What kind of informativity could drive phonetic duration modification: Focus structure or semantic likelihood?

Prosody is used to signal ‘informativity’, which has been separately approached in terms of information structure or information theory. As pointed out in [1], few studies combined both viewpoints in examining prosodic encoding. [1] used meaning-based contextual probability as an information-theoretical measure in an experiment and reported its influence on the fundamental frequency contour in different focus conditions in American English. A recent study of broadcast data in German also observed contributions of information status and trigram surprisal (i.e. structure-based) on syllable duration [2]. Inspired by [1], the current study revisited the role of information structure and information theory on prosodic encoding in German, by using a reading-aloud production experiment, and a meaning-based information-theoretic measure (i.e. likelihood of semantic association between two nouns) as in [1]. We hypothesized that (1) a focused component will exhibit longer duration than its non-focused counterpart, (2) a repeated word will have short duration and (3) a semantically less likely N1-N2 pairing will attenuate their durational differences and therefore attenuate the prominence relationship in each focus structure. Based on data from 10 participants, our preliminary findings provide support for (1), partial support for (2) but no support for (3).