Wilson, J.P., Baker, R.J., Whitehead, M.L. (1988). Level dependence of frequency tuning in human ears. In: Basic Issues in Hearing (Duifhuis, H., Horst, J.W., Wit, H.P., eds.), Academic, London, 80-87

... Discussion:

(1) The psychophysical pitch shift has been the topic of considerable controversy probably partly due to large individual differences. The widely quoted Stevens' rule that with increasing intensity pitch shifts downward below 1 kHz and upwards above, appears to derived chiefly from data gathered at higher sound levels. For levels as low as those used in the present study, however, many of the individual curves for low frequencies actually start out with a slight positive slope until about 60 dB SPL where the direction reverses (Morgan et al. 1951, Ward 1970, Terhardt 1974, Verschuure and van Meeteren 1975, Jesteadt and Neff 1982). This is less apparent in the data of Burns (1982). The most comprehensive study on a single subject (van den Brink, 1979) consistently shows a positive slope of pitch versus intensity over the range of frequencies used in the present study and between 10 and 60 dB SPL. Over octave intervals from 500 Hz to 4 kHz these values average 0.036, 0.046, 0.078 %/dB. Similar values of 0.03, 0.07, 0.05, 0.04 %/dB can be calculated from Verschuure and van Meeteren (1975) at 300, 500, 1k and 2 kHz respectively, for low sound levels. For a simple place model a decrease in frequency of tuning of the underlying auditory filters implies an increase in pitch, so that these values correspond well with the present findings. We would like, therefore, to suggest a new (WBW) pitch-intensity rule that "PItch usually increases with sound level up to about 60 dB SPL". Above 60 dB SPL the Stevens' rule appears to apply with pitch decreasing for low frequencies and increasing for high frequencies. With these relationships in mind much of the previous psychoacoustical data appears less idiosyncratic.

(2) Physiological recordings of iso-response contours to pure tones have not been determined with sufficient precision to be compared with the present data. Reverse correlation techniques using noise stimulation in cats (Evans, 1977), however, do show shifts of CF at high levels consistent with Stevens' rule.

(3) The shift downwards in frequency of an SOAE induced by a higher frequency suppressor tone (Wilson and Sutton, 1981, Fig.2a,c) may be a related phenomenon where in the present case the stimulus is also acting as "suppressor". Intriguingly, the slope of frequency shift at 3 kHz is comparable to the present data (-0.66 Hz/dB, -0.022 %/dB) and appears to flatten-out above 60 dB SPL.