Production and perception of Spanish voiced stops and approximants by L2 learners

Abstract

This study investigates second language (L2) Spanish learners’ acquisition of word-initial voiced stops [b d g] and intervocalic voiced approximants [? ð ?] at different points in a university curriculum to explore the potential relationship between their production and perception. Twenty-three L1 English-speaking L2 learners of Spanish enrolled in first-, third-, and fourth-year courses, respectively, at a university in the United States and four L1 Spanish speakers participated in the study. Participants’ production was assessed with a word list reading comprised of nonce words. For the perception task, participants listened twice to a recording of an L1 Spanish speaker reading the same words used in the production task. Each word contained a voiced stop in word-initial position and its corresponding approximant allophone in intervocalic position. Prior to completing the perception task, participants were informed that each word contained two tokens of the same consonant (e.g., two tokens of [b]) and they were to indicate on an answer sheet whether the two instances of the repeated consonant in each word were pronounced the same or differently. Acoustic analysis in Praat revealed that overall, learners produced word-initial [b d g] with significantly less prevoicing than L1 Spanish speakers and they infrequently produced target approximants as such. Acoustic analysis also indicated that first-, third-, and fourth-year learners achieved less spirantization of target intervocalic voiced approximants, when produced as such rather than as voiced stops, than L1 Spanish speakers. In addition, first-, third-, and fourth-year learners perceived a difference between target voiced stops and approximants 49%, 43%, and 58% of the time, respectively, following a U-shaped learning curve. Finally, Pearson’s correlation analyses suggest that the relationship between learners’ production and perception of word-initial voiced stops and intervocalic voiced approximants depends on their instruction level and the target sound/sound pair.