Abstract EN:
Intonational units possess both phonetic and meaning correlates. In bidialectal speakers, language use and socio-contextual cues have been shown to influence the mapping between intonational forms and meaning. Additionally, code-mixing frequency in daily life can alter the cognitive costs of language inhibition and selection. However, existing models of Intonational Phonology fail to account for these effects in bilingual intonation. Addressing this gap, the present work tested a hybrid exemplar model of intonational representation.
German–Italian bilinguals living in South Tyrol (ST) were taken as a case study due to their considerable variability along various sociolinguistic dimensions, among which are language dominance and code-mixing frequency. Previous studies have identified a rising L+H* pitch accent in both ST Italian and German, though this pitch accent is associated with Narrow Focus (NF) statements exclusively in ST German. In contrast, ST Italian displays a preference for a falling H+L* in the same pragmatic context. Thirty-six and forty-seven ST speakers took part, respectively, in a production and in a perception experiment on ST Italian intonation. The two experiments were conceived as a single experimental cycle and tested the influence of language dominance, code-mixing frequency, and socio-contextual cues on the association between L+H* and three information structures, including the aforementioned NF.
Through an interactive video game, a semi-read speech was elicited in a laboratory setting. This speech included Italian statements produced in three conversational contexts. The contexts, in turn, elicited the information structures under investigation. Stimuli had identical structures in the two experiments, i.e., Ha messo il NP1 sul NP2 (“pro_3SG put the NP1 on the NP2”). The two NPs consisted of single-word denominations of objects in the video game and could alternatively bear the nuclear pitch accent in the production experiment. In the perception experiment, the goodness-of-fit of auditory stimuli in the three conversational contexts had to be rated. Stimuli featured either L+H* or H+L* in sentence-final position (NP2). The experiment was completely web-based, and all written elements of the experimental setting were manipulated between participants. These elements were presented either in Italian or in German, serving as socio-contextual primes for the respective language systems (Social Prime). Conversational contexts and stimuli, in contrast, were always presented in Italian. Language dominance and code-mixing frequency of our participants were assessed through an expanded version of the Bilingual Language Profile – a standardized sociolinguistic questionnaire used in bilingual research – and then discretized for our purposes. We expected that German dominance, high code-mixing frequency, and German Social Prime would favor stronger cross-linguistic influence from German and determine a higher likelihood of L+H* in the NF context, compared to Italian dominance, low code-mixing frequency, and Italian Social Prime.
Regardless of socio-contextual cues, the results show that language dominance has the predicted impact on intonational patterns, pointing to cross-linguistic interference. Whereas the role of code-mixing frequency is less clear, the German Prime also favors the association between L+H* and NF, indicating that the influence of socio-contextual cues is partly independent of the target language. Crucially, the effect of Social Prime is modulated by the two usage-related variables (code-mixing and language dominance), and production and perception data closely mirror each other. Hence, these findings lend support to hybrid exemplar models and contribute to a deeper understanding of Intonational Phonology in bilingual speakers.