Jamshed Bharucha - US grants

The broad goals of this project are to understand how the brain encodes sound patterns and how the prior acquisition of some patterns influences the perception of others. Experiments and computer simulations will provide tests of a theory of these processes. The experiments will measure the speed and accuracy with which the perceptual system processes sounds as a function of the pattern within which the sounds are embedded. The theory holds that sounds will be processed more quickly and accurately and will be heard as more consonant when they occur in typical patterns than when they are in atypical patterns. These experiments will provide objective measures of the perceptual expectations that depend upon the kinds of patterns to which people have been previously exposed. The experiments will employ digitally synthesized musical sequences. A computer will record the speed and accuracy with which a person judges whether or not a designated sound is in tune. Computer simulations will generate precise predictions about how a network of neurons in the brain could encode patterns and consequently influence the perception of other patterns. The simulations will employ recently developed computational algorithms that model the organization of neurons into networks that exhibit intelligent behavior. This project will focus on neural networks that can encode sequential patterns. As an automatic consequence of encoding a large numbers of sequences, a network will learn the regularities inherent in the set of sequences as a whole. The sequential expectations generated by the network will then be compared with those of people as measured in the experiments described above. These results will contribute to our understanding of how the ability to learn sequential patterns depends upon the patterns to which we have previously been exposed. The results will thus have implications for how this ability might be improved by an appropriate schedule of exposure.

Many speech sounds and musical tones are composed of harmonics, which are frequencies that are roughly integer multiples of the lowest frequency present. The lowest or fundamental frequency of a harmonic sound usually determines its pitch. Yet when the fundamental frequency is removed while the upper harmonics remain intact (as is the case for low voices heard over the telephone), the pitch remains unchanged. Furthermore, shifting the upper frequencies causes the pitch to change in complex ways. This research will involve the refinement and testing of a theory of auditory perception that accounts for these and other phenomena. The theory postulates that as a result of the properties of auditory neurons and general principles of neural connectivity, exposure to speech results in the neural encoding of harmonic patterns. These neural encodings in turn filter the sounds we hear so that frequencies that are blocked or distorted during transmission are effectively restored by the auditory system. The research will test the theory using computer simulations of a neural-net model and using perceptual experiments. Understanding how the auditory system recognizes sound patterns may in the future guide the development of "smart" hearing devices that can compensate for selective loss of auditory function.