Coherence masking protection in brief noise complexes: effects of temporal patterns.

Three experiments examined listeners' thresholds for classifying the pitch of a target signal in a masking noise when it was presented alone as compared to when it was presented with a "cosignal." The target signal was a narrow band of noise centered on either 375 or 625 Hz and the masker was noise low-pass filtered at 1000 Hz. The cosignal provided no information about the pitch of the target signal but could potentially combine with it to form an auditory object; it was spectrally well separated from the target signal, consisting of a band of noise ranging from 2200 to 2900 Hz. Experiment 1 showed that identification thresholds were lower when the target signal was paired with the cosignal than when it was presented alone if the onsets and offsets of the target signal and cosignal were temporally synchronous. This is an instance of "coherence masking protection," a phenomenon that has previously been established in the perception of vowels [P.C. Gordon, Percept. Psychophys, 59, 232-242 (1997)]. The effect disappears when the cosignal leads and lags the target signal by short durations, a finding that also matches that observed previously with vowels. The finding that temporal relations between the components of a stimulus have similar effects on the perception of nonspeech noise complexes and speech sounds suggests that speech perception makes use of general auditory mechanisms for perceptual integration of this sort. Experiments 2 and 3 examine further the role of temporal relations between the onsets and offsets of the target signal and the cosignal in producing coherence masking protection. The results show that either onset synchrony or offset synchrony is sufficient to produce the effect when the cosignal is of greater duration than the target signal, but that only onset synchrony produces the effect when the target signal has greater duration than the cosignal. This pattern indicates that the target signal and cosignal do not contribute equally to the formation of auditory objects.