Perceptual and Decisional Separability
W. Todd Maddox
University of California at Santa Barbara
A fundamental issue in human perception is to determine how different stimulus dimensions interact during perceptual processing (e.g., Ashby & Townsend, 1986; Garner, 1970, 1974). Two main classes of stimulus dimensions have been identified. These have been variously termed unitary and analyzable ( Shepard, 1964), integral and nonintegral ( Lockhead, 1966), or integral and separable ( Garner, 1974).1 This chapter adopts Garner ( 1974) terminology. The traditional distinction between integral and separable stimulus dimensions emerges from the results of a series of experimental tasks ( Garner, 1974; Garner, Hake, & Eriksen, 1956) that purport to assess the extent to which a stimulus is processed by its dimensional values as opposed to as a unitary whole. If the results of these tasks suggest independent processing by dimensions, the stimulus dimensions are termed separable, if the results indicate processing as an unanalyzable whole, the dimensions are termed integral (e.g., Garner, 1974; Lockhead, 1966; however, see Cheng & Pachella, 1984, and J. D. Smith & Kemler-Nelson, 1984, for alternative processing theories; see also L. B. Smith, 1989, for a developmental processing theory). Prototypical separable dimensions are hue and shape (e.g., Garner, 1977). Prototypical integral dimensions are hue and brightness (e.g., Garner & Felfoldy, 1970; Hyman & Well, 1968).
This chapter examines four operational tests of separability: (a) the Filtering Task, (b) the Redundancy Task, (c) Direct Dissimilarity Scaling, and (d) the____________________