First the issue of binary branching onsets and codas...
I just think that the phonological arguments are persuasive on this, but I
am willing to set them aside in the interests of simplicity if it is
simpler to have a more phonemic-looking thing, eg. three onset nodes for
/spr-/ etc. Sarah? any opinion?
My main concern here is that the information we all want should be clearly
accessible or visible from the structures we've got. I'll give this a bit
more thought next week.
But yes, it may well be that /s/ in clusters with plosives is temporally
unlike other onset /s/'s. And /sl/ is different again, because it's
primarily in contrast with /Sr/ (both 'clear' vs. both 'dark' consonants),
and also with /f/.
(In any case /-Nkt, -lpt/ codas are irregular under this view... in fact
they capture Latin morphology, where /-t/ marks the past tense stem of
verbs.)
The other issue is the monomorphemic one.
Actually I think I should have been clearer -- it's not morphology per se
that's problematic here, but *inflexional morphology*. I agree with Jill
here: there may be no obvious phonetic difference in RP between "tacks"
and "tax" type things, but we ought to leave ourselves the possibility of
modelling this if need be. I need to think about this more as well,
because it leads to parsing problems to have an Appendix for inflexions.
(or it did last time we tried!!)
By the way:
I don't care about the names we give to features and values --- as long as
they're mnemonic and we don't confuse them.
About inheritance:
to say it trickles up and down is just to use a metaphor. A truer metaphor
here would be to say that the information is available at different places
in the structure. Of course, when you're parsing from phonemes, then
trickling up and down is a helpful image: but the representations are
static, hence the preference for a static metaphor. For example:
Rhyme <voi +>
/ \
Nucleus <voi +> Coda <voi +>
what we mean here is that <voi+> is available at all threeplaces in
structure. It might indeed "get there" from the coda in the parsing, but
in the representation, the information is available in all three places.
If we index it with eg. [1] (like I did in my document), then we are
saying that the information is referring to the same thing.
Atually the "link" metaphor from HTML and the WWW is quite a good one:
click on many buttons in several documents and if the link is the same,
you'll get to the same page/information.
Hope this helps, if you can be bothered to read this far!
Richard
Richard Ogden
rao1@york.ac.uk
http://www.york.ac.uk/~rao1/