Speech intelligibility and hemispheric asymmetry.
Doctoral thesis, UCL (University College London).
It is very rare, even in degraded listening environments, that we might confuse speech with a dog bark or vice versa, despite the fact that both are complex acoustic signals. Despite the solid assumption of left lateralisation in speech processing from clinical and anatomical observations, the results from brain imaging studies have been inconsistent. One possible cause for this controversy may come from the use of different imaging system. Using inadequate baselines, however, may bring more critical problem. In brain imaging studies, especially when cognitive subtraction is used, images of cognitive processes are generally derived by subtracting a control stimulus/task from an experimental counterpart. The two stimuli/tasks to be compared are expected to differ only in one factor/process and the difference in brain activations is thus considered to come from the particular difference between the two. This thus makes it difficult to find baseline stimuli/tasks that activate all but the process of interest. By far, spectrally rotated speech stands as a most satisfying control against intelligible speech as it is equally complex as speech but totally unintelligible. However, the spectral rotation so far has been a total rotation regardless of the source and the filter of sound, which are independent and heterogeneous by nature. A series of behavioural studies performed in this thesis showed that the source rotation did not significantly affect speech intelligibility whilst filter drastically decreased the intelligibility. Another possibility can be different brain imaging paradigms used. With carefully designed parametres using functional magnetic resonance imaging (fMRI), we confirmed that intelligible speech recruited predominantly the left superior temporal area, replicating the results from previous positron emission tomography (PET) and fMRI studies. Since the intervention of scanner noise has been an issue in auditory research using an MRI system, four imaging paradigms were compared and it is concluded that a sparse sampling with 8 seconds of repetition time had a clear advantage over longer repetition time with 16 seconds and a continuous sampling. This paradigm was used in the study investigating the effects of channel number and presence/absence of tonal variation on speech intelligibility. Intelligibility increased together with increasing number of band channels and showed drastic improvement especially in the range of 2 – 6 numbers of frequency channel bands. A brain imaging study followed with mixed subtraction and parametric designs and revealed that the right superior temporal gyrus responded most when pitch variation was provided in the speech, regardless of intelligibility, unlike the pitch variation in non-speech (spectrally rotated speech here). Increasing intelligibility with increased spectral detail showed linear increase in percent signal change in hemodynamic response in the left superior temporal gyrus. The current result supports a streamed hierarchical model, in which speech comprehension occurs predominantly in the left hemisphere.
|Title:||Speech intelligibility and hemispheric asymmetry|
|Additional information:||Authorisation for digitisation not received|
|UCL classification:||UCL > School of Life and Medical Sciences > Faculty of Brain Sciences > Psychology and Language Sciences (Division of) > Speech, Hearing and Phonetic Sciences|
Archive Staff Only