Speaker change in word recognition

Roessler, A. 1 , Gil López de Liaño, B. 2 & Sebastián Gallés, N. 1

1 Brain and Cognition Unit. Universitat Pompeu Fabra
2 Facultad de Psicología. Universidad Autónoma de Madrid

The auditory speech signal is highly variable and mechanisms to extract meaningful information are complex and not yet fully understood. Indexical information (speaker’s accent, emotional state) represents a source of information as well as variability. Recent literature has shown that fine-grained details of the auditory signal are retained in memory. In bilingual societies as in Barcelona, people are constantly exposed to two languages and importantly to slight (foreign-accented) mispronunciations. Furthermore, in bilingual contexts, speaker information correlates with language of communication. Therefore, comparing monolinguals with bilinguals might help to gather valuable insights into the role of variability in speech perception by investigating possible strategy-differences between mono-and bilinguals.

We conducted a series of auditory recognition memory experiments testing monolinguals (Madrid) and bilinguals (Barcelona). Participants were presented with a list of words of which half were repeated by either the same speaker or a different speaker. In a continuous design participants had to make an “old”-“new” judgement, while ignoring the speaker dimension. In experiments 1 and 2 the number of speakers was two. In experiments 3 and 4 the number of speakers was eight. The analyses of the reaction times showed significant repetition effects in all four experiments, but no traces of speaker effects. The analyses of the error rates (by means of d-prime scores) revealed an interesting pattern of results indicating different sensitivities to indexical cues in bilingual participants when compared to monolingual ones. These differences were more pronounced when multiple speakers were used (experiments 3 and 4).

The results indicate that bilingual recognition memory is less hindered by a change in speaker suggesting that indexical speaker information is not encoded as strongly as in monolinguals or alternatively, that bilinguals not be bothered by indexical information not relevant to the task.