Re[2]: ka and k^a [was: [tied] *kW- "?"]

At 6:57:04 PM on Friday, September 23, 2005, Patrick Ryan
wrote:

> From: "Miguel Carrasquer" <mcv@...>

>> (From: "Perception and representation of regular
>> variation: The case of final /t/, Meghan Sumner, Arthur
>> G. Samuel,
>> http://www.cs.sunysb.edu/~adaptation/sdarticle.pdf):

>> >For example, word-final /t/ can be produced
>> >in citation form with a fully released coronal stop,
>> >as a glottalized coronal stop that is coarticulated with
>> >both coronal and glottal place with no audible release,
>> >and as a glottal stop (e.g., the medial sound in the English
>> >word uh-oh). While it is generally accepted by linguists
>> >that these variants are common in American
>> >English, there is little research supporting this claim.

>> So, it appears, that Ladefoged's prescription are
>> basically unsupported by data. Just as I suspected.

I rather suspect that there's little published research
simply because the question is fairly uninteresting: anyone
who's paid any attention at all is aware at least of the
pre-glottalized allophone, and probably of the [?] allophone
as well. It would be nice to have more data, but the
qualitative picture is pretty clear.

>> >There is evidence that these three variants are regularly
>> >occurring in the Long Island dialect of American English
>> >(Huffman, personal communication, September 7,
>> >2004), the population examined in our study. Huffman
>> >(1998) found that all three variants occur regularly in
>> >the Long Island dialect of American English. Furthermore,
>> >at the end of an intonation phrase (i.e., word-final
>> >position before a pause), 70% of final-/t/ words are produced
>> >as glottalized stops, with both coronal and glottal
>> >articulation. The remaining 30% of final-/t/ words are
>> >split between the canonical [t] and the glottal stop.

> What I get from this is a contradiction of the assertions
> made by Ladefoged who makes a following consonant
> necessary to fulfill the conditions for [t] into [?].
> Here, a pause is necessary.

No, it isn't. The information given here merely happens to
be restricted to that context. (What does seem from other
sources very likely to be true, however, is that the highest
incidence of [?] is indeed found in precisely that context.)

> And even then, in only _15%_ of the instances does [t]
> become [?] -

You mean that /t/ is realized as [?]; there was no [t] in
the first place in these utterances.

> in the _Long Island_ area.

15% is rather a lot when set against the claim that it
happens in no U.S. variety. And at least one of the
references that I cited a few hours ago adds some
independent data. The Dautricourt abstract notes that:

In a dataset consisting of over 400 tokens, comprising all
/t,d#y/ word pairs in hour-long interviews with 16
speakers, four variants predominated in the following
decreasing order of frequency: glottal stop, palatal
affricates, alveolar stops, and (alveolar stop) deletion.

That is, in this specific context the glottal stop was the
single most frequent output.

Brian