To follow up on the posts about vowel fragments in the Iraqi woman’s speech…
It seems to me that her pronunciation of “street” is a nice illustration (all in one word!) of the difference between vowel epenthesis and vowel intrusion, a distinction that Nancy Hall argues for at length in her dissertation. My understanding of her work is that while the former arises for reasons having to do with syllable structure, the latter is motivated as a way to ensure perceptibility/recoverability of consonant clusters. Vowel intrusion is invisible to the phonology inasmuch as it fails to interact with processes that operate on higher-level prosodic structure.
Two examples of this invisibility come from Spanish. First, intrusive vowels that accompany taps are never counted in stress computation. In Spanish, main stress is confined without exception to a three-syllable window at the right edge of the morphological word (Harris 1995: 869). If the intrusive vowel in ártico [ˈaɾə.ti.ko] “Arctic” were to create a new syllable, then stress would fall outside the three-syllable window, yielding ungrammatical results: *[ˈa.ɾa.ti.ko]. Stress shift, *[a.ˈɾa.ti.ko], is an unattested repair strategy.
The second example is from the Spanish language game Jerigonza, often used by younger speakers as a secret speech code. In one version of the game, an epenthetic CV syllable is inserted to the right of every syllable boundary in a word. The consonant is from the set /p,t,k,tʃ/, and the vowel is a copy of the preceding syllable nucleus (Piñeros 1999). If the intrusive vowel in carta [kaɾə.ta] “letter” were syllabic, then CV-insertion would also target this nucleus. Jerigonza word formation yields [kaɾə.pa.ta.pa] instead of *[ka.pa.ɾa.pa.ta.pa], suggesting that the intrusive vowel is invisible. (In Section 3.3.1 of her dissertation, Nancy Hall also discusses Jerigonza data from Kekchi, a language with both epenthetic and intrusive vowels.)
Finally, it occurs to me that another way to get at the true status of the vocalic elements in the [st] clusters of the Iraqi woman’s pronunciation of English would be to check whether her L1 follows the predictions of Nancy Hall’s implicational hierarchy of vowel intrusion triggers (Section 1.4.4, example (45)):
Vowel intrusion triggers:
obstruents, if ever >
other approximants, nasals >
alveolar trill >
alveolar tap, voiced uvular fricative >
Among nasals: m > n
If the vowel in [st] is really intrusive, then wouldn’t the prediction be (on the grounds of harmonic completeness and modulo any gaps in the segmental inventory) that the speaker’s L1 should also have vowel intrusion with the segment classes that fall between obstruents and alveolar tap in the hierarchy?
References cited (in addition to N. Hall)
Harris, James. 1995. “Projection and Edge Marking in the Computation of Stress in Spanish”. The Handbook of Phonological Theory ed. by John Goldsmith, 867-887. Cambridge, Mass.: Blackwell.
Piñeros, Carlos-Eduardo. 1999. “Head Dependence in Jerigonza, a Spanish Language Game”. Advances in Hispanic Linguistics ed. by Javier Gutiérrez-Rexach & Fernando Martínez-Gil, 265-277. Somerville, Mass.: Cascadilla Press.