Negative masking and the units problem in audition.

Masking functions constructed from pedestal levels bracketing absolute threshold may exhibit negative masking, particularly when stimuli are defined in terms of amplitude (pressure). Three experimental conditions using 10-ms 1000-Hztones in quiet, 1000-Hztones embedded in continuous noise, and 6500-Hztones in quiet, yielded negative masking when amplitudes were used to define the stimulus, with the greatest amount of negative masking occurring with 6500-Hztones. Two models were applied to the data: the transduction model, which assumes direct coupling, and the sensory analytical model, which assumes differential coupling. Maximum likelihood estimates were derived to indicate goodness-of-fit, and the Akaike information criterion and Bayesian information criterion were utilised to adjust for model complexity. Overall, both models effectively accounted for the data, though the sensory analytic model provided the best fits to the data and has the added quality of being based on underlying physiological processes.