JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

2002, 78, 237–248

NUMBER

3 (NOVEMBER)

CATEGORIZATION, CONCEPT LEARNING, AND BEHAVIOR ANALYSIS: AN INTRODUCTION T HOMAS R. Z ENTALL , M ARK G ALIZIO ,
AND

T HOMAS S. C RITCHFIELD

UNIVERSIT Y OF KENTUCKY, UNIVERSIT Y OF NORTH CAROLINA AT WILMINGTON, AND ILLINOIS STATE UNIVERSIT Y

Categorization and concept learning encompass some of the most important aspects of behavior, but historically they have not been central topics in the experimental analysis of behavior. To introduce this special issue of the Journal of the Experimental Analysis of Behavior (JEAB), we define key terms; distinguish between the study of concepts and the study of concept learning; describe three types of concept learning characterized by the stimulus classes they yield; and briefly identify several other themes (e.g., quantitative modeling and ties to language) that appear in the literature. As the special issue demonstrates, a surprising amount and diversity of work is being conducted that either represents a behavior-analytic perspective or can inform or constructively challenge this perspective. Key words: categorization, concept learning, stimulus class, function transfer

Categorization is not a matter to be taken lightly. There is nothing more basic than categorization to our thought, perception, action, and speech. (Lakoff, 1987, p. 5) Concepts give our world stability. They capture the notion that many objects or events are alike in some important respects, and hence can be thought about and responded to in ways already mastered. Concepts also allow us to go beyond the information given; for once we have assigned an entity to a class . . . we can infer some of its . . . attributes. (Smith & Medin, p. 1)

There is, perhaps, no larger or more diverse literature within experimental psychology than that focused on categorization and concept learning. This topic is, to the casual observer, most directly associated with human cognitive psychology, within which the largest volume of research and theory building has
Order of authorship for this article was determined by random drawing. We are grateful to the numerous reviewers, who met the dual challenge of evaluating manuscripts according to JEAB’s usual expectations while embracing the conceptual and methodological diversity inherent in the topic, and to the authors of the articles contained herein, many of whom invested extra effort in tailoring their work to an unfamiliar audience. Address correspondence to Tom Zentall at the Department of Psychology, University of Kentucky, Lexington, Kentucky 40506 (e-mail: zentall@pop.uky.edu); Mark Galizio at the Department of Psychology, University of North Carolina at Wilmington, Wilmington, North Carolina 28403 (e-mail: galizio@uncwil.edu); or Tom Critchfield at the Department of Psychology, Illinois State University, Normal, Illinois 61704 (e-mail: tscritc@ilstu. edu).

taken place. Cognitive psychologists Laurence and Margolis (1999), in a book reviewed in the present issue, minced no words about the association: ‘‘Concepts are the most fundamental constructs in theories of the mind’’ (p. 3). Yet scholars from many research communities have struggled to come to grips with the complex repertoires that the topic encompasses. The heterogeneity of this research area is evident in the absence of a consensus definition of the term concept (see Palmer, this issue; Wasserman & Bhatt, 1992). Writers tend to stress the importance of concepts rather than specifying their defining features—perhaps, as Palmer speculates in his review of Margolis and Laurence (1999), ‘‘regarding the term as too familiar to need definition’’ (p. 598). Nevertheless, an introduction to this special issue demands at least an attempt to define its subject matter. Typically in cognitive psychology, categorization is regarded as a process of determining what things ‘‘belong together,’’ and a category is a group or class of stimuli or events that so cohere. A concept is thought to be knowledge that facilitates the categorization process (e.g., Barsalou, 1991, 1992). Consistent with the representational style of much cognitive theorizing, conceptual knowledge is often portrayed as existing independently of any particular behavior–environment relation. This is assumed partly because, once a categorization repertoire is in place, an individual may be able to categorize both previously en-

g. 1992). Herrnstein. 1987. they are clearly among the most interesting behavioral phenomena available for study. as an instructive case. The analytical challenge becomes more daunting. It also includes birds that are not dangerous. 155). 1978). it is possible to examine the circumstances under which we speak of conceptualization— that is. Reservations about the approach are warranted. Reeve. 1945) frequently espoused. For example. Thus. departure by suggesting that ‘‘when a group of objects gets the same response. brightness). see Deacon. novel stimuli to which the fruits of this history may transfer. Regardless of whether conceptual repertoires are uniquely human (many articles in the present issue suggest that they are not). we will follow that practice here. this issue. the balan category of the Australian aboriginal language Dyirbal.g. such as those that apparently unite many categories.. Rather than speculating about the status of hypothetical knowledge structures. and whether hierarchical relations apply to the category or instances within it (e. and so forth. Rosch.238 THOMAS R. support the spontaneous transfer of function from one member to another. the goal of many studies in cognitive psychology is to map the knowledge that humans presumably apply in already established patterns of categorization. what individuals are doing when they are said to behave conceptually. 1990). Such classes include stimuli involved in an explicit learning history plus. which of these entities are considered to be more or less typical of the category. A category is a class of stimuli that occasion common responses in a given context.g. size. For example. fire. ZENTALL et al. Thus. 1999. share members that may belong to different categories under different circumstances. this focus on knowledge may discourage attention to the role of experience in creating and maintaining conceptual behavior (Astley & Wasserman. as well as exceptional animals such as the platypus. merge and fracture. Because of their richness. 1997). Behavior analysts are likely to regard as foreign this practice of describing terminal performance without examining the necessary and sufficient conditions for its emergence. Many writers use the terms category and stimulus class more or less interchangeably. Barsalou (1992) has noted a tendency for competing cognitive theories to make similar predictions and to account equally well for data obtained from human subjects. AN OPERATIONAL APPROACH TO CONCEPT LEARNING From a behavior-analytic perspective.g. and how they came to behave in that way. categorization may be said to incorporate a pattern of systematic differential responding to classes of nonidentical.. and echidna’’ (Lakoff. as the essence of what it means to be human (e. countered stimuli and novel events. the present topic provides an opportunity to apply the operational analysis of psychological terms that Skinner (e. p. Some writers have gone so far as to label the capacity to glean abstract relations. 154). et al. their accounts can be difficult to distinguish empirically. 1996). Consider. Perhaps more important for present purposes. When the stimuli within and between categories vary along relatively simple dimensions (e.g. which ‘‘includes women. for a brief synopsis of some relevant theories). structured interviewing and other techniques may be used to determine what entities people include in a category like birds. Wasserman & Bhatt. stimuli (see Fields. and dangerous things. 5). see Palmer. though potentially discriminable. when they form a class the members of which are reacted to similarly. potentially. categorization is readily conceived in the same terms as stimulus discrimination and generalization.. ‘‘Generalization within classes and discrimination between classes— this is the essence of concepts’’ (Keller & Schoenfeld.. Although cognitive psychologists expend much energy debating the structure and contents of the knowledge that is assumed to underpin categorization and the means by which it is compared to new perceptual experiences (Laurence & Margolis. of course. Keller and Schoenfeld (1950) specified this very point of . and adaptability. as category membership is determined more complexly (e. bandicoot. p. this issue. suggesting to some observers that the latter are recognized via comparison to general information represented in memory. generativity... we speak of a concept’’ (p. 1950. wavelength. Any plausible account must also explain how categories add and lose members.

