Subject: Carrell/Opie research From: "Albert Bregman, Tel: 514-398-6103" <IN09(at)MUSICB.MCGILL.CA> Date: Tue, 3 Nov 1992 21:49:34 EST
November 3, 1992 Open letter to Tom Carrell and Jane Opie, I would like to draw the attention of the members of the AUDITORY list to your interesting finding just published in Perception & Psychophysics vol.52, no.4. You used used sentences made of sine-wave-analog-speech, which you referred to as "time-varying sinusoidal" (TVS) sentences. The sentences were amplitude modulated at various rates (e.g., 50 Hz) that were fixed over the course of the sentence. Modulated sentences (50 Hz and 100 Hz) were rated as more natural than unmodulated ones and more phonemes in them were correctly identified. AM of 200 Hz was worse than 50 or 100 Hz, and even worse than unmodulated sentences. Your discussion relates the results to "comodulation masking release" (CMR). In your view the common AM applied to all the component sinusoidal glides helped them to be heard as a single auditory object. This is a noteworthy finding. One thing to recall is that the AM creates sidebands that are not necessarily related harmonically to the 100 Hz modulation rate. At almost every instant, the signal would be quite inharmonic. Even so, the perceptual integration took place. You mention in your discussion that the next phase of the research involves looking at the intelligibility of such signals in the presence of other sounds. I don't know whether you have thought of this, but a natural development of this research would be to present mixtures of pairs of TVS sentences, and to look at their intelligibility as a function of the AM frequency separation between the members of a pair. This would be the analog of experiments done by Brokx/Noteboom, Scheffers, Chalikia/Bregman, Assman/Summerfield, and others, in which the F0 difference between members of a pair of mixed speech sounds was varied. These researchers used speech that was either natural or was synthesized so as to have a natural harmonic structure. Your method would be able to show whether it was the fact that the two speech signals were presented on different HARMONIC SERIES that was crucial or that at least some part of the within-signal integration derived simply from the fact that all the spectral regions showed the effects of a common amplitude modulation. If an effect of AM difference were found in your inharmonic stimuli, this would support the results by Bregman, Abramson, Doehring and Darwin on the fusion of amplitude modulated tones. This research found that the fusion of inharmonic complexes was affected by common (and in-phase) AM. Congratulations on your imaginative study. - Al Bregman P.S. Is anybody else doing research that pertains to these issues? If so, would you like to get some feedback from the group?