Joe:
> And Rosta wrote:
> >
> > But suppose instead that R were a generalization arrived at by generalizing
> > from particular instances (say, A, B, and C). There would be a stage at
> > which the learner has conservatively learnt A,B,C before making the
> > inductive leap to generalization R.
>
> This may or may not be true:
>
>    recall the discussion of connectionism [...]. One feature of
>    those models was the apparent collapse of the data/algorithm
>    distinction itself. The connection weights, in such models, act
>    as both knowledge store and knowledge-manipulation algorithm.
>    If real neural computation is indeed anything like connectionist
>    computation, the standard notion of an algorithm as a recipe
>    for acting on an independent data set also seems strictly
>    inapplicable. (Clark, Andy. 2001. Mindware: An Introduction
>    to the Philosophy of Cognitive Science. Oxford:
>    Oxford University Press, p. 97)

But in this case, how does this story capture the distinction between
knowing A,B,C without generalizing their pattern to D,E,F, and knowing
R, which covers A,B,C,D,E,F?

On a different point, it's views such as the one you quote that
make me think we have to countenance algorithms existing extramentally
in the abstract system but not in the mind.

> > [Let me irrelevantly interject Edwin's analysis of "hamburger": "handburger"
> > -- you eat it holding it in your hand, and ordinary assimilation
> > (e.g. handbag > hambag) yields the normal pronunciation.]
>
> This is a great example. Tell Edwin to keep up the great work!

I'll mention a similar one. Although I know that the phonological form
of _forecast_ is /fO:kA:st/ in canonical English (with a Southern English
or S. Hemisphere accent), I've always had a sense that it is /fO:gkA:st/,
and I was for many years mystified by this. But a couple of years it
occurred to me that as I child I might have analysed it as _forwardcast_
(i.e. an act of looking forward in time), which would give /fO:dkA:st/,
which through regular assimilation gives /fO:gkA:st/. The relative
[o]-like closeness of the realization of /O:/ that I was exposed to as
a child would tend to mask the acoustic difference between /fO:gkA:st/
and /fO:kA:st/.

--And.