Advantages of the Discriminability Criterion for a Loudness Scale

Abstract

It is argued that a loudness scale based on the discriminability criterion would be better than one based on those methods which assume that direct numerical responses adequately reflect the nature of the loudness experience. The evidence indicates that, first, there is just as good a logical basis for accepting the discriminability criterion as for accepting the direct response criterion. In both cases we have to assert a face validity which is arbitrary and may be incorrect. Second, the scales based on the discriminability criterion show less interobserver variability, are influenced less by specific experimental conditions, and are less affected by stimulus context than scales based on the fractionation procedure - the major direct response method used. Third, most of the direct response procedures give results which agree with scales based on a discriminability criterion. It is, in fact, only the ratio scaling or fractionation procedures which give results at variance with all the rest.