Neural Systems Underlying Speech Perception

Sheila E. Blumstein, Ph.D.

Brown University

Funded in September, 2003: $100000 for 3 years
LAY SUMMARY . ABSTRACT . HYPOTHESIS . FINDINGS . SELECTED PUBLICATIONS .

ABSTRACT

back to top

Neural Systems Underlying Speech Perception

Event-related functional magnetic resonance imaging (fMRI) will be used to investigate the neural mechanisms underlying normal speech processing. The specific aims of this research are to determine whether the neural mechanisms underlying the processing of the spectral and temporal properties of speech differ and to identify the neural systems underlying the processing of the internal structure of phonetic categories. To this end, the processing of the temporal/spectral properties of speech will be investigated by exploring the neural systems underlying two acoustic properties—voice-onset time, a temporal acoustic property that distinguishes voiced and voiceless phonetic categories in stop consonants in English, e.g., [t] vs. [d], and formant frequency space, a spectral acoustic property that distinguishes the phonetic categories giving rise to vowel quality, e.g., [i] as in beat vs. [I] as in bit.

Subjects will be presented with synthetic speech continua and will perform either a phonetic categorization task or a discrimination task. The long-range goal of this research is to understand the nature of the neural mechanisms underlying language processing after focal brain injury. This research will provide necessary baseline measures for such study.

HYPOTHESIS

back to top

Hypothesis:
The acoustic properties corresponding to the phonetic dimensions of speech will be processed preferentially by the left hemisphere irrespective of their acoustic structure, owing to the functional role that sound structure plays in language processing.  In contrast, right hemisphere activation will be modulated by different aspects of the acoustic signal, with increased activation for those acoustic properties that are defined in terms of fine spectral detail over a time window of 150 ms or greater. Modulation of activation in the left frontal areas will occur as a function of the phonetic category goodness.

Goals:
The goal of this research is to elucidate the neural systems underlying the auditory processing of speech. We will explore whether the computational mechanisms of the left and right hemisphere differ, preferentially processing different aspects of the acoustic signal contributing to the sound structure of language. We will also investigate whether the functional role of anterior and posterior brain structures in the processing of speech differs.  The results of this research will serve as the basis for exploring the neural systems underlying speech processing in patients with focal brain injury.

Methods:
Event-related fMRI will be used in normal subjects to measure dynamic changes in blood flow to regions of the human brain as subjects perform a series of speech processing tasks. Listeners will be presented with synthetic speech continua that vary between two phonetic categories by parametrically varying a particular acoustic parameter. Two tasks will be used: a phonetic categorization task and a discrimination task. In the phonetic categorization task, subjects will be required to categorize each auditorily presented stimulus by pressing one of two buttons corresponding to each phonetic category. In the discrimination task, subjects will be presented with adjacent pairs of stimuli from the acoustic continua and will be required to make same-different judgments by pressing the appropriate response button.  In both tasks, measures will be taken of performance and reaction time latency. In all experiments, a sparse sampling design will be used such that the test stimuli will be presented to listeners between volumetric acquisitions.  The imaging data will be collected using a 1.5 T system and will be analyzed using AFNI.

Follow-on Funding:
National Institutes of Health, NIDCD, principal investigator, Neural Basis of Lexical and Speech Processing, 2004-2009.

National Science Foundation, co-principal investigator, Major Research Infrastructure Program, awarded $2 million for purchase of a 3T magnet, 2005.

 

FINDINGS

back to top

We completed 4 experiments exploring the neural systems underlying two acoustic properties—voice-onset time, a temporal acoustic property that distinguishes voiced and voiceless phonetic categories in stop consonants in English, e.g., [t] vs. [d], and formant frequency space, a spectral acoustic property that distinguishes vowel quality, e.g., [i] as in beat vs. [I] as in bit

The results of a study on voice-onset time (VOT) perception using a phonetic categorization task indicated that VOT is processed primarily by the left hemisphere and that the neural system shows graded activation with details of phonetic category structure retained throughout the phonetic processing stream. In a second study investigating the discrimination of voice-onset time, two distinct hemispheric patterns emerged; left temporal regions appeared to be recruited in initial acoustic-perceptual analysis and right frontal regions appeared to be recruited with increased processing demands.

To examine the neural systems underlying the spectral properties of speech, we investigated whether the right hemisphere (RH) is recruited for the analysis of vowel quality, and whether its involvement is influenced by the acoustic space within which vowels reside, on the one hand, and by vowel duration, on the other. Results showed that the size of the vowel space affected hemispheric processing, with increased left hemisphere activation in frontal and temporal regions for vowels from a narrow vowel space compared to vowels from a wide vowel space. In the second experiment, results indicate strong bilateral activation, with increasing right hemisphere activation as the duration of the stimuli increased suggesting that the LH processes the spectral properties of vowels irrespective of vowel length, but the right hemisphere has a particular advantage at long vowel durations.

 

SELECTED PUBLICATIONS

back to top

Hutchison E.R., Blumstein S.E., and Myers E.B. An event-related investigation of voice-onset time discrimination. Neuroimage. 2008 Mar 1;40(1):342-52.

Blumstein S.E., Myers E.B., and Rissman J. The perception of voice-onset time: An fMRI investigation of phonetic category structure. J Cogn Neurosci. 2005 Sep;17(9):1353-66.

McNellis M. and Blumstein S.E. Self-organizing dynamics of lexical access in normals and aphasics. J Cogn Neurosci. 2001 Feb 15;13(2):151-70.

Burton M.W., Small S.L., and Blumstein S.E. The role of segmentation in phonological processing: An fMRI investigation.  J Cogn Neurosci. 2000 Jul;12(4):679-90.

Blumstein, S.E. 2000. Deficits of speech production and speech perception in aphasia. In R. Berndt (Ed.). Handbook of neuropsychology, 2nd edition, Vol.2, The Netherlands: Elsevier Science.