Abstract
People are more accurate in voice identification and emotion recognition in their native language than in other languages, a phenomenon known as the language familiarity effect (LFE). Previous work on cross-cultural inferences of emotional prosody has left it difficult to determine whether these native-language advantages arise from a true enhancement of the auditory capacity to extract socially relevant cues in familiar speech signals or, more simply, from cultural differences in how these emotions are expressed. In order to rule out such production differences, this work employed algorithmic voice transformations to create pairs of stimuli in the French and Japanese language which differed by exactly the same amount of prosodic expression. Even though the cues were strictly identical in both languages, they were better recognized when participants processed them in their native language. This advantage persisted in three types of stimulus degradation (jabberwocky, shuffled and reversed sentences). These results provide univocal evidence that production differences are not the sole drivers of LFEs in cross-cultural emotion perception, and suggest that it is the listeners' lack of familiarity with the individual speech sounds of the other language, and not e.g. with their syntax or semantics, which impairs their processing of higher-level emotional cues.