and to demonstrate that JEAB can be the locus for a productive exchange on a rich and compelling topic. & Robbins. TY PES OF CONCEPTS The work described in the present issue draws from multiple scholarly communities. Fortunately. provide a slippery foundation for an experimental analysis. The overarching purpose of this special issue. Wixted & Gaitan. Such paradigms. p. studied experimentally (e. 1977. with knowledge operationalized more or less as above. the authors of most articles in the present issue have. to 239 examples and nonexamples of categor y members (Barsalou. 1992. with feedback. from concepts to concept learning. The goal is ‘‘to establish high degrees of control over category knowledge’’ (Barsalou.g. Although Keller and Schoenfeld (1950) provided the outlines of a behavioral analysis of conceptual behavior more than half a century ago. Because the literature on categorization and concept learning is so vast. scholars outside behavior analysis have scarcely noticed. Although the scientific community with which JEAB is most associated has not made categorization and concept learning the centerpiece of its empirical and theoretical contributions. we seek to provide a cross section of contemporary research efforts and illustrate something of the breadth and value of this area of study. a remarkable amount of work is being conducted that expressly represents a behavior-analytic perspective. But taking issue with the theoretical perspective that has dominated the concepts and categories literature does not render the relevant behavioral phenomena any less provocative. or can constructively challenge this perspective. Palmer catalogues some key points of contention between behavior-analytic and cognitive views). Almost universally. see Gagne ´ . Unfortunately. these efforts require subjects to respond. this issue. both to plumb the abilities of specific species and to assess interspecies similarities and differences. To place this work into a broad context. of course. and what experiences and circumstances make these repertoires possible. One community of investigators examines categorization and concept learning in nonhumans. Many research traditions (including human cognitive psychology. In search of more secure theoretical footing. This subtle shift in emphasis. The preceding insights are not. and they are unlikely to do so in the absence of persuasive empirical evidence (Schwartz. but as a linguistic surrogate for learning histories that can be operationalized and. this issue) include attempts to create artificial categories uniting experimenter-selected rather than everyday stimuli. in many cases. below we provide some further orienting remarks about the area. Wasserman. is to bring together the best current work of behavioral researchers who seek to determine which species are capable of which conceptual repertoires. A key goal of this special issue is to promote a cross propagation of empirical and theoretical insights. The needed research requires a clearly defined subject matter. 2002). For these readers. We anticipate that some JEAB readers will have only passing acquaintance with this research area. therefore. which are comfortingly reminiscent of discrimination learning procedures. ‘‘knowledge’’ can be regarded.. in press). see Maddox. Skinner.CATEGORIZATION AND CONCEPT LEARNING they invite a thorough experimental and theoretical analysis. unique to behavior analysis. this issue. 1991). categories and concepts have been addressed only sporadically within behavior analysis. Moreover. defined largely in terms of abstract knowledge. can inform this perspective. A second community of researchers has been concerned with the for- . 31). even those who work in this area may be unfamiliar with the efforts of other investigators. recast the notion of knowledge in terms of necessary and sufficient conditions for conceptual behavior. in some fashion. diverts attention at least partly away from taxonomizing knowledge and toward identifying the functional relations between behavior and environment that provide the basis for conceptual behavior. This neglect may be understood partly as a rejection of the cognitive theoretical worldview that has helped to define the topic (in the present issue. and we have already suggested that concepts. objection through silence persuades no one. Maddox. form the basis of most of the research described in the present issue and constitute an important point of contact between diverse communities of researchers. not as an entity.

