Connectionism

[gthomson]

At 08:59 -0600 22/03/01, Sydney Lamb wrote:
>On Thu, 22 Mar 2001, dan everett wrote:
>
>>      The basic idea is that phonological representations require the use of
>>      variables and cannot be accounted for by different strengths of
>>      connections or statistical generalizations. Evidence for variables is
>>      problematic for connectionism.
>>      ...
>
>Variables are problematic only for a simple-minded version of
>connectionism such as the well-known version of Rumelhart &
>McClellan and their colleagues, with its limitation to three
>layers, etc. But that simple-minded version has to be rejected
>anyway as neurologically implausible.
>
>  - Syd Lamb

I don't know that it is clear what sense of "variables" is intended here.
Greg Thomson