1. Faculty of Arts, Humanities and Cultures
  2. School of Languages, Cultures and Societies
  3. Research projects
  4. Speech tempo perception and missing sounds

Research project

Speech tempo perception and missing sounds

  • Start date: 4 September 2017
  • End date: 1 July 2022
  • Funder: Leverhulme Trust
  • Primary investigator: Dr Leendert Plug
  • Co-investigators: Robert Lennon, Yue Zheng
  • External co-investigators: Rachel Smith (University of Glasgow)

Value

£195,000

Description

The aim of this research was to improve our understanding of how speech tempo is perceived by listeners. It addressed three questions.

1. What tempo measurement technique best reflects listeners’ tempo judgements? Measuring speech tempo generally involves counting linguistic units: e.g. syllables per second. Various specific techniques exist. We investigated experimentally which technique is most in line with listeners’ tempo estimates.

2. Is there evidence for listeners ‘restoring’ missing sounds in tempo perception? If listeners interpret a sport-like pronunciation of support as having two syllables, they have ‘restored’ a missing sound. We played listeners such pronunciations, convincing some that they heard sport and others that they heard support. We then assessed whether their interpretation affected their tempo estimates.

3. Is there evidence for acoustic reduction affecting tempo perception? Perhaps listeners do not count units when they estimate tempo, but judge more holistically how ‘rich’ a speech signal is. If so, utterances produced in different speaking styles should sound different in tempo. We tested this prediction experimentally.

Findings

We found that there is no single ‘best tempo measurement technique’: listeners are adept at detecting any differences that suggest one utterance is faster than another. Speech with more missing sounds can sound relatively fast or slow, depending on other factors that influence tempo perception.

We found that common measurement techniques produce closely correlating values. Therefore, for practical purposes, common techniques are equally good at capturing tempo variation.

We found evidence of listeners ‘restoring’ missing sounds when doing so allows them to distinguish two utterances in tempo that are otherwise very similar. We found evidence of speech style variation influencing tempo perception: listeners struggle to distinguish differences in pronunciation ‘precision’ from differences in ‘speed’.

Publications and outputs