Episodic and prototype models of category learning

The question of what processes are involved in the acquisition and representation of categories remains unresolved despite several decades of research. Studies using the well-known prototype distortion task (Posner and Keele in J Exp Psychol 77:353–363, 1968) delineate three candidate models. According to exemplar-based models, we memorize each instance of a category and when asked to decide whether novel items are category members or not, the decision is explicitly based on a similarity comparison with each stored instance. By contrast, prototype models assume that categorization is based on the similarity of the target item to an implicit abstraction of the central tendency or average of previously encountered instances. A third model suggests that the categorization of prototype distortions does not depend on pre-exposure to study exemplars at all and instead reflects properties of the stimuli that are easily learned during the test. The four experiments reported here found evidence that categorization in this task is predicated on the first and third of these models, namely transfer at test and the exemplar-based model. But we found no evidence for the second candidate model that assumed that categorization is based on implicit prototype abstraction.     

Visual noise reveals category representations

How are categories represented in human memory? Exemplar models assume that a category is represented by individual instances from that category that have been experienced. More generally, a category might be represented by multiple templates stored in memory. A new item is classified according to its similarity to these templates.Prototype models represent a category with a single summary abstraction (i.e., a single template), often the central tendency of the experienced items. A new item is classified according to its similarity to these category prototypes. Here, we show how a technique for correlating observers’ responses with external noise can be used not only to distinguish single- from multiple-template representations, but also to induce the form of these templates. The technique is applied to two tasks requiring categorization of simple visual patterns; the results demonstrate that observers used multiple traces to represent their categories, and thus highlight the procedure’s potential for use in more complex settings.     

Stages of abstraction and exemplar memorization in pigeon category learning

It has been proposed that human category learning consists of an early abstraction-based stage followed by a later exemplar-memorization stage. To investigate whether similar processing stages extend to category learning in a nonverbal species, we applied a prototype-exception paradigm to investigating pigeon category learning. Five birds and 8 humans learned six-dimensional perceptual categories constructed to include prototypes, typical items, and exceptions. We evaluated the birds' and humans' categorization strategies at different points during learning. Early on in both species, prototype performance improved rapidly as exception performance remained below chance, indicating an initial mastery of the categories' general structure. Later on, exception performance improved selectively and dramatically, indicating exception-item resolution and exemplar memorization. Abstraction- and exemplar-based formal models reinforced these interpretations. The results suggest a psychological transition in pigeon category learning from abstraction- to exemplar-based processing similar to that found in humans.     

Instance memorization and category influence: Challenging the evidence for multiple systems in category learning

A class of dual-system theories of categorization assumes a categorization system based on actively formed prototypes in addition to a separate instance memory system. It has been suggested that, because they have used poorly differentiated category structures (such as the influential “5-4” structure), studies supporting the alternative exemplar theory reveal little about the properties of the categorization system. Dual-system theories assume that the instance memory system only influences categorization behaviour via similarity to single isolated instances, without generalization across instances. However, we present the results of two experiments employing the 5-4 structure to argue against this. Experiment 1 contrasted learning in the standard 5-4 structure with learning in an even more poorly differentiated 5-4 structure. In Experiment 2, participants memorized the 5-4 structure based on a five minute simultaneous presentation of all nine category instances. Both experiments revealed category influences as reflected by differences in instance learnability and generalization, at variance with the dual-system prediction. These results have implications for the exemplars versus prototypes debate and the nature of human categorization mechanisms.     

Distinguishing prototype-based and exemplar-based processes in dot-pattern category learning

The authors contrast exemplar-based and prototype-based processes in dot-pattern categorization. In Experiments 1A and 1B, participants provided similarity ratings of dot-distortion pairs that were distortions of the same originating prototype. The results show that comparisons to training exemplars surrounding the prototype create flat typicality gradients within a category and small prototype-enhancement effects, whereas comparisons to a prototype center create steep typicality gradients within a category and large prototype-enhancement effects. Thus, prototype and exemplar theories make different predictions regarding common versions of the dot-distortion task. Experiment 2 tested these different predictions by having participants learn dot-pattern categories. The steep typicality gradients, the large prototype effects, and the superior fit of prototype models suggest that participants refer to-be-categorized items to a representation near the category's center (the prototype), and not to the training exemplars that surround the prototype.     

