Variation in prosodic planning among individuals and across languages

Résumé En

Previous research (Swets et al., 2007) found that working memory (WM) was associated with the manner in which silent readers in multiple languages package linguistic material together. Specifically, those with high WM were more likely to create larger linguistic packages than those with low WM, which in turn influenced the manner in which they interpreted syntactic ambiguities. One possibility that gained support in subsequent research on language production (Petrone et al., 2011) is that this packaging associated with WM is prosodic in nature. In the current study, we present evidence that the size of prosodic increments during language production, as measured by the occurrence of pauses, is associated with WM and speed of processing. This study adds to the increasing body of literature that seeks to better understand mechanisms of language production by investigating variation in planning tendencies among individuals. In addition, we examined whether pausing tendencies vary across languages. In our study, 31 German speakers and 32 French speakers described images presented on a computer monitor while their utterances were recorded. The images contained three objects apiece, and a speaker's task was to describe the images in such a way that a person listening to the utterance could follow the instruction to move the described objects (e.g., " The box moves below the train and the mouse moves above the train. ") To manipulate sentence complexity (and cognitive load), half of the descriptions required speakers to distinguish between similar looking objects (e.g., a cat with four legs vs. a cat with three legs). In addition to number of pauses, we also measured initiation time, and separately measured WM via a Reading Span task and processing speed via a Letter Comparison task. Analyses using linear mixed effects models revealed several key findings. First, we found significant cross-linguistic differences in initiation time and number of pauses: French speakers began their speech more quickly, but paused more often than German speakers, suggesting a greater degree of " incremental " planning. Despite these cross-linguistic differences, we nevertheless found evidence across languages that both WM and processing speed explain unique variance in our measure of pause frequency, and do so in the same direction: As WM and processing speed increase, the number of pauses speakers produce per utterance decreases. There is also evidence that both effects are stronger during the high-load, complex sentences. On the other hand, neither WM nor processing speed predicted variability in initiation time. These results lend support to the hypothesis that individual differences in WM and processing speed lead to differences in planning processes, such that higher WM and processing speed support the planning of larger prosodic " chunks " .