Speaker invariance for phonetic information: An fMRI investigation

Abstract

The current study explored how listeners map the variable acoustic input onto a common sound structure representation while being able to retain phonetic detail to distinguish among the identity of talkers. An adaptation paradigm was utilised to examine areas that showed an equal neural response (equal release from adaptation) to phonetic change when spoken by the same speaker and when spoken by two different speakers, and insensitivity (failure to show release from adaptation) when the same phonetic input was spoken by a different speaker. Neural areas that showed speaker invariance were located in the anterior portion of the middle superior temporal gyrus bilaterally. These findings provide support for the view that speaker normalisation processes allow for the translation of a variable speech input to a common abstract sound structure. That this process appears to occur early in the processing stream, recruiting temporal structures, suggests that this mapping takes place prelexically, before sound structure input is mapped onto lexical representations.

Keywords:

• Speech processing
• Speaker invariance

ACKNOWLEDGEMENTS

This research was supported in part by NIH NIDCD Grant RO1 DC006220 to Brown University. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Deafness and Other Communication Disorders or the National Institutes of Health. Many thanks to John Mertus for his help in programming the experiment and hardware support and to Kathy Kurowski for her assistance in recording the stimuli.