ISCA Workshop on Plasticity in Speech Perception (PSP2005)
Senate House, London, UK
In this article, we try to quantify the contribution of prosody (timing and melody) to the perception of a foreign accent, by using a prosodic transplantation paradigm. For the reported experiment, a dozen sentences which are spoken in almost the same way in Italian and in Spanish were recorded: e.g. ha visto la casa del presidente americano ('(s)he/you saw the American president's house'). They were read by 3 Spanish monolinguals, 3 Italian monolinguals and 3 Spanish/Italian bilinguals (once in the Spanish way, once in the Italian way). Prosody crossings were performed after an automatic segmentation of the sentences into phonemes: in order to copy the prosodic parameters of an utterance onto another one, a script was written for the PRAAT software, which enables speech handling and re-synthesis (extraction and then transplantation, phoneme by phoneme, of duration and then pitch). In addition to crosslanguage combinations, we crossed each original voice with the prosody of the same language of another speaker, to make sure that the impression of foreign accent comes from signal manipulation. Perceptual tests were run on 100 resulting stimuli, with three samples of individuals: for native listeners (20 Spanish and 20 Italian), the task consisted of judging whether the utterance was 'Spanish', 'Spanish with an Italian accent' (i.e. Spanish spoken by an Italian), 'Italian with a Spanish accent' (i.e. Italian spoken by a Spaniard) or 'Italian'; for 20 French listeners, the task consisted of telling if the mother tongue of the speakers they heard was 'very probably Spanish', 'probably Spanish', 'probably Italian' or 'very probably Italian'. It turned out that the original stimuli were well identified by the three groups of listeners, even though Italian subjects were inclined to answer 'Italian with a Spanish accent' as soon as the voice or prosody was Spanish Đ and symmetrically for Spanish subjects. The artefacts introduced by the acoustic manipulation of the stimuli did not affect the speakers' origin identification. For cross-language crossings, results suggest that prosody outweighs segmental cues in the perception of Spanish/Italian foreign accent. They were analysed statistically, and the corpus was investigated acoustically. In particular, the duration ratio between stressed and unstressed syllables is 1.4 for the Italian language, and only 1.1 for the Spanish language. In conclusion, all other things being equal and at least for bilinguals, prosody and particularly rhythm plays a major role in the perception of Spanish/Italian foreign accent.
Bibliographic reference. Vieru-Dimulescu, Bianca / Boula de MareŘil, Philippe (2005): "Contribution of prosody to the perception of a foreign accent: a study based on Spanish/Italian modified speech", In PSP2005, 66-68.