Auditory profile analysis: is there perceptual constancy for spectral shape for stimuli roved in frequency?

The ability to discriminate between two spectral shapes was investigated using a one-interval, two-alternative forced choice procedure. Each stimulus consisted of seven components equally spaced on a logarithmic frequency scale (frequency ratio of 1.2 between adjacent components). The two spectral shapes were produced by first generating a complex each of whose components was equally loud (when heard in isolation) and then increasing the level of either the third or the fifth component by a certain amount. To render pitch an unusable cue, the center frequencies of the stimuli were randomly varied between presentations (maintaining the ratio between adjacent components) over a range of about three octaves (279-2074 Hz). The data indicate that listeners were not able to distinguish between the two types of profiles even for very large increments in level of the third and the fifth components. The results suggest that there is no perceptual constancy for this type of spectral shape for stimuli roved over a sufficiently large frequency range. The spectral shape cannot be perceived independently of stimulus frequency. The results are related to the multi-channel model applied in profile analysis tasks.