Inference and classification learning of abstract coherent categories

Category learning research has primarily focused on how people learn to classify items using simple observable features. However, classification is only 1 way to learn categories. In addition, many concepts have an underlying coherence that explains the featural similarity among exemplars, such as abstract coherent concepts whose instances differ greatly on their observable features. In 3 experiments, the authors investigated how abstract coherent categories are acquired through 2 common means of category learning, classification and inference. Because inference promotes more focus on within-category information than does classification, they hypothesized that inference learning would lead to a better understanding of the underlying coherence of abstract coherent categories. All 3 experiments support this prediction.     

Collaboration facilitates abstract category learning

We examined the effects of collaboration (dyads vs. individuals) and category structure (coherent vs. incoherent) on learning and transfer. Working in dyads or individually, participants classified examples from either an abstract coherent category, the features of which are not fixed but relate in a meaningful way, or an incoherent category, the features of which do not relate meaningfully. All participants were then tested individually. We hypothesized that dyads would benefit more from classifying the coherent category structure because past work has shown that collaboration is more beneficial for tasks that build on shared prior knowledge and provide opportunities for explanation and abstraction. Results showed that dyads improved more than individuals during the classification task regardless of category coherence, but learning in a dyad improved inference-test performance only for participants who learned coherent categories. Although participants in the coherent categories performed better on a transfer test, there was no effect of collaboration.     

Learning abstract relations from using categories

When people learn categories, the importance of the features and relations in the category representation reflects both their diagnosticity for classification and their relevance to the use of the category. In earlier work in which the influence of category use on the representation has been shown, only cases in which the features and relations were simple, observable, and very specific were examined. Learners may begin to understand the underlying similarities of category members by using the categories. In the four experiments presented here, learners applied a simple category-specific formula to category members. The test results showed that the learners had incorporated relations among features from this use, including cases in which the relations were abstract. This learning occurred even though the relations were actually not predictive of category membership but just perceived to be so as a function of the use.     

The generalized polymorphous concept account of graded structure in abstract categories

Abstract categories present with graded structure. The extent to which feature commonality between exemplars and category provides a satisfying account of this graded structure varies from one abstract category to the other (Hampton, 1981). We investigate whether the incorporation of features that exemplars share with external categories yields an improved account of abstract categories' graded structures. In doing so, we follow the suggestion that abstract categories are relational in nature (Goldstone, 1996; Wiemer-Hastings & Xu, 2005). The generalized polymorphous concept model, which incorporates both types of features, is found to improve the account of typicality and category membership in three of seven studied abstract categories. These three categories are found to be the most abstract, suggesting that it is appropriate to think of abstract categories as varying along a continuum of abstractness/interrelatedness rather than as a distinct type of category altogether.     

Abstraction of prototypical information by adults and 10-month-old infants

The primary purpose of this study was to investigate whether preverbal infants, when presented with exemplars of an artificially constructed category, would abstract a prototypical representation of the category, and if so, whether this representation was formed by either "counting" or "averaging" the features that were varying among category members. Two experiments are reported. In Experiment 1, a set of stimuli was developed and tested for which it was demonstrated that adult subjects would readily abstruct either a modal or an average prototypical representation. The type of representation abstracted was found to be dependent on the discriminability of the feature values. In Experiment 2, 10-mo.-old infants were tested using a habituation paradigm with the stimuli developed in the first experiment. The results of this study indicated that the infants were also able to abstract the featural information that was varying among the exemplars of the category, and the infants formed an internal represenatation of the category by averaging feature values. Thus, the results clearly imply that infants are able to constructively process visual information and hence take a more active role in category formation than had been previously believed.     

Testing the speech unit hypothesis with the primed matching task: Phoneme categories are perceptually basic

The basic speech unit (phoneme or syllable) problem was investigatedwith the primed matching task. In primed matching, subjects have to decide whether the elements of stimulus pairs are the same or different. The prime should facilitate matching in as far as its representation is similar to the stimuli to be matched. If stimulus representations generate graded structure, with stimulus instances being more or less prototypical for the category, priming should interact with prototypicality because prototypical instances are more similar to the activated category than are low-prototypical instances. Rosch (1975a, 1975b) showed that, by varying the matching criterion (matching for physical identity or for belonging to the same category), the specific patterns of the priming × prototypicality interaction could differentiate perceptually based from abstract categories. Bytesting this pattern forphoneme and syllable categories, the abstraction level of these categories canbe studied. After finding reliable prototypicality effects for both phoneme and syllable categories (Experiments 1 and 2), primed phoneme matching (Experiments 3 and 4) and primed syllable matching (Experiments 5 and 6) were used under both physical identity instructions and same-category instructions. The results make clear that phoneme categories are represented on the basis of perceptual information, whereas syllable representations are more abstract. The phoneme category can thus be identified as the basic speech unit. Implications for phoneme and syllable representation are discussed.     

