Annoyance and spectral contrast are cues for similarity and preference of sounds

Previous research has suggested that perceived similarity is based on primarily cognitive processes, whereas preferences are based to a larger extent on affective processes. This was put to an empirical test utilizing 15 complex sounds as stimuli and 25 subjects for the assessments. Various versions of multidimensional scaling were used as a method of comparison. The results show that data analyses must take into account individual differences in similarity and non-preference. Contrary to the hypothesis expressed, both similarity and non-preference were found to be based mainly on affective responses because a major proportion of the explained variance originated from the perceived annoyance of sound. This was not true for perceived loudness or for the acoustic variables of Zwicker's loudness and Aures' sharpness. Spectral contrast calculated as the number of maxima-in the normalized Zwicker's specific loudness spectra was found to be the best acoustic candidate for explaining at the individual level what properties of sound cause them to be perceived as similar or non-preferred. (C) 2002 Academic Press.