Bidirectional decay of auditory memory traces for pitch in speech sounds

Previous studies have shown human listeners have greater detection sensitivity to pitch increment than decrement in successive sounds. Assuming deviance detection is based on the comparison between the memory trace of a recent stimulus and the neural representation of a new stimulus, one hypothesis is that this differential sensitivity between increment and decrement is caused by the downward decay of pitch’s memory trace. Under the same assumption, the present study found bidirectional—as opposed to the predicted unidirectional—decay of memory traces for pitch in speech sounds by measuring listeners' sensitivity to pitch change over varying time intervals with an AX discrimination task. Three properties in the AX task were randomly sampled from preset ranges: (1) f0 of the A token (163:320 Hz), (2) difference between the f0s of A and X (-30:30Hz), and (3) the inter-stimulus interval (ISI; 0:3 s). We found when the stimuli were in the lower portion of the speaker’s pitch range, listeners were less sensitive to pitch increments at larger ISIs. Meanwhile, when the stimuli were in the higher pitch range, listeners were less sensitive to pitch decrements at larger ISIs. These results suggest memory traces for pitch in speech sounds decay toward a center pitch.Previous studies have shown human listeners have greater detection sensitivity to pitch increment than decrement in successive sounds. Assuming deviance detection is based on the comparison between the memory trace of a recent stimulus and the neural representation of a new stimulus, one hypothesis is that this differential sensitivity between increment and decrement is caused by the downward decay of pitch’s memory trace. Under the same assumption, the present study found bidirectional—as opposed to the predicted unidirectional—decay of memory traces for pitch in speech sounds by measuring listeners' sensitivity to pitch change over varying time intervals with an AX discrimination task. Three properties in the AX task were randomly sampled from preset ranges: (1) f0 of the A token (163:320 Hz), (2) difference between the f0s of A and X (-30:30Hz), and (3) the inter-stimulus interval (ISI; 0:3 s). We found when the stimuli were in the lower portion of the speaker’s pitch range, listeners were less sensiti...