Subjects evaluated how well they liked each one of 38 short excerpts of Western music and also judged how well each excerpt was described by 23 adjectives. How well an excerpt was liked was negatively correlated with the use of the adjectives ‘unpleasant’, ‘complex’, ‘tense’, and ‘dissonant’. The use of the adjectives ‘melodic’, ‘pleasant’, ‘sentimental’, and ‘familiar’ were positively related to how well an excerpt was liked. The correlations between the preference judgments of different excerpts were taken as a measure of similarity between the excerpts. This measure of similarity was used in a multidimensional scaling analysis with the purpose of identifying dimension that may determine preferences for music. In the six dimensional space generated (stress value was .255) coordinates on three of the dimensions could be predicted, in part, by the use of the adjectives ‘sentimental’, ‘fast’, and a combination of ‘high pitched’, ‘calm’, and ‘sad’, respectively. Thus, some clues to the factors underlying musical preferences were obtained. Although a large number of dimensions were necessary and all of them could not be interpreted meaningfully in the present piece of research, this method may help discover a way of conceptualizing musical preferences with a more careful selection of excerpts and more detailed assessment of their qualities.