1. The neural correlates of low pitches produced by complex tones were studied by analyzing temporal discharge patterns of auditory nerve fibers in Dial-anesthetized cats. In the previous paper it was observed that, for harmonic stimuli, the most frequent interspike interval present in the population of auditory nerve fibers always corresponded to the perceived pitch (predominant interval hypothesis). The fraction of these most frequent intervals relative to the total number of intervals qualitatively corresponded to strength (salience) of the low pitches that are heard. 2. This paper addresses the neural correlates of stimuli that produce more complex patterns of pitch judgments, such as shifts in pitch and multiple pitches. Correlates of pitch shift and pitch ambiguity were investigated with the use of harmonic and inharmonic amplitude-modulated (AM) tones varying either in carrier frequency or modulation frequency. Pitches estimated from the pooled interval distributions showed shifts corresponding to "the first effect of pitch shift" (de Boer's rule) that is observed psychophysically. Pooled interval distributions in response to inharmonic stimulus segments showed multiple maxima corresponding to the multiple pitches heard by human listeners (pitch ambiguity). 3. AM and quasi-frequency-modulated tones with low carrier frequencies produce very similar patterns of pitch judgments, despite great differences in their phase spectra and waveform envelopes. Pitches estimated from pooled interval distributions were remarkably similar for the two kinds of stimuli, consistent with the psychophysically observed phase invariance of pitches produced by sets of low-frequency components. 4. Trains of clicks having uniform and alternating polarities were used to investigate the relation between pitches associated with periodicity and those associated with click rate. For unipolar click trains, where periodicity and rate coincide, physiologically estimated pitches closely follow the fundamental period. This corresponds to the pitch at the fundamental frequency (F0) that is heard. For alternating click trains, where rate and periodicity do not coincide, physiologically estimated pitches always closely followed the fundamental period. Although these pitch estimates corresponded to periodicity pitches that are heard for F0s > 150 Hz, they did not correspond to the rate pitches that are heard for F0s < 150 Hz. The predominant interval hypothesis thus failed to predict rate pitch. 5. When alternating-polarity click trains are high-pass filtered, rate pitches are strengthened and can also be heard at F0s > 150 Hz. Pitches for high-pass-filtered alternating click trains were estimated from pooled responses of fibers with characteristic frequencies (CFs) > 2 kHz. Roughly equal numbers of intervals at 1/rate and 1/F0 were found for all F0s studied, from 80 to 160 Hz, producing pitch estimates consistent with the rate pitches that are heard after high-pass filtering. The existence region for rate pitch also coincided with the presence of clear periodicities related to the click rate in pooled peristimulus time histograms. These periodicities were strongest for ensembles of fibers with CFs > 2 kHz, where there is widespread synchrony of discharges across many fibers. 6. The "dominance region for pitch" was studied with the use of two harmonic complexes consisting of harmonics 3-5 of one F0 and harmonics 6-12 of another fundamental 20% higher in frequency. When the complexes were presented individually, pitch estimates were always close to the fundamental of the complex. When the complexes were presented concurrently, pitch estimates always followed the fundamental of harmonics 3-5 for F0s of 150-480 Hz. For F0s of 125-150 Hz, pitch estimates followed one or the other fundamental, and for F0s < 125 Hz, pitch estimates followed the fundamental of harmonics 6-12. (ABSTRACT TRUNCATED)
Prediction of the Fundamental Period of Infilled RC Frame Structures Using Artificial Neural Networks
