Towards a hybrid model of speech prosody

Speech is conveyed through a richly detailed acoustic signal, and each spoken instance of a word varies. Yet humans recognize words quickly, with few errors and typically with no conscious effort. Linguists hypothesize that abstract categories aid the process of speech comprehension because complex details can be mapped to a category, after which the detail of an instance can be forgotten. At the same time, research shows acoustic detail guides comprehension in real time and is retained in implicit memory. This dissertation addresses language processing for speech prosody. This study consists of five perception experiments that test the representation of pitch accents in memory, and directly compares memory for pitch accents to memory for phonemes. Results show that (1) explicit memory for prosody is more accurate for categorical differences, rather than equivalently large within-category differences, and (2) that, unlike phoneme categories, listeners are able to explicitly remember within-category detail for pitch accents, but that this memory decays quickly. Based on these results, I propose a hybrid model of comprehension of prosody that involves activating both previously proposed abstract categories and gradient detail.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Linguistics

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2018

Author dc:creator

• Kimball, Amelia E.

Contributors dc:contributor

• Cole, Jennifer S.
• Dell, Gary
• Hualde, José Ignacio
• Watson, Duane

• prosody
• linguistics
• exemplar memory
• episodic memory
• Autosegmental Metrical Model
• hybrid memory model
• memory
• language
• intonation

Statement dc:rights

• Copyright 2018 Amelia Kimball

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/101291

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/101291