Three different methods of data collection in which subjects judged proximity between object pairs were examined. One method required subjects to partition objects into homogeneous subsets; the second entailed rating object pairs on a similarity-dissimilarity continuum; and the third involved comparing interobject proximities to a fixed standard. The three types of proximities were analyzed by the nonmetric multidimensional scaling procedure, and subsequent multidimensional representations were compared for accuracy to a criterion or true multidimensional configuration of the same objects. Significant differences in accuracy were found among the three methods, presumably due to differences in the extent to which subjects were able to describe their perceptions under the various methods. (Author/RC)