Loudness effect on pairwise comparisons and sorting
Abstract
Effect of loudness on the perceptual structure underlying a corpus of sounds is investigated by
two experimental methods: pairwise comparisons and sorting. Both methods are applied to a corpus of recordings
sounds presented with their ecological, non normalized loudness, and to the same corpus equalized in loudness. Two
types of perceptual structures (multidimensional scaling and hierarchical cluster analysis) are derived. Domination of
one auditory attribute – loudness – on less salient ones is discussed according to the two types of perceptual
structures. In the non-loudness-normalized corpus, loudness was the main factor explaining participant’s judgments
for both tasks. In particular, representations derived from sorting data almost solely reflected sound pressure level
difference between sounds. On the opposite, loudness normalization allowed the emergence of several predicted
auditory attributes that characterize the tested sounds. On that second corpus however, sorting data was found less
appropriate than pairwise data to provide interpretable continuous dimensions.