Against “broadening with sound level”

Abstract

Spectral resolution is fundamental to audition. Auditory-nerve-fiber responses provide a window onto cochlear tuning. Typically, resolution is quantified using threshold tuning curves. Superficially, these suggest dramatic and systematic filter broadening with increasing intensity, a view often invoked to explain psychoacoustic phenomena. We recorded auditory-nerve-fiber spike times from normal-hearing chinchillas in response to 0.01-20 kHz wideband noise. We gathered data over a wide intensity range from each fiber. Linear filter estimates were obtained at each level by reverse correlation. We characterized their magnitude and phase response. We considered responses from 175 fibers (CF range: [0.1, 3.0] kHz; all SR groups), with noise level varying over a 40 to 100-dB range. 10-dB bandwidth varied minimally with sound level over a 40 to 50-dB range. Some fibers demonstrated phase pivoting with level; however, most did not follow this simple scheme. Common teaching is “auditory filters broaden with level,” usually in reference to threshold tuning curves. Here, gain functions constructed from responses to a fixed-level input show this broadening is not as severe as threshold tuning curves might suggest. Broadening is very subtle over a large range of “low” to “mid” input intensities (<70-dB SPL), before rapidly increasing at “high” levels (>80-dB SPL).