experience with both positive and negative stimuli appears to be necessary.g. Because stimuli often are said to be similar on the circular basis that subjects respond similarly to them. background. 1965). such as a person.240 THOMAS R.g. Loveland. report that discrimination training between categories is needed for pigeons to respond differentially to untrained members of the training set versus untrained members of a different training set (i. in the acquisition of simple discriminations. Although most laboratory concept tasks involve static stimuli. attempt to identify simple stimulus features that guide categorization. from a subject’s perspective. Sutton and Roberts. category membership may be said to be determined. research with nonhumans involves the discrimination of photographs or drawings that contain a particular type of object. Finally. Finally. 1976). Much of their work bears on questions about categorization and concept learning. Newman & Baron. a large community of cognitive psychologists seeks to understand categories and concepts in humans. Fremouw. In perceptual concepts. In relational concepts. Thus. training animals to respond to a single set of stimuli may be insufficient to ensure differential responding among multiple sets. as Vonk and MacDonald (this issue) note. in associative concepts. by what the experimenter recognizes as an image of a person). category distinctions may not be made until members of at least two categories are compared). Herbranson. subjects may show a high degree of class-consistent responding to novel exemplars of the stimulus sets (Herrnstein.. the pictures in the positive set (to which responding is reinforced) and in the negative set (to which responding is extinguished) vary in terms of perspective. Even when complex stimuli are employed. number of relevant items portrayed. In the present issue. establishing a response in the presence of one stimulus does not ensure differential responding when other stimuli are introduced (e. For example. stimuli are grouped primarily on the basis of shared physical features. many studies employ stimuli that are not defined by a few simple features. Following this kind of training. In such cases. below we emphasize three broad types of relations that appear to unite events within a category. it is more difficult to think in terms of simple stimulus control processes.e. similarity is construed here as relatively little separation along a well-defined physical dimension. and Shimp (this issue) show that pigeons can accurately categorize dynamic properties such as stimulus movement and direction. experience with both ‘‘examples’’ and ‘‘nonexamples’’ appears to be essential. whose conceptual abilities have been studied infrequently. or a common consequence with which they are correlated). imental contingencies. Perceptual Concepts The classes of stimuli that are united in perceptual concepts may be said. mation of stimulus classes (including equivalence classes) primarily in humans. Vonk and MacDonald extend this type of procedure to gorillas. a common response that they engender. stimulus variation within a picture set may be as great as the variation between the sets and. it seems unlikely that each separate picture becomes an independent discriminative stimulus. when many pictures are used (Herrnstein & Loveland. and so forth. 1964. concept learning mirrors discrimination learning in tantalizing ways. Straightforward examples can be found in the present issue in articles by Maddox (line length) and by Fields and colleagues (gradations between two photographs as altered by commercial morphing software).. from those that do not contain the object. used several hundred). in the present issue. Perhaps analogously. In the interest of modeling categorization as it occurs in the natural environment. To avoid unnecessary entanglement in theoretical debates between and within communities. and apparently unsuccessful. in the exper- . color. however. In studies like that of Herrnstein and Loveland (1964). ZENTALL et al. it is not the physical features of stimuli per se but the relations among these features that are grouped. and some of it addresses these topics explicitly. to bear physical similarity to one another. For example. In such cases. & Cable.. Herrnstein and Loveland (1964) taught pigeons to discriminate photographs containing a person from photographs that did not (in such a case. stimuli are grouped on the basis of shared function (e. One noteworthy feature of the Vonk and MacDonald study is an extensive.

Sands..CATEGORIZATION AND CONCEPT LEARNING Relational Concepts Compared to perceptual concept learning. Typically. because arbitrary relations were thought to provide a tool for studying symbolic processes relevant to language and cognition (e. In the present issue. a choice of Comparison Stimulus B1 is reinforced.. the correct comparison is conditional on the particular sample presented. Another set of relations could then reinforce a choice of C1 given Sample B1 and C2 given Sample B2. the subject chooses Comparison C1). For example. With this procedure. given Sample Stimulus A1. arbitrary match-tosample training is used to establish at least two conditional discriminations involving physically unrelated stimuli. Rivera. 1988). researchers of animal learning and cognition have been investigating similar phenomena in nonhumans.g. or an object depicted in a photograph. naturalistic drawing. Perhaps the strongest evidence of same– different learning by pigeons comes from research in which the simultaneous display of similar objects is discriminated from displays in which all of the objects are clearly different from each other. 2000). 1997). Wasserman. Associative Concepts In associative concept learning. & Delius. the sample consists of a pair of objects that are either identical or different. relational concept learning makes use of more abstract properties of the stimuli. in which absolute properties of the stimuli may guide responding. choosing B2 is reinforced. Also in the present issue. This interest arose.g. Chimpanzees acquire such an abstract conditional discrimination and can transfer it to novel exemplars (Thompson. Same–different concept learning can be studied using conditional discriminations in which subjects must respond to the comparison that matches the sample. 1994) seminal work on stimulus equivalence has generated enormous interest. & Boysen.g. Cook. the 241 reinforced choice would be BB. in the behavior-analytic community.. when trained with two novel hues (Zentall & Hogan. but rather cohere because of shared functional properties. the subject chooses Comparison A1). 1994). . Even better transfer can be found when pigeons are trained on an identity task with a large number of stimuli (Wright. given A1 as the sample. Sidman designated these reflexivity (e. Cook. Sidman’s (e. For example. symmetry (e. For example. in the study of non-similarity-based stimulus classes. and given Sample A2. if shown Sample AA and given a choice between BB and CD. 1997). In the present issue. One of the simplest and most studied relational concepts is same versus different. relative to an appropriate control. Thus... the stimuli within classes bear no obvious physical similarity to one another. Barnes-Holmes.g. Pigeons trained in this way learn to discriminate the training exemplars and also to discriminate novel stimuli involving the same relations as in training (e. and transitivity (e. There is also evidence that some species are capable of second-order same–different learning.. As research on stimulus equivalence has evolved into a major focus in the experimental analysis of human behavior. given B1 as the sample. three kinds of untrained performances often emerge. Katz. and the correct comparison consists of two stimuli that are different from the sample but that bear the same relation to each other as those in the sample. and Peissig report that detecting same and different arrays occurs very rapidly and thus does not appear to require the sequential comparison of the stimuli in the array. as defined by Sidman (e. the subject chooses Comparison A1).g. they show facilitated acquisition. 1978). in which the relation between two objects must be matched rather than the objects themselves. Oden. & Cavoto. in part. Roche. 1971. possibly analogous performances are shown in human ‘‘equivalence–equivalence responding’’ (Stewart. & Smeets). will illustrate one way that associative concept learning can be studied in the laboratory.g. After such training. On each trial. given A1 as the sample. feature..g. after pigeons have been trained to match two shapes. Young. A brief review of the basic methods used to study stimulus equivalence. geometric shape. Cook reports that the resulting same– different categories are general and can include stimuli that vary along different perceptual dimensions such as texture. Sidman. it is not the match between stimuli but the match between the stimulus–stimulus relations that must be learned.

