Department of Educational Psychology, School of Education, Waseda University
1986 Volume 28 Issue 1 Pages 11-20
(compatible with EndNote, Reference Manager, ProCite, RefWorks)
(compatible with BibDesk, LaTeX)
Twelve pentagons, which were each variations of a standard pentagon, were rank ordered by 37 university students according to their similarity to the standard. The data were analyzed by a nonmetric maximum likelihood scaling method for rank order data. We proposed various assumptions about the manners in which physical difference measures between two stimuli were defined. The physical difference measures defined were each assumed to be linearly related to the perceived dissimilarity. Fitting this linear function model to the rank order data has revealed (a) that the measure defined on the distance between a pair of corresponding vertices of two pentagons, using the difference limen as a unit, after translating one stimulus pattern so that its centroid may coincide with that of the other, works best;(b) that the linear function of this measure is a fairly good account of the data, but that it is not good enough to explain all meaningful variations in the data.