Séminaire 29/09/2023

Séminaire co-organisé par le laboratoire SCALab et le laboratoire STL

The role and sources of individual variability in the production and processing of speech prosody

Amalia Arvaniti

Radboud University, Netherlands

 This talk deals with the role that individual variability plays in the production and processing of prosody, intonation in particular, and the possible sources behind it. The studies I will present deal with the posited phonological contrast between H* and L+H* in English: somewhat informally, these are a high and a rising accent respectively and are said to indicate that the accented item is new in discourse (H*) or has a corrective or contrastive function (L+H*). Although the H* ~ L+H* contrast first posited in Pierrehumbert (1980) is generally accepted for English, the empirical evidence for it is slim, while some researchers (most notably Ladd 2008) dispute its existence altogether; a comparable contrast does not feature in British accounts of English intonation either. The present research sheds light on the reasons for the disagreement between analyses and the discrepancies between analyses and empirical evidence, by examining both production data from British English unscripted speech and perceptual data, which link the processing of the two accents to the participants’ levels of empathy, musicality, and autistic-like traits. By analysing the production data separately for phonetic realization and pragmatic function we show that Pierrehumbert’s original analysis holds only partially for British English: the accents used to indicate new information are falling rather than high, while the rising (L+H*) accents are used to both mark contrast and highlight unexpected information, though the extent to which speakers use L+H* for either purpose depends greatly on their individual style. This optionality does not apply to the same extent to corrective accents which typically take a L+H* shape. In terms of perception, the extent to which participants attended to the differences between the accents depended on individual traits: highly empathetic individuals were more sensitive to the function of the accents, while individuals with high musicality or more autistic-like traits were more sensitive to F0 shape.  As a result of these differences, individual speakers may learn different grammars such that for some the two accents do form a phonological contrast, while for others they do not.  Taken together these results point to the fallacy of trying to connect intonation categories with very narrow pragmatic functions and indicate that a better understanding of how phonological categories are formed can be gained by paying closer attention to individual differences and their sources.