Kastak. associative concepts develop fairly readily through one-to-many training. & Kastak. this issue. Steirn.. because research has suggested that not all of the formal properties of equivalence classes (as defined by Sidman & Tailby. 1995.. Over the past 25 years or so. 1997). then it should be harder to discriminate between those samples than between samples associated with different comparison selections. A variety of means exist to assess the associative relations that emerge in nonhumans following many-to-one training. 1982) are readily demonstrated in nonhumans. Lane & Critchfield. see Hayes. a small circle). recent years have seen a convergence of interests between equivalence researchers and researchers who study associative concepts in animal cognition (Zentall & Smeets. & Urcuioli.g. 2001). 1997. 1992. 1985). Nuefeld.. For example. Carrigan & Sidman.g. et al. One possible explanation of this failure emphasizes not a generic deficit but rather competing control by factors such as stimulus location.. Sidman. Saunders & Green. 1998. a red field and vertical lines) to the same comparison selection (e. and fracture. Arntzen & Holth. the demonstration of all three relations was considered to be the definition of an equivalence class (e. Reeve. In the present issue. & Urcuioli. Sherburne. Sidman & Tailby. symmetry. Although baseline matching transferred to novel stimulus locations. Common-response training. this issue. 1997. and transitivity without explicit reinforcement of these relations. Sidman. 1996). & Zentall. for example. Zentall.. a green field and horizontal lines) to another comparison selection (e. Several lines of evidence indicate that in nonhumans a relation develops between the samples associated with a common comparison (e. C. Pilgrim & Galizio. symmetrical relations may fail to emerge in nonhumans after training involving arbitrary stimuli (Sidman et al. Steirn. 1997). see Arntzen & Holth. and different samples (e. in which a common sample is associated with two or more comparison selections (e. Sherburne..g. 1993. 1982). 1991).g. Zentall. Fields. no evidence of the emergence of symmetry was obtained in any of their experiments. ZENTALL et al. in humans. 1982).. preconditions of class merger. For example. arrange the requisite common response by reinforcing the matching of two or more samples (e. Symmetrical responding was directly reinforced with some stimulus sets before testing for symmetry with new sets. 2000). Many more studies have shown functional substitutability . Surprisingly.g. leaving the distinctions between equivalence and functional classes blurred at best (see Sidman.g. Zentall. Some studies influenced by this convergence now suggest the capacity of certain nonhuman species to acquire equivalence relations (see Kastak & Schusterman..g. similar results have not been found with pigeons. Kirk. & Steirn. types of training and testing formats.g. see also McDaniel. To illustrate. If relations have formed between samples associated with the same comparison selection. and the stimulus control processes that promote or preclude class formation.. Some commentators have wondered whether stimulus equivalence is a uniquely human phenomenon (e.g. 1999.. expansion. we consider some of the procedures through which such functional classes have been examined. JacksonSmith. 1994.g. and further training relates Sample A to new Comparison D. Recent research suggests that this is indeed the case (Kaiser. Common-sample training. and. 1989). Sample B will be matched to Comparison D without additional training (Urcuioli. many studies have been directed toward evaluating the mechanics of equivalence class formation (e. historically. & Damico-Nettleton. & Willson-Morris.242 THOMAS R. a large circle). Lionello-DeNolf and Urcuioli describe an experiment in which pigeons were tested for symmetry after multiple sample-location training designed to reduce control by a particular location. Another approach involves comparing the discriminability of samples associated with a common comparison selection with that of samples associated with different comparison selections. just as is the case for humans (e. 2001). 1989). if Samples A and B share a common comparison selection... Humans have shown the emergence of reflexivity. of class members. Schusterman. Sherburne. As the preceding case illustrates. Stimuli can become united into a class by virtue of association with a common response. Many-to-one conditional discrimination procedures. It is worth noting that. Arntzen & Holth.

Reeve. and then reversed again. Their findings suggest that one-to-many training may interfere with generalized categorization in humans. the valence associated with each set was reversed. repeatedly. Zentall. Symmetry training. in spite of the fact that stimuli differed as much within sets as between sets in terms of physical similarity and reinforcement histories. In the present 243 issue. In each task. and Delius report that similar results can be obtained using simultaneous discriminations. That is. concept learning was evaluated through selection-based repertoires involving the matching of one class member to another. Following training. red and vertical) were reinforced with peas. budgerigars were . When animals learn to match. Clement. Mackay.CATEGORIZATION AND CONCEPT LEARNING even when conditions are otherwise similar to those under which associative concepts have been shown to develop during many-to-one training (Urcuioli & Zentall. in which the stimuli within a class spontaneously share whatever functions each has acquired separately (Dougher & Markham. and Stoddard (1989). can serve to form a stimulus class consisting of those stimuli..e. Lehr. Articles in the present issue indicate that if chimpanzees or sea lions are presented with a novel sample and one novel and one familiar comparison (i. following this training. 2000). Common-outcome training. Siemann.. B → C) in a conditional discrimination. Zentall. The use of simultaneous discriminations permits the presentation of novel pairings of the training stimuli. and Hogan (1982) trained pigeons on two identity-matching tasks. a comparison already associated with a different sample). to produce comparison stimuli during match-to-sample training. Kawashima. one with lines and the other with hues. Choice by exclusion. et al. In the present issue. 1993). Such repertoires have been described as part of a more general transfer of function (also called transformation of function or inheritance of meaning). the untrained relation A → C emerged as well. Recent research suggests that symmetry training. suggesting that these arbitrarily assigned stimuli had become two functional stimulus classes.. the number of trials needed for responding to follow suit decreased to only a few trials. although interspecies similarities may yet exist. For example.g. Kastak & Schusterman). McIlvane. Another procedure for developing stimulus classes is to treat a set of stimuli the same through a number of transformations (Vaughan. Manabe. Fields. Transfer of function. Serial reversals. correct choices of one of the comparisons (e. The asymmetry of these effects hints at differences in the way humans and other animals form these associative concepts. and Staddon (1995) described a procedure in which. Vaughan randomly assigned photographs of trees to two arbitrary sets. Stimuli can cohere into classes by way of their association with a common outcome. it is assumed that they learn to select the comparison based on its acquired association with the sample. In many of the preceding examples. Across reversals of stimulus sets. 1996. see Dube. Jagielo.. the positive stimulus from one training pair could be presented with the negative stimulus from a different training pair. they will avoid the familiar comparison (that with an experimentally defined function) and choose the novel (undefined) one. Maguire. A and B.g. the animals will tend to choose by exclusion (Beran & Washburn. For example. Jitsumori. Edwards. in which A → B and B → A relations are taught. if one stimulus was also associated with a new stimulus (e. Control procedures reveal that this is not simply a preference for novel comparisons but rather a general capacity to match undefined stimuli. Dymond & Rehfeldt. whereas correct choices of the other comparisons (e.g. 1988). pigeons showed positive transfer to trials in which hues and lines associated with a common outcome could be matched. report that the likelihood of humans acquiring a generalized catergorization repertoire (one that applies across several different mixed perceptual-associative categories) is positively related to the number of sample stimuli used during training and is negatively related to the number of comparison stimuli. Following training. green and horizontal) were reinforced with wheat. and Weaver (in press) found that. Such choices therefore suggest an indirect measure of concept learning. For analogous findings with humans. and then trained pigeons to respond to those in Set A but not to respond to those in Set B.

