Critical bands and mixed-frequency scaling: sequential dependencies, equal-loudness contours, and power function exponents.

Stimuli of random intensities and various but predictable frequencies were presented for repeated magnitude estimations on the same scale (mixed-frequency scaling). The frequencies for a particular judgment session were selected so that they lay either inside each other's critical bands or outside them. Contrastive dependencies of current magnitude estimation responses on previous stimuli of a different frequency were significantly affected by whether the two frequencies were inside or outside each other's critical bands, while assimilative dependencies were not. This reinforces the idea that such dependencies are sensory in nature and arise from a different mechanism than do the assimilative dependencies. Mixed-frequency scaling of loudness also gives rise to cross-frequency matching functions from which equal-loudness contours can be calculated. These contours calculated from the present judgments are similar to those produced from other methods, even when they are extrapolated to untested intensities. Power function exponents for loudness scaled in this way are larger for lower frequencies, as has been found in previous studies, and are consistent with the flattening of the calculated equal-loudness contours for low frequencies as intensity increases.