are we making since yet?
faber at HASKINS.YALE.EDU
Wed Apr 20 16:14:52 UTC 2005
David Bowie wrote:
> From: "Arnold M. Zwicky" <zwicky at CSLI.STANFORD.EDU>
>> meanwhile, ron butters wrote:
>>> The pin/pen "neutralization" is pretty widespread. In my
>>> experience, it is not a merger in the sense that one or the other
>>> phones is invariably selected..; rather, it is realized variably as
>>> anything between [I] and [E].
>> so long as the means of /I/ and /E/ are not significantly different,
>> this is a true merger; you have one phoneme with variation through
>> the [I]-to-[E] space, and this phoneme occurs in words that have /I/
>> or /E/ for other speakers. i believe this is the case for many
>> southern speakers. i also believe that there are southern speakers
>> with two phonemes, but with a lot of variation for each and with a
>> *lot* of overlap between their phonetic ranges.
> And that's why i don't agree with the definition of "merger" that you
> give above--you can have two phonemes, the means of which differ
> insignificantly, but that have significantly different distributions.
> From before my dissertation, i've used z/t-tests to test for merger,
> but i'm becoming more and more convinced there's got to be some *way*
> better statistical tests for it--any ideas? (My father-in-law, a
> mathematician, has floated some to my mind bizarre but maybe workable
> ideas past me.)
Marianna Di Paolo and I (with the advice of a statistician here at
Haskins) used discriminant function analysis for this.
Alice Faber faber at haskins.yale.edu
Haskins Laboratories tel: (203) 865-6163 x258
New Haven, CT 06511 USA fax (203) 865-8963
More information about the Ads-l