Markham and Markham explore some of the necessary and sufficient conditions for transfer of function in humans. and Randle. but the cognitive literature merits attention for at least two reasons. Gagne ´ proposes an experiential basis for this transfer of function to which readers may readily apply familiar behavior principles (e. who describe a required to make distinct vocal responses to each of two samples. construed somewhat differently. Second. Urcuioli et al.g. In the present issue. Several papers in the present issue incorporate Sidman’s framework (e. Barnes-Holmes. below).g. conditional relations and the matching law). also plays a role in some cognitive accounts. the circumstances that facilitate transfer of function are poorly understood (e. 2001)— both propose that the emergence of equivalence relations reflects previously established higher order operant behavior.. then paired shock with one member of a class. this issue).244 THOMAS R. this literature describes many empirical phenomena that any noncognitive theory must explain (e. Urcuioli et al. describe a successful attempt to replicate this effect in pigeons. stresses the importance of providing individuals with extended exposure to experimental manipulations and of evaluating individual-subject response functions. but without differential vocalization requirements. substituting high-rate versus low-rate pecking patterns for differential vocalization. trained to one stimulus. . Griffee & Dougher).. His quantitative models of perceptual concept learning draw heavily on signal-detection principles and the effects of consequences on behavior. Sidman’s (2000) stance is that equivalence relations represent a fundamental outcome of a reinforcement contingency such that all elements of the contingency (conditional and discriminative stimuli. and behavior analysts may be surprised to find that they share many assumptions with some of their cognitive counterparts. regardless of its theoretical underpinnings. ZENTALL et al. In the present issue. Two competing views—naming theor y (Horne & Lowe. The elicitation function appeared spontaneously in other members of the same class (Dougher. see Concepts and Language. but an operant response function. see Markman. OTHER ISSUES Theoretical Views Much research on categorization and concept learning is guided by cognitive theories that run counter to traditions in behavior analysis (Palmer. propagated through the putative class nevertheless. Horne. Differential vocalizations spontaneously transferred to the new stimuli.. Greenway. As Gagne ´ (this issue) suggests. and reinforcers) are posited to become equivalence class members. the prerequisite operant is a bidirectional relation between objects and the speaker-listener behavior (names) they occasion (note that naming. 1991). In the case of naming theory. Readers of research on stimulus relations (such as equivalence) in humans will discover the influence of three main theories. 1994). Maddox and Gagne ´ provide cases in point. 1996) and relational frame theory (Hayes. Markham. Barsalou. apparently on the basis of association with common comparison selections. The results provide fodder for interesting speculation about what exactly creates and defines stimulus classes. They describe a case in which stimulus class formation (on the basis of common respondent functions) apparently failed. new samples were associated with the same comparisons (creating many-to-one training). making it a conditioned stimulus for skin conductance responses indicating fear elicitation. and address the possibility that adventitious reinforcement might create spurious cases of function transfer. Harris. Transfer of function also has been a focus in many studies of stimulus equivalence. For example. In general. see Barnett & Ceci. Augustson. 1992. Later. & Roche. for instance. Fields et al.. Maddox. & Wulfert.g. First. An example of research conducted within the framework of naming theory is provided in present issue by Lowe. such transfer-of-function outcomes are reminiscent of the shared meaning evident in hybrid lexical concepts. In the present issue. offer an extended analysis of conditions under which such effects might arise. Dymond & Rehfeldt. cognitive psychology is not monolithic.g. 2000). responses.. Also in this issue. 2002. Dougher and colleagues first created classes of arbitrary stimuli using stimulus equivalence procedures.

