With loudness-equalized car interior noise as sound stimuli, the work focused on apphcabihty of nonmetric multidimensional scaling （NMDS） to sound quality research. It is presented that NMDS is an effective tool for subjective assessment research of sound quality, when listeners are clustered through correlation between hsteners. With this key clustering process, the perceptual structure of car interior noise with strong consistency is revealed. The results shown that： for car interior noise, ‘preference＇ can be regarded as a main sub-dimension in ‘similarity＇ space and listeners could be divided into two groups; For both groups of listeners, the perceptual ‘preference＇ can be characterized as one-dimensional descriptor, with one group has a positive relation to ‘low-frequency＇, while another group has a negative relation to ‘low- frequency＇.
