Acoustic cues for the perception of the information status of words in speech

Mahrt, Tim

Permalink

https://hdl.handle.net/2142/100934 Copy

Description

Title

Acoustic cues for the perception of the information status of words in speech

Author(s)

Mahrt, Tim

Issue Date

2018-04-09

Director of Research (if dissertation) or Advisor (if thesis)

Cole, Jennifer S.

Doctoral Committee Chair(s)

Cole, Jennifer S.

Committee Member(s)

• Hualde, José I.
• Watson, Duane G.
• Hasegawa-Johnson, Mark

Department of Study

Linguistics

Discipline

Linguistics

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Speech Prosody
• Intonation
• Information Structure
• Speech Perception

Abstract

In English, speakers highlight new or semantically focused information by prosodic enhancement–greater pitch, greater duration, and greater intensity–as compared to words that are already given in, or inferable from the prior discourse context. If speakers signal discourse meaning in their speech, then listeners may use that information in comprehending utterance meaning. This dissertation includes four studies that investigate the perception of four different categories of discourse meaning (i.e., meaning related to information structure for a noun in subject position: broad focus, contrastive focus, narrow focus, and discourse-given). In the first experiment, the perception of the subject noun’s information structure in terms of these categories was examined in the absence of any explicit discourse context. Participants were randomly assigned to an experiment testing one pair of information structure categories, for a total of six experiments covering all possible pairings of the four information structure categories. Participants heard recorded sentences in one of two conditions that differed in the information structure of the subject noun. The results show that response accuracy varied greatly depending on the pair of information structure categories that was tested, suggesting that meaning related to information structure is not evoked from acoustic cues alone. In the second experiment, naturally produced utterances were resynthesized in pitch, duration, and intensity to examine the individual contribution of each acoustic cue to the perception of two information structure categories: broad focus and contrastive focus–one of the pairs that was well distinguished in the first experiment. Pitch was found to be the key feature that when changed, led to changes in the perception of information structure. In experiments three and four, the perception of all four information structure categories of the subject noun was again examined, now with an explicit discourse context presented to participants. Participants were again randomly assigned to experiments testing two of the four discourse categories, and were presented with a series of two question-answer mini-dialogues. The two mini-dialogues differed in whether the question was explicitly about the subject noun (Who broke the record?) or was about the sentence as a whole (What happened?). On each trial, one question-answer pair was congruent in that the accentuation of the subject noun in the answer was appropriate for the question that preceded it, while the other question-answer pair was prosodically incongruent, i.e., the accentuation of the subject noun was not appropriate for the preceding question. In these experiments, accent contrasts were naturally produced and involved all acoustic prosodic dimensions: pitch, duration, and intensity. The participants’ task was to choose the mini-dialogue that was prosodically congruent. Similar to the results of experiment 1, accuracy depended on the pair of information structure conditions that were tested in the experiment, providing more evidence that listeners do not perceive specific discourse meaning solely on the basis of prosodic cues. The four studies here show that while there may be acoustic cues that differentiate information structural meaning, the prosodic cues in an utterance are appropriate for a variety of discourse meanings. The results suggest that listeners have biases about the prototypical prosodic cues associated with particular information structural meanings. These findings suggest that listeners do not have strong associations between discrete information status categories and the patterning of acoustic prosodic that they use in comprehending discourse meaning.

Graduation Semester

2018-05

Type of Resource

text

Permalink

http://hdl.handle.net/2142/100934

Copyright and License Information

Copyright 2018 Tim Mahrt

Owning Collections