g. The common names were sufficient to establish stimulus classes determined by sorting the stimuli in a category match-to-sample test. this issue). 1996). and Gagne ´ describes a model of conceptual combination. Language-based categories often take on a hierarchical structure involving nonequivalence relations between members. 1996. this is most evident in connectionist paradigms that focus on computer simulations of neural networks (e. 2001. It is worth noting that both relational frame theory and naming theory have been criticized for vagueness about the origins of higher order operants that supposedly underpin stimulus class formation. They use the terms mutual entailment and combinatorial entailment. 1995). Maddox describes a quantitative model of learning perceptual categories. Lane. Computer simulations also play a role in some areas of concept learning research with nonhumans (e. including equivalence. and semantic priming (Hayes & Bisset. and much research has examined arbitrary stimulus relations in humans as a theoretical basis of language and cognition (see Hayes et al.g. Their experiment illustrates the feasibility of conducting studies that bear on the plausibility of accounts based on higher order operants. In human cognitive psychology. relational frame theory is designed to account for a broad range of relations including equivalence phenomena. 1971. Barsalou. see Barsalou. & Critchfield. Wynne. & McIlvane. but cases worthy of examination can be identified in human cognitive psychology. demonstrations of category clustering in free recall (Galizio. but might not be perceptually related to other category members like cabinets or beds. This work is in its infancy. all primates are mammals. Fields et al. 1998) have all been accomplished with equivalence class procedures. Stewart. For example.. For example. The category furniture includes exemplars such as chairs that are perceptually related to one another. et al. It is encouraging that several phenomena typically observed in lexical classes have been modeled with equivalence relations. Quantitative modeling. Thus. in two papers in the present issue. Horne & Lowe... ‘‘greater than’’) entailed relations are different from those that emerge from equivalence relations. and . presented a model of concept learning with roots in the RescorlaWagner (1972) account of respondent conditioning.. Sidman. fast lexical mapping (Wilkinson. & Pilgrim. has not played a large role in the study of conceptual behavior. 2001). Reeve. 1998). In the present issue. for example.CATEGORIZATION AND CONCEPT LEARNING variation of the common-response procedure in which children were trained to make one common naming response to a set of physically unrelated stimuli and another common naming response to a different set of stimuli. and support accounts of. Innis. From this vantage point.. because in some relational frames (e. rather than symmetry and transitivity. Fields.. Although not targeting these theories per se. equivalence is one of many possible relational operants that Hayes et al. formal models. explore the merger of perceptually based classes with equivalence classes in generalized equivalence classes. Relational frame theory posits that much complex behavior. everyday language and cognition. 1993. Some aspects of the theory are illustrated in the attempt by Stewart et al. 245 in the style typical of many investigations published in JEAB. Formal Modeling Some investigations of concept learning test. (2001) call a frame of coordination. and Gagne ´ . but not all mammals are primates. 2000). Gluck and Bower (1988). is better understood as the result of ‘‘arbitrarily applicable relational responding’’ called a relational frame (relational frames also are invoked in some cognitive accounts. and it remains to be seen how adequately laboratory-generated stimulus relations can simulate. however. all birds have wings but not all wings are found on birds. Dube. or support the development of.g. (this issue) to craft a behavior-analytically based model of analogy. Clow. For instance. lexical classes often include both members that are related perceptually and members that are not (e. (this issue) describe some of the training conditions under which a generalized repertoire for categorizing perceptually defined stimuli is likely to arise. Several papers in the present issue extend such analyses to additional aspects of language-based categories. 1993). Concepts and Language Conceptual behavior often is thought to be linked to complex human capabilities such as language.g.

Barsalou (1992. 1991). members of lexical classes often exchange multiple functions. subordinate. One goal of the special issue. accessible introduction to mainstream cognitive work. Class members to which no discriminative function had been explicitly trained almost always inherited one such function. occurred in only about half of these cases. Systematic programs of research that complement. Studies of conceptual combination almost always focus on listener behavior (i. Consider. 4). for example. The present discussion is intended to place the articles of the special issue into a context of interest to journal readers. and basic level categorization. that the Margolis and Laurence (1999) volume reviewed in this issue. As Gagne ´ ’s article indicates. 11) and Honig and Fetterman (1992). the present issue so forth. and they highlight an important possible contribution of stimulus class research. The Stewart et al. see also a 1993 special issue of The Psychological Record. These additional functions were not mutually exclusive. as arranged in the laboratory. Vol. ulus class research provide a possible means of modeling speaker effects. If nothing else. comprehension. Stimuli to which one discriminative function had been explicitly trained almost never acquired a function that had been trained to another stimulus in the same class.e. data relevant to these complex effects are scarce in the literature on stimulus class formation. For a review of the literature on categorization and concept learning in nonhumans.g. potentially allowing the blending of functions through transfer. and extend those sampled in this special issue are called for.. ultimately. see Gagne ´ . Griffee and Dougher use differential reinforcement contingencies to establish hierarchical category relations that provide an elegant model of superordinate. therefore. 1978). but it barely scratches the surface of a complex topic. this issue) or do not distinguish clearly between speaker behavior (production) and listener behavior (e.. The preceding cases all involve attempts to generalize from concept learning. Unfortunately. we recommend Roberts (1998. in which naturally occurring abilities are brought into the laboratory and experimentally analyzed.. We hope that. consists of over 600 densely packed pages. challenge. model of analogy in this issue provides another relevant example. We did not attempt a comprehensive survey of this research area because the relevant literature is too extensive. to naturally occurring language phenomena outside the laboratory. which includes only selected seminal papers from only one research community. we recommend Sidman (1994). For an introduction to the research on stimulus equivalence in humans. Barsalou. making an example from Clow (2000) worthy of mention. 2 and 7) provides a brief. ZENTALL et al.g. Finally. In the present issue. Function-transfer procedures in stim- . we suggest a volume edited by Zentall and Smeets (1996. 43. exam headache).246 THOMAS R. reflecting the blending of discriminative repertoires. Gagne ´ illustrates the opposite approach. but conjunctive transfer. For discussion of the common research agendas that have united animal learning and human equivalence communities. is to promote further reading in this area. chap. CONCLUSION We have presented a brief introduction to some of the problems that have concerned investigators of categorization and concept learning. chap. and transfer of function may not always reciprocate among stimuli within a class (see also Rosch. separate training procedures established additional discriminative functions for two class members. No. In the present issue. Gagne ´ ’s research promotes testable predictions about the learning histories that may underlie conceptual combination involving the merger of lexical categories through noun–noun combinations (e. After humans had acquired equivalence classes. and the reports collected here illuminate some of the many questions worth addressing. no consideration of categorization and concept learning will be complete without some attention to the voluminous literature in human cognitive psychology. this special issue will encourage new applications of the experimental analysis of behavior to the phenomena and processes involved in categorization and concept learning. These outcomes broadly mimic patterns seen in lexical classes. Extensive surveys can be found in Smith and Medin (1981) and the Margolis and Laurence (1999) volume reviewed in the present issue.

