Roy D. Patterson
MRC Appl. Psychol. Unit, 15 Chaucer Rd., Cambridge CB2 2EF, UK
``Damped sounds,'' constructed by repeating a short segment of a sinusoid with an exponential decay, were compared with ``ramped sounds,'' constructed by reversing each damped sound in time. As the half-life of the decay function decreases through values less than the repetition period, the continuity of the sinusoid is disrupted and its distinctive sound quality fades away. However, these changes occur far sooner for the damped sound. Damped/ramped pairs with the same carrier frequency and repetition rate have identical power spectra, so simple spectral models predict no perceptual difference for these sounds. Excitation patterns generated with filterbanks and leaky integrators show that damped sounds have sharper spectral peaks than corresponding ramped sounds, so auditory models with simple leaky integrators predict the reverse of what is heard. A new model that employs strobed temporal integration to stabilize neural activity patterns suggests that ramped stimuli sound more sinusoidal because they produce more time intervals near the carrier period, although not in the carrier channel, and the perception is more continuous because the strobe threshold function is asymmetric in time.