Re: THEORY: storage v computation (was: RE: Language revival)
|From:||David G. Durand <david@...>|
|Date:||Monday, November 29, 1999, 18:06|
At 3:14 PM -0500 11/28/99, Nik Taylor wrote:
>And Rosta wrote:
>> But the
>> lesson is clear: in production, irregularity appears to confer
>> no benefits (if the experimental findings are correct).
>But, it confers no HARM either. This seems to suggest that both regular
>and irregular forms are stored, i.e., "cook/cooked" and "run/ran" are
>both stored without computation. If one were computed, and the other
>stored, one would expect one to be faster than the other.
The experiment as described cannot distinguish between the two cases (with
the result that was given). You would need to compare the creation of
rule-generated forms of non-words as well as actual words. (Presumably
derived forms of non-words will not be stored).
And, of course, you would have to measure, and correct for, any possible
delays caused by non-word prompts.
Like most experimental psych experiments, by the time you can get an
unambiguous result, the experiment will be artificial enough that the
result's application will be rather limited.
I tend to think that a mix of computation and memorization is involved,
varying with the frequencies of the words, but that's just an opinion.
David Durand firstname.lastname@example.org \ david@dynamicDiagrams.com
http://www.cs.bu.edu/students/grads/dgd/ \ Director of Development
Graduate Student no more! \ Dynamic Diagrams
MAPA: mapping for the WWW \__________________________