A test of independence of dimensions in multidimensional scaling

Abstract

Models for multidimensional scaling use metric spaces with additive difference metrics. Two important properties of additive difference metrics are decomposability and intradimensional subtractivity. A prediction was derived from these properties and tested experimentally. Rectangles varying in area and shape served as stimuli. Dissimilarity judgments were obtained by both rating and pair comparison procedures. The assumptions of the model were violated by most of the Ss. Apparently this violation was due to an interaction between the two dimensions. © 1971 Psychonomic Journals, Inc.