Tone bursts of 2, 4, 6, and 8 kHz, with rise-fall times of 20 msec and durations of 200 msec, were presented to human subjects, through earphones, at sensation levels (SLs) of 10, 30, 50, or 70 dB. The tones were delivered once per second, in blocks of 60, at a specific frequency and SL. Sixteen combinations of frequency and SL were employed, and were presented in a counterbalanced order, with four replications per combination for each subject. Simultaneously with each burst, the slow, evoked response (V potential) was recorded between vertex and mastoid. Evoked response amplitudes (N1 − P2) were found to decrease with increasing frequency, particularly at higher sensation levels. The slopes of the input-output functions also decreased with frequency, though differences between subjects influenced slope even more than frequency. The input-output data, when fitted by straight regression lines, generally showed the same goodness of fit, whether plotted on logarithmic or linear ordinates, versus SL.