Elsevier

Cognition

Volume 65, Issues 2–3, January 1998, Pages 167-196
Cognition

Alternative strategies of categorization

https://doi.org/10.1016/S0010-0277(97)00043-7Get rights and content

Abstract

Psychological studies of categorization often assume that all concepts are of the same general kind, and are operated on by the same kind of categorization process. In this paper, we argue against this unitary view, and for the existence of qualitatively different categorization processes. In particular, we focus on the distinction between categorizing an item by: (a) applying a category-defining rule to the item vs. (b) determining the similarity of that item to remembered exemplars of a category. We begin by characterizing rule application and similarity computations as strategies of categorization. Next, we review experimental studies that have used artificial categories and shown that differences in instructions or time pressure can lead to either rule-based categorization or similarity-based categorization. Then we consider studies that have used natural concepts and again demonstrated that categorization can be done by either rule application or similarity calculations. Lastly, we take up evidence from cognitive neuroscience relevant to the rule vs. similarity issue. There is some indirect evidence from brain-damaged patients for neurological differences between categorization based on rules vs. that based on similarity (with the former involving frontal regions, and the latter relying more on posterior areas). For more direct evidence, we present the results of a recent neuroimaging experiment, which indicates that different neural circuits are involved when people categorize items on the basis of a rule as compared with when they categorize the same items on the basis of similarity.

Introduction

Since the beginning of the experimental study of categorization in psychology (Hull, 1920), there has been a tendency to assume that all acts of categorization are accomplished by the same means. In the seminal studies of Bruner et al. (1956), people were assumed to rely primarily on rules when categorizing novel items; in the current models of Estes (1994)and e.g. Nosofsky (1992), people are assumed to categorize new objects solely on the basis of their similarity to remembered exemplars of known categories; and in the discussions by Keil (1989)and Rips (1989), categorization inevitably comes down to applying a `theory' of the category. The alternative to such unitary views, of course, is that there are multiple strategies of categorization. This issue of unitary vs. multiple strategies is of foundational importance. If there are multiple strategies or procedures but we act as though there is only one, then results from different situations will no doubt conflict, with few or no generalizations emerging and no true accumulation of research findings1.

Though much of the research motivated by the various unitary views is of great importance, it is not difficult to devise counterexamples to the idea of a unitary view. One can generate pairs of extreme situations, such that just about everyone would agree that categorization is based on one strategy in one case and on a different procedure in the other case. Thus, if a person has to categorize two-digit numbers as odd or even, presumably all researchers would agree that the categorizer does it by applying the rule of `Divisible by 2 or not' (Armstrong et al., 1983); but if that same person has to categorize novel people with respect to whether they are as friendly as the neighbors on their block, presumably most would agree that the categorizer now relies on memories of his or her neighbors (Kahneman and Miller, 1986). Note, however, an important aspect of this counterexample—the different putative categorization procedures are applied to different kinds of categories, where one category is part of a rule-based formal system and the other one is completely ad hoc. If we ask instead whether there are multiple categorization procedures that are routinely applied to the same categories, then the issue of unitary versus multiple strategies is very much alive. It is this more stringent question that is the subject of the present paper.

Another aspect of the issue of one or many procedures is whether the multiple procedures are qualitatively different from one another. Some have argued for multiple categorization strategies that are qualitatively similar. For example, Nosofsky et al. (1994)have proposed a rule-plus-exception model, in which the representations used by the two procedures—rules and stored exemplars—differ quantitatively rather than qualitatively (e.g. the representation of a simple rule specifies a single attribute value, whereas the representation of a remembered exemplar might specify two or three attribute values). The obvious alternative is that there are qualitatively different procedures, that is, procedures that contain different processes.

In this article we argue that people can apply multiple procedures of categorization to the same items, perhaps even simultaneously, and that the procedures are qualitatively different from one another. Our specific agenda for the remainder of the paper is as follows. In the next, or second section we select two categorization procedures for examination—applying a rule vs. determining similarity to remembered examples—and characterize each procedure in more detail. In the third section we review evidence from cognitive studies that supports the claim that the two procedures of interest are applied to the very same categories. This review will accomplish two goals. Firstly, it will integrate findings from experiments that have used artificial materials with results from studies that have employed natural categories, showing striking convergence between two literatures that have heretofore been kept separate. Secondly, our review will reveal some of the processes that comprise the two procedures of interest, which will show that the procedures are indeed qualitatively different. In the fourth section we consider a different kind of evidence for a qualitative difference between the procedures of interest. Specifically, we will review results from neuropsychology and neuroimaging experiments which indicate that the procedures at issue are mediated by different neural structures. The fifth and final section summarizes our main points, and notes some related research.

Section snippets

The general distinction

It is time to be more precise about possible categorization procedures. A review of the literature on concepts and categorization (Smith and Medin, 1981) suggests at least three distinct procedures. In deciding whether a test object belongs to a particular category, one may:

  • 1.

    Determine whether the test object fits a rule that defines the category (the rule specifies the necessary and sufficient conditions for category membership);

  • 2.

    Determine the similarity of the test object to remembered exemplars

Cognitive studies of rule application vs. exemplar similarity

The cognitive experiments that have dealt with the issues of interest divide into two sets, depending on whether they have employed artificial or natural categories in their research. In what follows, we consider these two sets in turn.

