

c) is this object of vision relevant in qualifying as adequate a non-interpretative

reproduction of what is seen?

d) can this object of vision be involved in the so-called “change of aspect”?

e) does the vision of this object require “imagination”?

f) are previous "concepts" required in order to see this object?

g) can this object of vision be perceived also through other kinds of perception (is

multimodal perception possible?).

Many of these distinctions can be usefully compared with analogous distinctions drawn

by cognitive theories of vision. In order to show these similarities, a main stream paper by

Pylyshyn ("Is Vision Continuous with Cognition? The Case for Cognitive Impenetrability of

Visual Perception?", Behavioural and Brain Sciences, 1999 (22), pp. 341-365) is chosen as a

term of comparison. The following analogies can be grasped:

1) what W. identifies as the “purely optical” component of vision through criteria b)

and c) can be compared with the purely bottom-up component of vision identified by Pylyshyn;

2) some kinds of “change of aspect” (in particular, the black-white cross example) are

described by W. as influenced by associations, but not by thoughts. In fact they are different

from the “purely optical” in virtue of the criterion b), but also from other kinds of objects

involved in the change of aspect, in virtue of other criteria (e.g., e)). This particular level can be

compared to what Pylyshyn describes as cognitively impenetrable, but nonetheless

computationally top-down;

3) Pylyshyn maintains that, although the organization of early vision is cognitively

impenetrable, cognition can allocate cognitive resources (attention) throughout the levels of

vision. Analogously, W. maintains that criterion b) does not actually discriminate among

objects of vision: any object of vision can be “enlightened”;

4) according to Pylyshyn, a distinctive property of the contents of early vision is that

they can not be object of multimodal perception; analogously, W. denies this possibility only

for the contents of the “purely optical” vision.

Given this detailed comparison, we suggest two strategies in order to make sense of a

prima facie contrast between methodology and actual analysis: W. draws a methodological

distinction between conceptual descriptions and scientific explications, but nowadays scientific

explanations resemble W's descriptions at a surprising degree.

The first strategy tries to give a new interpretation of W.'s methodological remarks.

Some of these remarks, while denying that science (physiology) can be considered a supercriterion

in conceptual analysis, does not exclude the possibility of some evidential connections

between philosophy and science. In particular, some remarks seems to suggest that a good

conceptual analysis can make a scientific hypothesis more likely than another (however, W.

does not explain why this should happen) and this could be the case when, for example, we

compare computational theories of vision with ecological theories à la Gibson.

The second strategy does not concern the Wittgensteinian exegesis, but the status of the

hypotheses-results of the cognitive sciences involved in the comparison. Only some of these

thesis have some kind of empirical confirmation. Some others could be safely classified just as

conceptual. Thus the analogy with W.'s descriptions would be less puzzling. However, the

evaluation of this strategy involves quite deep considerations concerning different levels of

analysis in cognitive sciences and can be merely mentioned here.