Re: Images Vs. Symbols

From: HARNAD Stevan (harnad@cogsci.soton.ac.uk)
Date: Wed Jun 05 1996 - 18:51:21 BST


> From: "Porter, Julie" <JAP295@psy.soton.ac.uk>
> Date: Fri, 24 May 1996 09:08:13 GMT
>
> "A picture is worth a thousand words" refers to the imagery debate;
> therefore, whether mental imagery is best explained by depictive (ie,
> similar to actual pictures or images, they resmble the spatial
> features of an object physically) or propositional (ie, symbol
> strings, which similar to words, share the fact that they are only
> arbitrarily related to an object) representations.

The words' shapes do not resemble what the words refer to.

> Pylyshyn argued
> that mental imagery cannot be pictorial since he believed that
> images, unlike propositions, require a homonculus ("a little man in
> the head") to look at them, and to interpret and understand them.
> Hence, to Pylyshyn a "picture" is worth nothing on its own, since it
> objectively means nothing without a propostionally-based
> interpretation. Instead he saw images as "epiphenomenal"; an
> unimportant by-product of some other underlying process, images
> themselves are of no consequence during the actual act of processing
> information.

Good.

> Basically, propositionalists, perhaps rightfully,
> advocated the possibility of using symbols in any task which Kosslyn
> believed necessitated the "priveleged properties" of depictive
> representations; with the additional advantage that propositions can
> "stand alone", without the necessity of an internal viewer, since as
> in a machine, the computation involved in symbol manipulation does
> not require a "mind's eye". Therefore, it is possible to argue that
> a propositional representation can offer as much information as would
> be avaliable in a picture.

The last statement is true, but it does not follow from the fact that
images may need homunculi and propositions don't. It follows from the
power of computation (and language) to compute (and express) ANY
proposition whatsoever. Hence anything that a picture can convey
directly, by resemblance, a long enough string of propositions could
convey to as close an approximation as one wishes.

> However, as an end-product, propositions
> are cumbersome. Pictures can encompass a vast amount of information
> in a more immediately avaliable and accessible form, whereas to
> represent certain things symbollically would require a vast amount of
> additional symbolically-based information. Eg, when we perform a
> task such as identifying a face, we do so by directly picturing the
> face, rather than labouriously working through a list of descriptive
> features. Clearly the latter would prove innappropriate.

Correct.

> Therefore,
> although experiments such as the "mental scanning" studies did reveal
> that propositions, in the form of "mental lists", were equally
> capable of supposedly visually-based tasks; as a usable
> end-product(regardless of trhe processes that may create it), an
> image-based representation is preferable for such a task. Therefore,
> in this respect a picture may prove worth a thousand words, although
> it appears that the worth of both types of representation (which each
> appear equally probable), may be task dependent.

Very good. To push it over into an A, you need to relate it to the
general issue of analog processing vs. the power of symbol processing.



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:45 GMT