Acoustic-prosodic Disambiguation of Direct and Indirect Speech Acts

The same surface form used for a speech act meant to be interpreted literally may be used for a speech act that takes on additional indirect meaning. We investigated acoustic and prosodic differences between the realizations of such direct versus indirect speech acts. We conducted a production experiment with seven native General American English speakers, each engaging in fourteen pairs of dialogues designed to elicit realizations of direct and indirect readings of fourteen surface yes/no questions. Our analyses of the acoustic-prosodic features of these utterances show that 1.) utterances realized with a low boundary tone are substantially more likely to have an indirect reading, 2.) the main pitch difference between direct and indirect speech acts is realized in the f0 of the final high peak, and 3.) impoliteness in dialogue context contributes to greater final f0 rise in a particular subclass of utterances.