My favorite topic, vowel insertion in second language speech. Thanks for pointing this tidbit out Eric!
The first thing I’d like to say is that I’m very surprised that the Iraqi woman epenthesized instead of prothesizing. As far as I know, Broselow (1987) and others (Heidi Fleischhacker? I don’t have her paper here) have shown that while Egyptian Arabic speakers often epenthesize into English loanwords with initial /s/-clusters, Iraqi speakers usually prothesize. Examples (from Broselow, and see these authors for more on the epenthesis vs. prothesis distinction):
But that aside, I’d also like to point out that I think the quality of the svarabhakti vowel between the [t] and the flap is quite different from that between the [s] and the [t]. I’m surprised Eric transcribed it as [ɪ] (little uppercase I), because both impressionistically and visually, it seems like a different vowel to me–perhaps schwa? I’ll leave it at that for the moment, but in fact, if they are different, that’s important.
I’ve spent a long time thinking about Travis Bradley’s, Diamandis Gafos’s, and Nancy Hall’s work, and in comparison to my experimental work on vowel insertion in the production of non-native speakers, I’m starting to come to the tentative conclusion. In the case of copy vowels, or svarabhakti, the coordination of the two consonants is also bound up with the following vowel, so that the vowel gesture begins before the first consonant has been completed. In these cases, if the coordination pattern is such that there’s not close transition between the two consonants (that is, C1 is released), then the quality of the excrescent vowel is similar to the following full vowel because it essentially is that vowel.
In the case of my experimental work, and perhaps in the case of this Iraqi speaker, two things seem to be going on. First, there is a phonotactic restriction on [st] clusters, which is manifested as a prohibition on the close coordination of these consonants. Consequently, speakers do not produce these consonants with sufficient overlap, which leads to a period of open vocal tract between the consonants and the percept of a transitional vocoid (in fact, this kind of coordination seems to be the norm for languages like Sierra Popoluca, Piro, Moroccan Arabic, etc.) Since the transitional vocoid in this case is not a copy vowel, it is not necessarily going to have the quality of the following vowel. It’s going to have some properties that reflect the state of the open vocal tract at the moment that it is produced. (I see this in the acoustic characteristics and ultrasound images for my English speakers, who are asked to produce Slavic-like stimuli such as /zgomu/.)
Now, I guess I haven’t really answered Eric’s question about why the durations of the first transitional vocoid and the second copy vowel are different, except to say that I think that they arise from two different processes. I don’t know if Iraqi Arabic has a copy vowel before flaps as a regular phonological process, but if it does, then I would venture to guess that in the case of the copy vowel, they’re doing something practiced and productive in the grammar (and that has something to do with the aerodynamics of flaps–Travis?), whereas the transitional vocoid between [s] and [t] is the result of a different phonological process that occurs infrequently and in special cases, like second language acquisition or loanwords.
UPDATE (5:15pm, EST): I knew I left something out! Thanks Sharon, for reminding me. I originally meant to say at the end of the post that in all of my work, I never actually rule out true epenthesis as a possibility in non-native speech. While most of my English speakers seem to be producing a vowel that does not have the acoustic or articulatory characteristics of lexical schwa when repairing words like /zgomu/, there are others that definitely do seem to be epenthesizing. So, if epenthesis is a productive process repairing turning CCC sequences into CVCC in Iraqi Arabic, then yes, certainly, that would explain the length difference.
But I still don’t think that first vowel has the same quality as the second [i] in the Iraqi woman’s production of street…