[SaDPhIG] paper for Monday?

[colavin at ling.ucsd.edu]

Well, I like the idea but at some point, I a chance to get input about
these questions (this is just a draft of some ideas i have been struggling
with):


Hayes and Wilson use frequency information over natural classes to model
judgments over attested and unattested sequences.


1)In classic generative models, phonotactics are modeled as restrictions
only. Why? Is there psycholinguistic evidence for this? Or is it theory
internal?

2)Usually, attested sequences are assumed to be the reflection of  type
and token frequency of segments while unattested sequences are the
reflection of grammatical constraints that are not frequency based
(because speakers have no frequency information to distinguish unattested
sequences). Hayes and Wilson use frequency information over natural
classes to model judgments over attested and unattested sequences.  Why is
this not surprising?

3)Assuming different processes for attested and unattested sequences
implies that  there must be two different processes for harmony and
disharmony:
1.Disharmony (OCP in semitic languages) can naturally be modeled with
restrictions
2.reflecting restrictions on co-occurrence
3.Harmony, which is a pattern apparent in attested sequences, would have
to be modeled as frequencies over segments or, rather unintuituively, as
large set of restrictions on disharmony.

4)In fact, Hayes and Wilson do not really model attested sequences in a
satisfactory manner and my own research  shows that incorporating the
small restrictions that reflect over-representation sequences reduces the
correlation of the model with unattested sequences.


Rebecca

> May I suggest that we read this paper for Monday by Gabriela Caballero,
> a paper which will be the basis of her job talk on Friday?
>
> http://linguistics.berkeley.edu/~caballero/Site/Papers_files/morphology_paper_november08.pdf
>
> Sharon
>
> _______________________________________________
> SaDPhIG mailing list
> SaDPhIG at ling.ucsd.edu
> http://pidgin.ucsd.edu/mailman/listinfo/sadphig
>