Margolis & S. Laurence. Equivalence classification by California sea lions using class-specific reinforcers. 111–126.. Concepts: Core readings. The transfer of respondent eliciting and extinction functions through stimulus equivalence classes. Cambridge. Laurence (Eds. 239–254. Concepts: Core readings (pp. Galizio. The Psychological Record. Amsterdam: Elsevier. R.). 128. (2002). Maguire. R. The combination of response functions in merged equivalence classes: A behavioral paradigm for conceptual combination. Barsalou. J. In A.. Herrnstein. L. S. S. 183–204.) Principles of psychology.. Hillsdale. Herrnstein. 23. (2001). (1998). J.. (1987). Gathercole. Perceptual learning in pigeons: Decreased ability to discriminate samples mapped onto the same comparison in many-to-one matching. K. F. Dymond. & Laurence. M. Honig. S. 50. In G. 8. S.. 47. Keller. & Wulfert. (1992). & Stoddard.. . R. C.. W. Generalization of crossmodal stimulus equivalence classes: Operant processes as components in human category formation. D. S.. A. K. M. 37–50. S. C. 125–144. Journal of the Experimental Analysis of Behavior. F. there is work enough for all who are interested. 63. structure. & Schoenfeld. S. M. 31. A. Pigeon same-different concept learning with multiple stimulus classes. 131–158. & Margolis. & P. E. The Psychological Record.. Clustering in artificial categories: An equivalence analysis.). Cognitive aspects of stimulus control.. R. 51. S. J. A. K. fire. and linguistic vagary in concepts: Manifestations of a compositional system of perceptual symbols. L. & Hogan.. taxonomic. Astley. L. M. Jr. (1982)..). P. Psychonomic Bulletin & Review.. S. G.). F. Steirn. On the origins of naming and other symbolic behavior. From conditioning to category learning: An adaptive network model. Journal of Experimental Psychology: Animal Behavior Processes. A. M. & Roche. E. & Holth. no matter how vast the community of scholars who investigate categorization and concept learning. & Zentall.. M. Journal of the Experimental Analysis of Behavior. & Bower. (1992). Levels of stimulus control: A functional approach. D. W. 76. Dube. & Kastak. D. 23.. S. Stewart. R.. E. (2001). Journal of Experimental Psychology: Animal Behavior Processes. 133–166. & Staddon.. (1992). M. The symbolic species: The co-evolution of language and the brain. Kawashima. 48. NJ: Erlbaum. D. 267–279. When and where do we apply what we learn? A taxonomy for far transfer. 137–152). Barnett. & Fetterman. & Cable. J. R. H. R. & Sidman. M. Zentall. T. (1994). E.. S. (1993). 549–551. San Diego: Academic Press. 27. D. H. & Rehfeldt. 65. Mediating associations. C. L. E. Clow. Understanding complex behavior: The transformation of stimulus functions. (1991).. 8. T. C. Journal of the Experimental Analysis of Behavior.. R. New York: Plenum. L. Zentall & P. 385–392. S. M. (1989). N. Journal of Experimental Psychology: Animal Behavior Processes. E.. V. Acquired equivalence and distinctiveness in matching-to-sample by pigeons: Mediation by re- 247 inforcer-specific expectancies. New York: Appleton-Century-Crofts. W. W. K. Conway. D. G. J. D. Stimulus classes and the untrained acquisition of function. Augustson. (1998). R. Differential vocalization in budgerigars: Towards an experimental analysis of naming. R. Sherburne. NJ: Erlbaum. E. Cognition. Journal of Experimental Psychology: General. H. In T. L. & Loveland. Journal of the Experimental Analysis of Behavior. 1–64). A. Collins. Carrigan. 61(3B).. R. & Pilgrim. (1989).. (1997). Deriving categories to achieve goals. Dougher.. Greenway. J. A. Hillsdale. Clow. Science. 37. Complex visual concepts in the pigeon. P. N. The Behavior Analyst. Kaiser. Derived stimulus relations produce mediated and episodic priming. 2. M. K. Stimulus class formation in humans and animals (pp. 58. J. E. Gelman & J. and mutual exclusivity assumptions as initial constraints on word meanings. Markham. MA: MIT Press. Manabe.. H. Kastak. Concepts and cognitive science.). Journal of the Experimental Analysis of Behavior. Hayes. Dissertation Abstracts International.. J. Natural concepts in pigeons. Stimulus class formation and stimulus-reinforcer relations. Journal of the Experimental Analysis of Behavior. (2000). Gluck.. Women. Herrnstein. New York: Norton. W. S. A. E. & Markham. Chicago: University of Chicago Press.. C. In E.. A. P. T. MA: MIT Press. The whole-object. F. R. 612–637. Cambridge. H. J. (1988). & Wasserman. Markman. Schusterman. & Critchfield. G. Stimulus class formation in humans and animals (pp. 285–302. M. (1990). C. S. Conditional discrimination and equivalence relations: A theoretical analysis of control by negative stimuli. (1964). Journal of the Experimental Analysis of Behavior.. & Cavoto. 309–320. 21–41. (Eds.. Theories of memory (pp. Nonhumans have not yet shown stimulus equivalence.. T. (1999).. Barsalou. R. Lane. C. Journal of the Experimental Analysis of Behavior. (1950. Journal of Applied Behavior Analysis. 185–241. Katz. (2001).. (1997).. and non-similarity-based categorization. J.. Relational frame theory: A post-Skinnerian account of human language and cognition. 417–433. REFERENCES Arntzen.. J. (1997)... J..). Hayes. Amsterdam: Elsevier. T.. Horne. & Lowe. A. W. 70. Dougher. B. S.CATEGORIZATION AND CONCEPT LEARNING should demonstrate that. L. M. essentialism. McIlvane. Barnes-Holmes. (1997). Hillsdale. Innis. Lane. R. D. E. G. 244-259. 65–76. W. Zentall & P. Probability of stimulus equivalence as a function of training design. 8. (1995). Psychonomic Bulletin & Review. & Critchfield. Smeets (Eds. The psychology of learning and motivation (Vol. (1991). T. E. R. & Bisset. 617–630. H. R. R. 3–81). (1996). 29–89). 4. D. M. 378–381. In T. Cognitive psychology: An overview for cognitive scientists. J. (2000). Smeets (Eds. Jagielo. Loveland. Classification of vowels and consonants by individuals with moderate mental retardation: Development of arbitrary relations via match-to-sample training with compound stimuli. Deacon. J.. Barsalou. T.. In S. (1999). (1996). 111–133). Lakoff. J. and dangerous things: What categories reveal about the mind. Flexibility. & Ceci. J. B. W. 146. W. NJ: Erlbaum. Hayes. Mathis (Eds. Mackay. Margolis. C. P. Psychological Bulletin. (1996). 1619. Edwards. pp.. (1976). Bower (Ed. S. M. (1998). Cook. 51.. 609–614.