Neuropsychological and neuroimaging studies of rule application vs. exemplar similarity

Another way to determine whether rule application and exemplar similarity are qualitatively different strategies is to ascertain whether they are implemented by different neural structures in the brain. In what follows, first we consider some indirect evidence from neuropsychology (i.e. the study of selective deficits due to brain damage), and then present some direct evidence from a neuroimaging experiment from our laboratory.

Summary

Studies with artificial categories provide evidence that rule application and exemplar similarity are qualitatively different categorization procedures (Allen and Brooks, 1991; Regehr and Brooks, 1993). The most straightforward piece of evidence is that subjects instructed to use a rule differ in predictable ways in their dominant categorization of certain items from subjects instructed to rely on their memories. Other experiments with artificial categories provide further dissociations between

Acknowledgements

Some of the research reported in this paper was supported in part by a grant from the Office of Naval Research and in part by a grant from the National Institute on Aging.

References (58)

  • S.W. Allen et al.

    Specializing the operation of an explicit rule

    Journal of Experimental Psychology: General

    (1991)
  • Awh, E., Smith, E.E., Jonides, J., 1995. Human rehearsal processes and the frontal lobes: PET Evidence. In: Grafman,...
  • L.R. Brooks et al.

    Role of specific similarity in a medical diagnostic task

    Journal of Experimental Psychology: General

    (1991)
  • Bruner, J.S., Goodnow, J., Austin, G.A. (1956). A study of thinking, Wiley, New York,...
  • M.C. Bushnell et al.

    Behavioral evidence of visual response in monkey cerebral cortex. I. Modulation in posterior parietal cortex related to selective visual attention

    Journal of Neurophysiology

    (1981)
  • J.D. Cohen et al.

    Activation of prefrontal cortex in a non-spatial working memory task with functional MRI

    Human Brain Mapping

    (1994)
  • M. D'Esposito et al.

    The neural basis of the central executive system of working memory

    Nature

    (1995)
  • Diamond, A., 1988. The development and neural bases of memory functions as indexed by the AB and delayed response tasks...
  • Estes, W.K., 1994. Classification and cognition, Oxford University Press, New York,...
  • Farah, M.J., 1990. Visual agnosia: disorders of object recognition and what they tell us about normal vision, MIT...
  • M.J. Farah et al.

    A computational model of semantic memory impairment: Modality specificity and emergent category specificity

    Journal of Experimental Psychology: General

    (1991)
  • Fuster, J.M., 1995. Memory in the Cerebral Cortex, MIT Press Cambridge,...
  • Hull, C.L., 1920. Quantitative aspects of the evolution of concepts, an experimental study. Princeton, NJ:...
  • J.J. Jaeger et al.

    A positron emission tomography study of regular and irregular verb morphology in English

    Language

    (1996)
  • D. Kahneman et al.

    Norm theory: comparing reality to its alternatives

    Psychological Review

    (1986)
  • Keil, F.C., 1989. Concepts, kinds, and cognitive development, MIT Press, Cambridge,...
  • J.A. Kolodny

    Memory processes in classification learning: An investigation of amnesic performance in categorization of dot patterns and artistic styles

    Psychological Science

    (1994)
  • B.J. Knowlton et al.

    Intact artificial grammar learning in amnesia: dissociation of classification learning and explicit memory for specific instances

    Psychological Science

    (1992)
  • S.M. Kosslyn et al.

    Visual mental imagery activates topographically organized visual cortex: PET investigations

    Journal of Cognitive Neuroscience

    (1993)
  • Cited by (241)

    • How does the brain learn environmental structure? Ten core principles for understanding the neurocognitive mechanisms of statistical learning

      2020, Neuroscience and Biobehavioral Reviews
      Citation Excerpt :

      The attention-dependent learning system shares similarities with Baars (1988; 2005) global workspace theory of consciousness, in which consciousness is construed as a limited capacity attentional spotlight that “enables access between brain functions that are otherwise separate” (Baars, 2005, p.46). What determines the mode of learning (explicit vs. implicit) likely depends at least in part on the type of input to be learned; some types of structures appear to require attention to adequately process and encode the patterns, such as nonadjacent dependencies (de Diego-Balaguer et al., 2016d), global patterns (Bekinschtein et al., 2009), cross-modal dependencies (Keele et al., 2003), and rule-based processing (Hendricks et al., 2013; Smith et al., 1998). Other factors that may affect the involvement of automatic versus attention-dependent mechanisms include whether learning is assessed through the use of direct/explicit judgments versus indirect measures such as reaction times (Batterink et al., 2015) and whether learning requires generalization or transfer to new stimulus sets (Hendricks et al., 2013).

    • The effect of category learning on attentional modulation of visual cortex

      2017, Neuropsychologia
      Citation Excerpt :

      The remaining 80 creatures were composed of three dimensions of color (red or purple), global shape (horse or fish), and markings (spots or stripes); the values of these dimensions determined category membership. Binary dimension values like these have been used in many past studies of categorization (Allen et al., 1991; Folstein et al., 2011; Hammer et al., 2010; Medin and Schaffer, 1978; Nosofsky et al., 1994; Regehr and Brooks, 1993; Shepard et al., 1961; Smith et al., 1998; Smith and Minda, 1998). Stimuli were created by combing the values of color, shape, and markings factorially, resulting in eight equally probable combinations, or cases.

    • When more is less: Limited consideration

      2017, Journal of Economic Theory
    View all citing articles on Scopus
    View full text