Perception of Pitch in Pulse Wave Forms whose Power Spectra are Flat

These experiments were performed to explore the upper limits of pitch perception based on the time waveform of the stimulus. Flanagan and Guttman, using pulses with various polarity patterns, showed that pitch may be assigned to a sound based on either the time waveform or the Fourier spectrum. Their results show these two mechanisms in conflict in the region between 100 and 200 hertz. The stimulus used here consisted of a train of uniformly spaced pulses of equal amplitude whose polarities were randomly chosen. Such a signal is shown to have a flat power spectrum and as such should offer no competition to the time-based pitch mechanism. Subjects listened alternately to the stimulus and to a unipolar pulse generator whose pulse rate was under their control. The subjects were instructed to match the two pitches. Some musically talented subjects were found who could match the pitch even when the random polarity pulse rate was as high as 9.5 kHz. Matching is defined to occur when an integral relation exists between the pulse rates as set by the experimenter and the subject Matching at high rates was found to occur at stimulus levels as low as 10 db or when both the stimulus and matching signals were high-pass filtered at 8 kHz . Investigation of the short term spectrum of the random polarity pulse train shows that there are clues to the pulse rate even though the long term power spectrum is flat. One of these clues is the different probability distribution of the amplitude peaks at frequencies of nfp/2 where fp is the pulse rate. This variability of amplitude distribution can be reduced, and the frequencies which then occur changed, by using polarity patterns randomly chosen from a certain set of patterns. However, experiments performed using pulse patterns showed little or no change in subjects' ability to match.