Low pitches of frequency-transposed stimuli: Salience estimates from a full population-interval model

Abstract

Transposed stimuli consisting of half-wave rectified modulators of high frequency carriers allow comparisons of pitch perception across frequency regions. While listeners can discriminate low F0-pitches of simple transposed stimuli (1 Fm-Fc), most cannot detect low pitches of complex transposed stimuli [Fm=300, 400, 500 Hz, Fc=4, 6.35, 10.08 kHz; Oxenham et al., PNAS (2004)]. Auditory nerve (AN) responses to a low-frequency harmonic complex (F0=100 Hz;n=3–5) and its transposition were simulated using filtering (middle ear, 48 gammatone filters, 144 fibers, human CF distribution, 100 Hz to 18 kHz), rectification, adaptive gain control, spontaneous activity (three classes), 65 dB SPL simulated level. Pitch salience was estimated by analyzing population-interval histograms using subharmonic interval sieves. The harmonic complex produced high salience (2.05), well above assumed pitch detection threshold (∼1.3). Simple transposed stimuli produced lower saliences (1–1.37). Some were below threshold, inconsistent with psychophysical pitch discrimination results. The complex transposed stimulus produced an F0-salience (1.31) near threshold, qualitatively consistent with their mixed detection results. Dramatic differences are thus seen between low-frequency and transposed stimuli. Clearly, for these and other high-frequency stimuli, full AN population-interval models can produce salience estimates that differ substantially from simple intuitions and/or partial implementations. Interpretational caution is therefore counseled.