Multidimensional analyses of circuit quality judgments.

Abstract

Listeners judged paired comparisons of speech samples transmitted through 22 circuits, each of which introduced a different type of distortion. One group of listeners made preference judgments and another group made similarity judgments. A different multidimensional scaling procedure was applied to each set of data. Both of these procedures represent the stimuli as points in a multidimensional subjective space whose coordinates correspond to the subjective attributes underlying the judgments. The procedure applied to the preference judgments also represents the listeners as vectors in the same subjective space. A listener's vector is located according to the relative weight he assigns to each attribute. The results of these analyses revealed a high degree of correspondence between the three-dimensional solutions representing each type of judgment. The underlying attributes common to the two spaces are interpreted as over-all clarity, a distinction between signal distortion and background distortion, and subjective loudness. Listeners generally agree in their preferences for signals according to over-all clarity. Individual differences in preference for either signal or background distortion account for most of the variability between listeners. Some additional variability is introduced by individual differences in preferred loudness.