False prototype enhancement effects in dot pattern categorization

Results from the classic dot pattern distortion paradigm have sometimes yielded prototype enhancement effects that could not be accounted for by exemplar models of categorization. However, in these experiments the status of the prototype was confounded with certain stimulus-specific properties as well as with the frequency of presentation of the prototype during testing. In two mock-subliminal experiments, participants made categorization judgments to patterns that were generated as prototypes, low-level distortions, or high-level distortions. The participants rated the prototypes as being more likely to be members of a category, although no patterns were presented during training, and there was no objective category structure. In two other experiments, greater prototype enhancement effects were observed when the prototype and low-level distortions were presented with greater frequency during transfer. These results suggest that classic prototype enhancement effects may not be due to the abstraction of a prototype at time of original learning, but rather to other factors not formalized in extant models.     

The modulating influence of category size on the classification of exception patterns

Generalization gradients to exception patterns and the category prototype were investigated in two experiments. In Experiment 1, participants first learned categories of large size that contained a single exception pattern, followed by a transfer test containing new instances that had a manipulated similarity relationship to the exception or a nonexception training pattern as well as distortions of the prototype. The results demonstrated transfer gradients tracked the prototype category rather than the feedback category of the exception category. In Experiment 2, transfer performance was investigated for categories varying in size (5, 10, 20), partially crossed with the number of exception patterns (1, 2, 4). Here, the generalization gradients tracked the feedback category of the training instance when category size was small but tracked the prototype category when category size was large. The benefits of increased category size still emerged, even with proportionality of exception patterns held constant. These, and other outcomes, were consistent with a mixed model of classification, in which exemplar influences were dominant with small-sized categories and/or high error rates, and prototype influences were dominant with larger sized categories.     

The modulating influence of category size on the classification of exception patterns

Generalization gradients to exception patterns and the category prototype were investigated in two experiments. In Experiment 1, participants first learned categories of large size that contained a single exception pattern, followed by a transfer test containing new instances that had a manipulated similarity relationship to the exception or a nonexception training pattern as well as distortions of the prototype. The results demonstrated transfer gradients tracked the prototype category rather than the feedback category of the exception category. In Experiment 2, transfer performance was investigated for categories varying in size (5, 10, 20), partially crossed with the number of exception patterns (1, 2, 4). Here, the generalization gradients tracked the feedback category of the training instance when category size was small but tracked the prototype category when category size was large. The benefits of increased category size still emerged, even with proportionality of exception patterns held constant. These, and other outcomes, were consistent with a mixed model of classification, in which exemplar influences were dominant with small-sized categories and/or high error rates, and prototype influences were dominant with larger sized categories.     

Comparing categorization models

Four experiments are presented that competitively test rule- and exemplar-based models of human categorization behavior. Participants classified stimuli that varied on a unidimensional axis into 2 categories. The stimuli did not consistently belong to a category; instead, they were probabilistically assigned. By manipulating these assignment probabilities, it was possible to produce stimuli for which exemplar- and rule-based explanations made qualitatively different predictions. F. G. Ashby and J. T. Townsend's (1986) rule-based general recognition theory provided a better account of the data than R. M. Nosofsky's (1986) exemplar-based generalized context model in conditions in which the to-be-classified stimuli were relatively confusable. However, generalized context model provided a better account when the stimuli were relatively few and distinct. These findings are consistent with multiple process accounts of categorization and demonstrate that stimulus confusion is a determining factor as 10 which process mediates categorization.     

