The open post-tonic high front vowel in trochaic bisyllables: a study of native speaker and english learner corpora

Abstract

This is a phonetic study across three oral corpora: one with recordings of naturally spoken American English (the SBCSAE corpus), another with interviews with advanced Brazilian learners of English (the LINDSEI-BR corpus), and a third one, composed by recordings of spontaneous speech in Brazilian Portuguese (the C-ORAL-BRASIL corpus). The main purpose was to investigate the realization and duration and second formant patterns of the phonologically lax high front vowel in free final position, [:]#, in English trochaic bisyllables, i.e., happy and study. The analysis of similar Portuguese [:]#-WORDS (such as gente and sabe), was also necessary for comparative purposes, since the project for this thesis originated from the impressionistic contention that native speakers of English consistently produce [:]#, even in their most casual speech, while, on the other hand, the English spoken by Brazilians and the Portuguese normally spoken in Brazil both often seem to exhibit [:]#- WORDS with an overly short or simply unrealized [:]#. The results show that native speakers of English do produce [:]# consistently but that the vowel duration varies considerably (it varied between 33 ms and 124 ms in SBCSAE samples). In LINDSEI-BR the variation was even greater. Brazilian learners of English did not realize the [:]# in 10 words out of 32 analyzed. In all those cases of unrealized [:]#, there was phonetic similarity between the intervocalic consonant in the [:]#-WORD and the sound immediately following it (e.g. [p] and [b] in happ(y) because). The two sounds around [:]# may represent the same phoneme, e.g., (/r/ [:]# /r/), ver(y) rich; (/l/ [:]# /l/), reall(y) like. On the other hand, following pauses seem to have greatly favored the full realization of [:]# in the LINDSEI-BR, e.g., She was very... The C-ORAL-BRASIL data indicate that, for Brazilian Portuguese, the expectations about [:]# should be readjusted for a different standard: The common behavior found in the corpus was the non-realization of the free vowel, e.g., the word gente was produced as [et] more frequently than as [e.t:]. A further reduction was observed in some other samples analyzed, apparently caused by a considerable distance between two stressed syllables. In a segmentation of the phrase Sabe que que é? the canonical sequence of syllables Sa-be-queque-é? was reduced to Sa-que-q'jé?