Eye-tracking data on the immediate contribution of prenuclear accents and f0-interpolations to utterance interpretation in German

Previous research has shown that Germans produce contrastive topics with prenuclear accents exhibiting larger f0-excursions, later peaks, and longer durations than non-contrastive topics (Braun, 2006). Contrastive topic accents hence resemble L*+H, non-contrastive topic accents L+H*, but expert labelers experience difficulties in consistently labeling this distinction (ibid). Offline acceptability judgments in contrastive contexts showed that naive listeners preferred resynthesized stimuli with a larger f0-excursion on the prenuclear accent over ones with only later alignment or longer duration (ibid.). Why is it that this pragmatic distinction is signaled by a gradient scaling cue instead of a more categorical alignment cue? However, the scaling of the prenuclear peak is not a gradient cue if evaluated in relation to the scaling of the nuclear accent, though. Since judgments are performed at utterance end, relative peak scaling becomes meaningful. The current study is concerned with the online processing of prenuclear accents (L+H*,L*+H) and the following f0-interpolation that are not contaminated by the nuclear accent. We exploit the fact that contrastive topics semantically co-occur with contrastive foci (resulting in parallel constructions, A did X and B did Y). We monitored eye-movements of 33 participants to displays with four objects while they listened to minidialogues. 1a contains a contrastive topic accent L*+H, followed by a high plateau (hat pattern), 1b a noncontrastive topic accent (L+H*), followed by a dipping pattern, and 1c a contrastive focus on the subject (H*L-). Accented syllables are underlined. Acoustic analyses confirmed that trough and peak were aligned latest with respect to the stressed syllable for L*+H, significantly earlier for L+H* and earliest for H*L(> 20ms between conditions, p < 0.01). Visual displays all contained a contrast pair (e.g., tent–house, verified by means of a web experiment). In fillers, mostly one of the members of the contrast pair was mentioned. (1) Precontext: Ich habe gehört, Sabine wollte sich ein neues Zelt kaufen. (I have heard, Sabine wanted to buy a new tent.) Target: Und ich habe gehört (And I have heard)