Multidimensional interpretation of rising and falling tunes for requests and offers

Résumé En

Intonational contours convey multidimensional meaning. While it has been long assumed that the semantics of tunes are driven compositionally by integrating the meaning of pitch accents and phrasal/boundary tones [1], perception of meta-linguistic information such as valence of speaker emotion is known to affect listeners' judgment of speech act (e.g., statement vs. question: [2]). Social factors also affect tune choices: the speaker may choose different tunes (e.g., for requests and offers) according to their familiarity with the listener [3], and the perlocutionary meaning is a function of both sentence type and tune [4]. The present study investigates the interpretations of rising and falling tunes of specific kinds of requests and offers in American English (e.g., Can [you/I] bring [me/you] some water?). In particular, it tests how the perceived speaker mood, degree of sincerity and projected authority interact with one another (see example in (1)). The choice of the three dimensions is in line with the literature (e.g., [5]). Since the tested requests and offers belong to polar (yes/no) questions, the intonational deviation from the default rising tune [1] was predicted to evoke negative assessment of speaker mood and perception of relatively lower authority of the listener (or higher authority of the speaker). The effect of falling tune on the perceived degree of sincerity may differ between requests and offers, because an offer (Can I bring you some water?) with a falling tune may sarcastically convey the lack of intention to make an offer, whereas a request (Can you bring me some water?) is less likely to be uttered without intending to make a request, even with a falling tune. A total of 96 request-offer pairs were recorded with both rising [H* L-H%] and falling [L* L-L%] tunes by two female native speakers of AE. 240 participants were randomly assigned to the 24 presentation lists for an online survey until each list reached 10 participants. Each participant rated a total of 96 items, which were presented as a sequence of 6 blocks that clustered 16 items by utterance type (request/offer) and question type (mood/authority/sincerity). On each trial, participants pressed the play button to listen to the sentence, saw the question, and rated the sentence using the slide bar underneath it (Fig. 1). Each participant provided a rating on only one of the three questions for a given version of an item (see (1) for an example set of questions that were cycled across lists). Responses from 237 participants (observation N=22677) showed the predicted main effect of falling tune, which led to overall negative assessments of mood and sincerity (both p