DICHOTIC PITCHES AS ILLUSIONS OF BINAURAL UNMASKING - I - HUGGINS PITCH AND THE BINAURAL EDGE PITCH

The two most salient dichotic pitches, the Huggins pitch (HP) and the binaural edge pitch (BEP), are produced by applying interaural phase t ransitions of 360 and 180 degrees, respectively, to a broadband noise. This paper examines accounts of these pitches, concentrating on a ''c entral activity pattern'' (CAP) model and a ''modified equalization-ca ncellation'' (mE-C) model. The CAP model proposes that a dichotic pitc h is heard at frequency f when an individual across-frequency scan in an interaural cross-correlation matrix contains a sharp peak atf. The mE-C model proposes that a dichotic pitch is heard when a plot of inte raural decorrelation against frequency contains a peak at f. The predi ctions of the models diverge for the BEP at very narrow transition ban dwidths: the mE-C model predicts that salience is sustained, while the CAP model predicts that salience declines and that the dominant perce pt is of the in-phase segment of the noise. Experiment 1 showed that t he salience of the BEP was sustained at the narrowest bandwidths that could be generated (0.5% of the transition frequency). Experiment 2 co nfirmed that the pitch of a BEP produced by a 0.5% transition bandwidt h was close to the frequency of the transition band. Experiment 3 show ed that pairs of simultaneous narrow 180-degree transitions, whose fre quencies corresponded to vowel formants, were perceived as the intende d vowels. Moreover, the same vowels were perceived whether the in-phas e portion of the noise lay between the two transition frequencies or o n either side of them. In contrast, different patterns of identificati on responses were made to diotic band-pass and band-stop noises whose cutoff frequencies corresponded to the same formants. Thus, the vowel- identification responses made to the dichotic stimuli were not based o n hearing the in-phase portions of the noise as formants. These result s are not predicted by the CAP model but are consistent with the mE-C model. It is argued that the mE-C model provides a more coherent and p arsimonious account of many aspects of the HP and the BEP than do alte rnative models. (C) 1998 Acoustical Society of America.