Sequence effects in categorization of simple perceptual stimuli.
Neil Stewart, Gordon D. A. Brown, Nick Chater · Journal of Experimental Psychology Learning Memory and Cognition · 2002
Categorization research typically assumes that the cognitive system has access to a (more or less noisy) representation of the absolute magnitudes of the properties of stimuli, and that this information is used in reaching a categorization decision.However, research on identification of simple perceptual stimuli suggests that people have poor representations of absolute magnitude information and that judgments about absolute magnitude are strongly influenced by preceding material.The experiments presented here show strong sequence effects in categorization tasks.Classification of a borderline stimulus was more accurate when preceded by a distant member of the opposite category than by a distant member of the same category.It is argued that this category contrast effect cannot be accounted for by extant exemplar or decision-bound models.The effect suggests the use of relative magnitude information in categorization.A memory and contrast model illustrates how relative magnitude information may be used in categorization. Sequence Effects 3 Sequence Effects in Categorization of Simple Perceptual StimuliCategorization models are often divided into two general classes, each including a wide range of specific accounts: parametric Thurstonian decision-bound models (e.g., Ashby & Townsend, 1986), and non-parametric exemplar models (e.g., Medin & Schaffer, 1978;Nosofsky, 1986).However, all extant models of categorization assume that items can be represented in terms of their (more or less noisy) absolute location in a multidimensional space.This absolute location information is then assumed to be used in the decision process (either directly, as in exemplar models, or indirectly, in relation to decision bounds).Thus two key assumptions are (a) that the absolute location of a stimulus in multidimensional space, though possibly noisy, is available when a categorization or identification decision is made, and (b) that absolute location information provides the sole basis for categorization decisions.Formal models of categorization and identification based on these assumptions have a long history of successful application to a wide range of experimental paradigms (see Estes, 1994, for a review).Here we present evidence that relative magnitude information, derived from comparison of the current stimulus to recent stimuli, is also used in categorization. Difficulty in Determining Absolute MagnitudesRegarding the first assumption above, participants often have difficulty making accurate estimates of the absolute values of stimuli along simple perceptual dimensions, particularly in the absence of contextual information.For example, in a series of classic experiments by Garner (1954), participants' judgments of whether comparison tones were more or less than half as loud as a given reference tone were completely determined by the range of the comparison tones (see also Helson, 1964).Such a context effect should not be evident if participants did have access to absolute magnitude information.Baird, Green, and Luce (1980) demonstrated that two-thirds of the variability in loudness estimates was explained by the variability in the previous estimate when loudnesses were similar,