The Representation of the Pitch of Vowel Sounds in Ferret Auditory Cortex

Abstract

According to ANSI, pitch is defined as that attribute according to which sounds can be ordered on a scale from low to high. However, most studies of pitch so far asked merely whether a subject can detect a pitch change, not whether he can also determine the direction of the pitch change. Using behavioural testing after operand conditioning we have shown that ferrets can discriminate rising from falling pitches in artificial vowels, although behavioural thresholds (with Weber fractions of similar to 0.4) may seem high compared to humans. We have further studied the cortical representation of these sounds in five distinct areas of ferret auditory cortex, Al, AAF, PPF, PSF, and ADF. Of over 600 units which responded to our artificial vowels, approximately half were sensitive to pitch changes. Thirty-eight percent of these exhibited high-pass, another 38% low-pass characteristics. The remaining 24% had non-monotonic pitch tuning curves. Our results did not reveal a clear `pitch area'. Rather, pitch sensitive neurons could be found throughout all cortical areas studied. We used neurometric analysis to make the observed neural sensitivities more directly comparable to the animal's behavioural performance. Only a small number of units yielded neurometric curves which approached or matched the animal's psychometric functions. Furthermore a unit's neurometric could account for the animal's behavioural capabilities only around a small range of `reference pitch' values around the steepest part of the unit's pitch tuning curves. We also developed a ``population neurometric{''} analysis, with which we decoded the activity of up to 60 simultaneously recorded units. These population responses often performed substantially better than the individual constituent neurons. Populations also performed well over a wider range of reference pitches. Interestingly, spike count or spike latency based decoding of the population response yielded very similar neurometric curves.