Temporal processing of the pitch of complex tones.

Abstract

The effect of tone duration on fundamental frequency (F0) discrimination is greater for complexes containing unresolved harmonics than for those containing resolved harmonics [Plack and Carlyon, J. Acoust. Soc. Am. 98, 1355-1364 (1995)]. Three experiments explored this effect further. The first experiment measured sensitivity (as d') to fundamental frequency (F0) differences for two complexes, both with an F0 of 250 Hz. The first complex was low-pass filtered at 1875 Hz to create a resolved complex and the second was bandpass filtered between 5500 and 7500 Hz to create an unresolved complex. The harmonics for the resolved complex were selected so that no two harmonics were the same between the two observation intervals. Performance for both complexes was measured for tone durations of 20, 40, 80, and 160 ms. For the unresolved complex, the effect of duration was greater than that for the resolved complex and greater than the predictions of a "multiple-looks" model assuming either peripheral (before sampling) or central (after combining samples) sources of variance. The second experiment replicated these results using an F0 of 62.5 Hz with the cutoff frequencies of the bandpass filters divided by four, confirming that the effect is related to resolvability and not to spectral region. In the final experiment, F0 discrimination for pairs of complexes separated by a temporal gap was measured relative to that for one complex. Performance for the resolved and unresolved complexes was similar: Very little effect of gap duration was observed and the results were consistent with the predictions of the peripheral-variance multiple-looks model. Taken together, the results suggest that the pitch mechanism for resolved harmonics uses a relatively short sampling window of around 20 ms, while the mechanism for unresolved harmonics may use a more complex strategy for optimizing the combination of information over time, perhaps involving a flexible integration time.