& Robbins. E. A. B.. G.. M. R. S. Jr. Sidman. 725–744. (1971). (1978). J. T.. New York: Appleton-Century-Crofts. J. Wixted. and children.. Rivera. Zentall.). 60.). In A. 14. Journal of the Experimental Analysis of Behavior. 225–238. & Willson-Morris. M. Why I am not a cognitive psychologist. N. Journal of Comparative and Physiological Psychology. K. Developmental Review. Evidence for common coding in manyto-one matching: Retention. (1985). Jackson-Smith. MA: Harvard University Press. Clement. B. (1999). A test of comparisonstimulus substitutability following one-to-many matching by pigions. Schwartz. K.. A discrimination analysis of training-structure effects on stimulus equivalence outcomes. 182–194. Sidman. Journal of Experimental Psychology: Animal Behavior Processes. In W. 43.). R. B. T. (in press). Classical conditioning: Vol. Lazar. New York: Cambridge University Press. Skinner.. & Tailby. Urcuioli. R. Dube.. & Bhatt. F. & Urcuioli. M. A. M. (1991). Stimulus class formation in humans and animals. W. D. 64–99). K. & Wagner. F. E. Journal of the Experimental Analysis of Behavior.. Conditional discrimination vs. (1965). Animal Learning & Behavior. Nuefeld. E. 5–13.. Fetterman (Eds. Zentall. Psychological Review. P. Rauzin. 23. D. 23–44. S. A. New York: Norton.. R.. 14.. (1993). T. ZENTALL et al. (1981). (1992). Psychonomic Bulletin & Review. Zentall. A. Zentall. Principles of categorization. 30.. S. Common coding in pigeons assessed through partial versus total reversals of many-to-one conditional and simple discriminations. 264–273. Rescorla. Language-naive chimpanzees (Pan troglodytes) judge . M. Categories and concepts. Conceptualization of natural and artificial stimuli by pigeons. R. R. Wynne. M. 194–201. J. Wright. Thompson. J. W. & Zentall.. 74. & Boysen.. A. Received June 19. (1978). L. P.. (1989). relations between relations in a conceptual matchingto-sample task. Sands. F. R.). 127–146. R. & Urcuioli. Rosch & B. 5–22. T. R. 233–237. S. (1972). R. Journal of the Experimental Analysis of Behavior. M. & Galizio. Amsterdam: Elsevier. L. H. (1996).. matching to sample: An expansion of the testing paradigm. B. (1995). (1988). Tailby. The operational analysis of psychological terms. (1995). M. Boston: Authors Cooperative. P. Steirn.. (1994). R. D. 37. J. 23.. Urcuioli. 59–63. Rosch. Equivalence relations and the reinforcement contingency. M. G. A crossdisciplinary perspective on studies of rapid word mapping in psycholinguistics and behavior analysis. Honig & J. M. J. (1993). Smith. 117–137. Perspectives on language and thought (pp. 2002 Final acceptance July 1. Wilkinson. & Weaver. Zentall. P. (2000). 15. Principles of animal cognition.. Hillsdale. Sidman. (Eds. Oden.. 17. S. Pilgrim. Saunders. & Gaitan. 1–10. 52. 31-43. T. Zentall. 72–106). C. 2002 Byrnes (Eds. T. M. A. Skinner. S. Formation of equivalence sets in pigeons. T. R. Animal Learning & Behavior. Concept learning by pigeons: Matching-to-sample with trial-unique video picture stimuli. & Steirn. L. L. D. (1988). & McIlvane.. Equivalence relations and behavior: A research story. McDaniel. & Green. T. A search for symmetry in the conditional discrimination of rhesus monkeys. S. W.. Sherburne. Reading and auditory-visual equivalences.. Journal of the Experimental Analysis of Behavior. G. A. M. Cognitive aspects of stimulus control (pp. R. & Baron. Journal of Experimental Psychology: Applied. NJ: Erlbaum. 37. (in press). & Hogan.. (1996). Journal of the Experimental Analysis of Behavior. C. J. Journal of Experimental Psychology: Animal Behavior Processes. (2002). Journal of Speech and Hearing Research. Wasserman. Wasserman. L.. (1997). 207–217. NJ: Erlbaum. The Psychological Record. T. Journal of Experimental Psychology: Animal Behavior Processes. 16.248 THOMAS R. F. baboons. Kirk. M. Sidman. J. Many-to-one and one-to-many associative learning in a naturalistic task.. 7(3). Sherburne. In E. A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement... Prokasy (Eds. T. Journal of the Experimental Analysis of Behavior. B. 36–42. Journal of the Experimental Analysis of Behavior. Cognition and categorization (pp. Vaughan. Stimulus generalization along the dimension of angularity. W. D. V. K. 72. & Damico-Nettleton. (1982).. J. (2001). Cook... & Smeets. M. Reversal of baseline relations and stimulus equivalence: I. Current research and theory (pp. 21. Newman.. W. 63. R. (1998). Lloyd (Eds. C. R. 16... 2. Cunningham. Sidman.. J. S. Animal Learning & Behavior . Sixmember stimulus classes generated by conditional-discrimination procedures. 43. 177– 186. 270–277.. 125–148. P. H. 27–48). Boston: McGraw-Hill. Black & W.. Journal of Experimental Psychology: Animal Behavior Processes. Adults. Reinforcement accounts for transitive inference performance. (1982). J. P. L. A. P. 203–223). Behaviorism.). E. E. & Carrigan. Same/different concept learning in the pigeon: The effect of negative instances and prior adaptation to the transfer stimuli. 436–444. W. Cognitive theories of reinforcement history surrogates: The case of likelihood ratio models of human recognition memory. Sidman. J.. 21–42. & Delius. Cambridge. & Medin. 5. Roberts. R. (1977). intertrial interference. J. N. and transfer. R. E. Hillsdale. F. (1945)... E. Common coding in a many-to-one delayed matching task as evidenced by facilitation and interference effects. Symmetry training in pigeons can produce functional equivalences. R. Psychology of learning and behavior . Animal Learning & Behavior.

Sign up to vote on this title
UsefulNot useful