Accommodation of L2 speech in a repetition task
Résumé
This paper investigates phonetic accommodation between non-native and native speakers in a repeti
-
tion task. The data is a fragment from the ANGLISH corpus (Tortel 2008). 40 French speakers (10 male
intermediate and 10 male advanced learners, 10 female intermediate and 10 advanced learners) were
asked to repeat a set of 20 sentences produced by British native speakers, resulting in 815 recordings
lasting on average 4 seconds. Subjects could read the text corresponding to the sound file that was played.
Stimuli are short sentences from the EUROM Corpus (Chan et al. 1995) which increases the comparabi
-
lity of the data with existing speech corpora for natives.
Previous findings for accommodation in spontaneous speech indicate that female informants tend to
converge more than male informants (Babel 2009) and advanced learners to converge more than inter
-
mediate learners for ‘greater L2 usage and proficiency are associated [...] with increased L2 production
experience’ (Best & Tyler 2007). We expect female speakers to converge more whatever their level but
level to play a role in the type of accommodation observed. Convergence in vowel duration in order
to sound more native-like (Zając 2013) and accommodation for low vowels especially within the F1
dimension (Babel 2009) are more likely to be observed. F1-F2 Euclidean distance will be compared for
levels and gender.
We will analyse convergence with the realisations of read speech recorded before this repetition task
that we have used as a baseline. The idea is to determine whether phonetic accommodation occurs straight
after auditory exposure to the model talker. Statistical and acoustic analyses will be based on tools such
as Praat (Boersma & Weenink 2015), Rstudio (RStudio Team 2015) or WinPitchPro (Martin 2005).
In addition, low frequency words tend to be more subject to imitation than high frequency words (Gol
-
dinger 1998). We will look at the CELEX database and extract the least frequent words of the corpus to
determine whether this assumption is proved to be validated in the case of imitation of L2 speech