>
> That's very interesting. Do you think they would claim that the
> results generalize beyond morphology, so that properties inherited by
> lexical items process faster than properties specified within the entry
> itself?
>
> One reason I'm curious is that it is often possible to find, over the
> grammar, various small regularities that are sufficiently underwhelming
> that one is unsure about whether it is worthwhile recognizing them. The
> claims you cite would suggest that there is a functional advantage to be
> had from maximizing regularity-recognition, however minor they are.
You only have to look at irregular past tenses in E, which pattern
into (fairly untidy) sub-regularities. Some of these are productive
(DOVE), so there *must* be a generalisation (at least for anybody
for whom DOVE is possible).
>
> --And.
>
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|