Frequency discrimination at different frequency levels as indexed by electrophysiological and behavioral measures.

The present study systematically compared the neural and behavioral accuracy of discriminating a frequency change ("deviant") in a repetitive tone ("standard") across a frequency range of 250-4000 Hz. The sound structure (pure sinusoidal vs. harmonically rich tones) and the magnitude of frequency change (2.5%, 5%, 10%, 20%) were also varied. The accuracy of neural frequency-change detector was determined by comparing the auditory event-related potentials (ERP) elicited by deviant and standard stimuli in the absence of attention. In a separate behavioral task, subjects were to indicate when they noticed a frequency change. The ranges of the across-subject means of ERP parameters across the conditions were: the mismatch negativity (MMN) amplitude -0.9 to -4.9 microV, latency 125-218 ms, the P3a amplitude 0.3-3.2 microV, latency 239-304 ms. The ERP latency was shortest for the standard-stimulus frequency from 1000 to 2000 Hz suggesting that automatic frequency discrimination was the most accurate in that range. The ERP latencies and amplitudes correlated with the hit rate (HR) and reaction time (RT), with highest correlation found between the MMN amplitude and the HR (r=0.8). The harmonical tones elicited MMN and P3a with shorter latencies and larger amplitudes, than did pure sinusoidal tones in all frequency bands. The results may have implication to pitch-perception theories.