Talker adjustment to perceived communication errors

Abstract

During conversation, talkers may adapt their speech in a variety of ways. One form of adaptation is clear speech, in which a talker hyperarticulates when faced with specific communication challenges. In this study, we investigated how phonetic adaptation is affected by the nature of apparent miscommunication. To do so, we utilized a common feature of American English dialects, in which intervocalic /t/ is realized as a voiced flap rather than an unvoiced, fully articulated stop before unstressed syllables, causing words like petal and pedal to sound similar. Flapped /t/ may become more distinct compared to /d/ during clear speech. Using a faux-computer training paradigm, we provided artificial feedback on subjects’ speech. When subjects said a word (e.g., petal) the computer would either interpret the word correctly (i.e., it guesses petal) or incorrectly (it guesses pedal, kettle or ???). We measured talkers’ produced word length, initial-syllable vowel length, and medial consonant stop closure or flap consonant duration in both their first and second utterances of each trial. We found that participants produced longer words, vowels, and stop closure durations in response to feedback that contained an error compared to correct feedback. Participants also changed their flapped productions to stops most often and lengthened their stop closures to a greater degree following voicing errors compared to the other error conditions. Our findings indicate that interlocutors’ productions may change along different acoustic dimensions depending on the kind of miscommunication they experience. Broadly, our findings indicate a need to consider how phonetic adaptation in the form of clear speech is affected by the nature of miscommunication during conversational interaction.