Natural categories

The hypothesis of the study was that the domains of color and form are structured into nonarbitrary, semantic categories which develop around perceptually salient “natural prototypes.” Categories which reflected such an organization (where the presumed natural prototypes were central tendencies of the categories) and categories which violated the organization (natural prototypes peripheral) were taught to a total of 162 members of a Stone Age culture which did not initially have hue or geometric-form concepts. In both domains, the presumed “natural” categories were consistently easier to learn than the “distorted” categories. Even when not central, natural prototype stimuli tended to be more rapidly learned and more often chosen as the most typical example of the category than were other stimuli. Implications for general differences between natural categories and the artificial categories of concept formation research were discussed.     

The nature and structure of infant form categories

The nature of form categories in 3 to 4-month infants was studied using the visual preference for novelty in the familiarization-novelty paradigm. Novelty preference indicates habituation to and recognition of the familiar. In a series of experiments employing three form categories composed of dot patterns, generalized habituation to new category members was used to assess categorization behavior in the recognition of visual forms. At 3 to 4 months of age, infants did not initially show any systematic preferences for “good” or symmetrical examples of a category relative to “distorted” examples (Experiment 1) and this was true for all three form categories used (i.e., square, triangle, and diamond). Evidence for categorization was seen in the recognition performance of 3- to 4-month infants (Experiment 2). Infants showed generalized habituation to the previously unseen category prototypes following exposure to six exemplars within each of the three form categories. Given evidence that infants could discriminate between the prototype and other category members (Experiment 3), “inability to discriminate” was ruled out as an explanation for this form categorization or generalized habituation effect. Four subsequent experiments were conducted to determine whether infants exhibit a prototypicality structure for their remembered categories and whether certain conditions which have been shown to enhance prototypicality effects with adults have similar effects with infants. No evidence of a prototypicality structure was found for the form categories of infants when the number of exemplars during familiarization was limited to 6 and the test for form recognition followed immediately (Experiment 4). However, a prototypicality structure for the remembered form categories was found when a 3-min delay was introduced between familiarization and tests for form recognition (Experiment 5), when 12 exemplars were presented during familiarization (Experiment 7), or when the prototype was included as one of the six exemplars during the familiarization period (Experiment 6).     

Array training in a categorization task

Two components of categorization, within-category commonalities and between-category distinctiveness, were investigated in a categorization task. Subjects learned three prototype categories composed of moderately high distortions, by observing arrays containing patterns that belonged either to a common prototype category or to three different categories; a third group learned patterns presented one at a time, mirroring the standard paradigm. Following 6 learning blocks, subjects transferred to old patterns and new patterns at low-, medium-, and high-level distortions of the category prototype. The results showed that array training facilitated learning, especially when patterns in the array belonged to the same category. Transfer results showed a strong gradient effect across pattern distortion level for all conditions, with the highest performance obtained following array training on different category patterns and worst in the control condition. Interestingly, the old training patterns were classified worse than new low and no better than medium distortions. Neither this ordering nor the steepness of the gradient across prototype similarity for each condition could be predicted by the generalized context model. A prototype model better captured the steep gradient and ordinal pattern of results, although the overall fits were only slightly better than the exemplar model. The crucial role played by category commonalities and distinctiveness on categorical representations is addressed.     

Atypical categorization in children with high-functioning autism spectrum disorder

Children with autism spectrum disorder process many perceptual and social events differently from typically developing children, suggesting that they may also form and recognize categories differently. We used a dot pattern categorization task and prototype comparison modeling to compare categorical processing in children with high-functioning autism spectrum disorder and matched typical controls. We were interested in whether there were differences in how children with autism use average similarity information about a category to make decisions. During testing, the group with autism spectrum disorder endorsed prototypes less and was seemingly less sensitive to differences between to-be-categorized items and the prototype. The findings suggest that individuals with high-functioning autism spectrum disorder are less likely to use overall average similarity when forming categories or making categorical decisions. Such differences in category formation and use may negatively impact processing of socially relevant information, such as facial expressions. A supplemental appendix for this article may be downloaded from http://pbr.psychonomic-journals.org/content/supplemental.     

