Overall similarity and the identification of separable-dimension stimuli: A choice model analysis

Abstract

Two subjects identified a set of perceptually confusable stimuli varying along two continuous, separable dimensions. The data were analyzed using various restricted versions of the similarity choice model (Luce, 1963; Shepard, 1957). In the main version, the multidimensional scaling approach to modeling similarity was applied and yielded excellent quantitative accounts of the identification data. A surprising result was that the Euclidean metric provided a far better description of psychological distance relationships than the city-block metric, a finding that contrasts with virtually all previous conclusions regarding the appropriate Minkowski r-metric for separabledimension stimuli. Possible reasons for the discrepancy between the present findings and previous conclusions are considered in the General Discussion section, with Shepard's (1964) seminal study examined in depth. In an initial theoretical section, it is also shown that, for factorial stimulus sets, a choice model in which subjects identify values along each component dimension independently is a special case of the full choice model in which overall similarity relationships among stimulus wholes determine identification performance. © 1985 Psychonomic Society, Inc.