A lenição da oclusiva alveolar não vozeada por brasileiros aprendizes de Inglês como L2. ; The voiceless alveolar stop lenition by Brazilian learners of English as L2.

The aim of this dissertation is to investigate the lenition of the voiceless alveolar stop /t/ in intervocalic position within the word produced by Brazilian learners of English as a second language (L2) as in whatever [wʌtˈɛvɚ] [wʌɾˈɛvɚ] [wʌˈɛvɚ]. Consequently, due to the difference in allophonic distribution in the two languages, a relevant question is: how do Brazilian learners process the lenition of intervocalic /t/? Such a question was investigated based on two phonological models: the generative phonology (CHOMSKY; HALLE, 1968) and the usage phonology (BYBEE, 2016). The /t/ lenition was determined as the dependent variable, which presents the possible lenition path: (0) unreduced [t], (1) alveolar tap [ɾ], (2) glottal stop [ʔ] and (3) deletion [Ø]. In order to achieve the established goals 7 independent variables were determined, 4 of which were interlinguistic: adjacent vowels, grammatical category, lexical item and frequency of the lexical item, and 3 extralinguistic: proficiency level, type of exposure to input and individual. In order to reach the proposed objectives, data were collected through interviews and task application with 24 students from two language schools in the city of Montes Claros-MG. The data were initially evaluated in a categorical variable. Two levels of our dependent variable were determined: (0) for absent lenition and (1) for present lenition. The data were classified from acoustic analysis performed using PRAAT software (BOERSMA; WEENINK, 2010). The results obtained suggest that the general hypothesis was confirmed: the /t/ lenition was attested in the speech of Brazilian learners. However, unlike the proposed path, only stage 1, which corresponds to the production of the alveolar tap, was found in our data. Therefore, there was no occurrence of the glottal stop and no total deletion of /t/, as initially proposed. Usage-based models, which incorporate language use as well as phonetic gradient, allowed us to explain the lenition of intervocalic /t/ in the data more ...