Combining exemplar-based category representations and connectionist learning rules.

Adaptive network and exemplar-similarity models were compared on their ability to predict category learning and transfer data. An exemplar-based network (Kruschke, 1990a, 1990b, 1992) that combines key aspects of both modeling approaches was also tested. The exemplar-based network incorporates an exemplar-based category representation in which exemplars become associated to categories through the same error-driven, interactive learning rules that are assumed in standard adaptive networks. Experiment 1, which partially replicated and extended the probabilistic classification learning paradigm of Gluck and Bower (1988a), demonstrated the importance of an error-driven learning rule. Experiment 2, which extended the classification learning paradigm of Medin and Schaffer (1978) that discriminated between exemplar and prototype models, demonstrated the importance of an exemplar-based category representation. Only the exemplar-based network accounted for all the major qualitative phenomena; it also achieved good quantitative predictions of the learning and transfer data in both experiments.     

Human adults and human infants show a "perceptual magnet effect" for the prototypes of speech categories, monkeys do not

Many perceptual categories exhibit internal structure in which category prototypes play an important role. In the four experiments reported here, the internal structure of phonetic categories was explored in studies involving adults, infants, and monkeys. In Experiment 1, adults rated the category goodness of 64 variants of the vowel i parallel on a scale from 1 to 7. The results showed that there was a certain location in vowel space where listeners rated the i parallel vowels as best instances, or prototypes. The perceived goodness of i parallel vowels declined systematically as stimuli were further removed from the prototypic i parallel vowel. Experiment 2 went beyond this initial demonstration and examined the effect of speech prototypes on perception. Either the prototypic or a nonprototypic i parallel vowels was used as the referent stimulus and adults' generalization to other members of the category was examined. Results showed that the typicality of the speech stimulus strongly affected perception. When the prototype of the category served as the referent vowel, there was significantly greater generalization to other i parallel vowels, relative to the situation in which the nonprototype served as the referent. The notion of a perceptual magnet was introduced. The prototype of the category functioned like a perceptual magnet for other category members; it assimilated neighboring stimuli, effectively pulling them toward the prototype. In Experiment 3, the ontogenetic origins of the perceptual magnet effect were explored by testing 6-month-old infants. The results showed that infants' perception of vowels was also strongly affected by speech prototypes. Infants showed significantly greater generalization when the prototype of the vowel category served as the referent; moreover, their responses were highly correlated with those of adults. In Experiment 4, Rhesus monkeys were tested to examine whether or not the prototype's magnet effect was unique to humans. The animals did not provide any evidence of speech prototypes; they did not exhibit the magnet effect. It is suggested that the internal organization of phonetic categories around prototypic members is an ontogenetically early, species-specific, aspect of the speech code.     

Are there representational shifts during category learning?

Early theories of categorization assumed that either rules, or prototypes, or exemplars were exclusively used to mentally represent categories of objects. More recently, hybrid theories of categorization have been proposed that variously combine these different forms of category representation. Our research addressed the question of whether there are representational shifts during category learning. We report a series of experiments that tracked how individual subjects generalized their acquired category knowledge to classifying new critical transfer items as a function of learning. Individual differences were observed in the generalization patterns exhibited by subjects, and those generalizations changed systematically with experience. Early in learning, subjects generalized on the basis of single diagnostic dimensions, consistent with the use of simple categorization rules. Later in learning, subjects generalized in a manner consistent with the use of similarity-based exemplar retrieval, attending to multiple stimulus dimensions. Theoretical modeling was used to formally corroborate these empirical observations by comparing fits of rule, prototype, and exemplar models to the observed categorization data. Although we provide strong evidence for shifts in the kind of information used to classify objects as a function of categorization experience, interpreting these results in terms of shifts in representational systems underlying perceptual categorization is a far thornier issue. We provide a discussion of the challenges of making claims about category representation, making reference to a wide body of literature suggesting different kinds of representational systems in perceptual categorization and related domains of human cognition.     

