Early-holistic processing or dimensional similarity?

Abstract

In this commentary, we consider current evidence for the theory that multidimensional stimuli are initially processed as unanalyzed wholes. The theory appears unable to explain observers' behavior in various speeded and unspeeded tasks: (a) Perceivers can differentiate tiny changes made between confusable dimensions; (b) the fits of distance metrics to rating data appear to be largely under the control of optional processes; and (c) direct predictions from early-holistic models, as instantiated in the Euclidean hypothesis, are disconfirmed in tests using the rotation paradigm. In contrast, evidence has mounted to support the view that perception relies on a set of primary dimensions, processed within context-bound constraints. D. G. Kemler Nelson (1993) has been critical of our approach; we respond to each of her objections. Kemler Nelson (1993) summarizes the traditional view of how perceptual dimensions interact, what we call the earlyholistic theory. According to this theory, stimuli formed from interacting dimensions (e.g., colors varying in saturation, brightness, and hue) are first perceived holistically, as if they are unanalyzed points in a dimensionless space and then, with cognitive effort or processing time, are analyzed into their dimensional components. Early-holistic theory has been a fertile approach to the study of multidimensional perception. Over the past 20 years, Kemler Nelson and her colleagues have assembled a plethora of data and interpreted all of it through this system. In this short article, we are obviously unable to address this large body of data in detail. Nonetheless, by considering some of the main features of Kemler Nelson's (1993) summary we hope to raise substantive doubts about the continued usefulness of the theory. Early-holistic models are predicated on the logic of converging operations, which considers no single outcome to be decisive in accepting or rejecting a theory (Garner, 1974; Garner, Hake, & Eriksen, 1956). It should come as no surprise, therefore, that it is possible to question individual findings and interpretations even if the entire theoretical approach were sound. Nonetheless, we believe that the doubts we raise are so fundamental that they enervate the actual structure of the theory.