Categorization generated by extended prototypes—An axiomatic approach

We suggest a model of categorization based on prototypes. A set of entities, identified with some finite dimensional Euclidian space, is partitioned into a finite number of categories. Such a categorization is said to be generated by extended prototypes if there is a set of distinguished entities, one for each category, such that the categorization is determined by proximity to these prototypical entities. The prototypes are called extended since they are described by points in a higher dimensional space than the entities. Sufficient conditions for a categorization to be generated by extended prototypes are provided. These conditions are also necessary if the prototypes are in general position.     

Similarity-scaling studies of dot-pattern classification and recognition.

Classification performance in the dot-pattern, prototype-distortion paradigm (e.g., Posner & Keele, 1968) was modeled within a multidimensional scaling (MDS) framework. MDS solutions were derived for sets of dot patterns that were generated from prototypes. These MDS solutions were then used in conjunction with exemplar, prototype, and combined models to predict classification and recognition performance. Across 3 experiments, an MDS-based exemplar model accounted for the effects of several fundamental learning variables, including level of distortion of the patterns, category size, delay of transfer phase, and item frequency. Most important, the model quantitatively predicted classification probabilities for individual dot patterns in the sets, not simply general trends of performance. There was little evidence for the existence of a prototype-abstraction process that operated above and beyond pure exemplar-based generalization.     

Fruits and vegetables categorized: an application of the generalized context model

In the study reported in this paper, we investigated the categorization of well-known and novel food items in the categories fruits and vegetables. Predictions based on Nosofsky’s (1984, 1986) generalized context model (GCM), on a multiplicative-similarity prototype model, and on an instantiation model as applied in Storms, De Boeck, and Ruts (2001) were compared. Despite suggestions in the literature that prototype models predict categorization from large categories better than exemplar models do, our results showed that the exemplar-based GCM yielded clearly better predictions than did a (multiplicative-similarity) prototype model.     

Human adults and human infants show a “perceptual magnet effect” for the prototypes of speech categories,monkeys do not

Many perceptual categories exhibit internal structure in which category prototypes play an important role. In the four experiments reported here, the internal structure of phonetic categories was explored in studies involving adults, infants, and monkeys. In Experiment 1, adults rated the category goodness of 64 variants of the vowel /i/ on a scale from 1 to 7. The results showed that there was a certain location in vowel space where listeners rated the /i/ vowels as best instances, or prototypes. The perceived goodness of Iii vowels declined systematically as stimuli were further removed from the prototypic Iii vowel. Experiment 2 went beyond this initial demonstration and examined the effect of speech prototypes on perception. Either the prototypic or a nonprototypic IM vowel was used as the referent stimulus and adults’ generalization to other members of the category was examined. Results showed that the typicality of the speech stimulus strongly affected perception. When the prototype of the category served as the referent vowel, there was significantly greater generalization to other /i/ vowels, relative to the situation in which the nonprototype served as the referent. The notion of aperceptual magnet was introduced. The prototype of the category functioned like a perceptual magnet for other category members; it assimilated neighboring stimuli, effectively pulling them toward the prototype. In Experiment 3, the ontogenetic origins of the perceptual magnet effect were explored by testing 6-month-old infants. The results showed that infants’ perception of vowels was also strongly affected by speech prototypes. Infants showed significantly greater generalization when the prototype of the vowel category served as the referent; moreover, their responses were highly correlated with those of adults. In Experiment 4, Rhesus monkeys were tested to examine whether or not the prototype’s magnet effect was unique to humans. The animals did not provide any evidence of speech prototypes; they did not exhibit the magnet effect. It is suggested that the internal organization of phonetic categories around prototypic members is an ontogenetically early, species-specific, aspect of the speech code