How to switch on and switch off semantic priming effects for natural and artifactual categories: activation processes in category memory depend on focusing specific feature dimensions

"Semantic priming" refers to the phenomenon that people react faster to target words preceded by semantically related rather than semantically unrelated words. We wondered whether momentary mind sets modulate semantic priming for natural versus artifactual categories. We interspersed a category priming task with a second task that required participants to react to either the perceptual or action features of simple geometric shapes. Focusing on perceptual features enhanced semantic priming effects for natural categories, whereas focusing on action features enhanced semantic priming effects for artifactual categories. In fact, significant priming effects emerged only for those categories thought to rely on the features activated by the second task. This result suggests that (a) priming effects depend on momentary mind set and (b) features can be weighted flexibly in concept representations; it is also further evidence for sensory-functional accounts of concept and category representation.     

Categorization and representation of physics problems by experts and novices

The representation of physics problems in relation to the organization of physics knowledge is investigated in experts and novices. Four experiments examine (a) the existence of problem categories as a basis for representation; (b) differences in the categories used by experts and novices; (c) differences in the knowledge associated with the categories; and (d) features in the problems that contribute to problem categorization and representation. Results from sorting tasks and protocols reveal that experts and novices begin their problem representations with specifiably different problem categories, and completion of the representations depends on the knowledge associated with the categories. For, the experts initially abstract physics principles to approach and solve a problem representation, whereas novices base their representation and approaches on the problem's literal features.     

Concept learning and feature interpretation

Models of categorization often assume that people classify new instances directly on the basis of the presented, observable features. Recent research, however, has suggested that the coherence of a category may depend in part on more abstract features that can link together observable features that might otherwise seem to have little similarity. Thus, category learning may also involve the determination of the appropriate abstract features that underlie a category and link together the observable features. We show in four experiments that observable features of a category member are often interpreted as congruent with abstract features that are suggested by observable features of other highly available category members. Our discussion focuses on the implications of these findings for future research.     

The influence of category coherence on inference about cross-classified entities

A critical function of categories is their use in property inference (Heit, 2000). However, one challenge to using categories in inference is that most entities in the world belong to multiple categories (e.g., Fido could be a dog, a pet, a mammal, or a security system). Building on Patalano, Chin-Parker, and Ross (2006), we tested the hypothesis that category coherence (the extent to which category features go together in light of prior knowledge) influences the selection of categories for use in property inference about cross-classified entities. In two experiments, we directly contrasted coherent and incoherent categories, both of which included cross-classified entities as members, and we found that the coherent categories were used more readily as the source of both property transfer and property extension. We conclude that category coherence, which has been found to be a potent influence on strength of inference for singly classified entities (Rehder & Hastie, 2004), is also central to category use in reasoning about novel cross-classified ones.     

Learning nonlinearly separable categories by inference and classification

Previous research suggests that learning categories by classifying new instances highlights information that is useful for discriminating between categories. In contrast, learning categories by making predictive inferences focuses learners on an abstract summary of each category (e.g., the prototype). To test this characterization of classification and inference learning further, the authors evaluated the two learning procedures with nonlinearly separable categories. In contrast to previous research involving cohesive, linearly separable categories, the authors found that it is more difficult to learn nonlinearly separable categories by making inferences about features than it is to learn them by classifying instances. This finding reflects that the prototype of a nonlinearly separable category does not provide a good summary of the category members. The results from this study suggest that having a cohesive category structure is more important for inference than it is for classification.     

Development of an abstract category representation for the spatial relation between in 6- to 10-month-old infants

Nine experiments examined the formation of an abstract category representation for the spatial relation between by 6- to 10-month-old infants. The experiments determined that 9- to 10-month-olds, but not 6- to 7-month-olds, could form an abstract category representation for between when performing in an object-variation version of the between categorization task. The results also demonstrated that 6- to 7-month-olds could form category representations for between in the object-variation version of the between categorization task but that such representations were specific to the particular objects presented. The evidence confirms that representations for different spatial relations emerge at different points during development, and suggests that each representation undergoes its own period of development from concrete to abstract.     

Representation at different levels in a conceptual hierarchy

The present study examines the influence of hierarchical level on category representation. Three computational models of representation – an exemplar model, a prototype model and an ideal representation model – were evaluated in their ability to account for the typicality gradient of categories at two hierarchical levels in the conceptual domain of clothes. The domain contains 20 subordinate categories (e.g., trousers, stockings and underwear) and an encompassing superordinate category (CLOTHES). The models were evaluated both in terms of their ability to fit the empirical data and their generalizability through marginal likelihood. The hierarchical level was found to clearly influence the type of representation: For concepts at the subordinate level, exemplar representations were supported. At the superordinate level, however, an ideal representation was overwhelmingly preferred over exemplar and prototype representations. This finding contributes to the increasingly dominant view that the human conceptual apparatus adopts both exemplar representations and more abstract representations, contradicting unitary approaches to categorization.     

The development of spatial category representations from 4 to 7 years

Representation of spatial categories was assessed in 4‐ to 7‐year‐olds. Across nine spatial categories (In, On, Under, In Front, Behind, Above, Below, Left, and Right), children were asked to pick the odd‐one‐out from four images, three of which displayed the same spatial relationship between two objects, and one which showed a different spatial relationship. Results support our proposed model of spatial category representation. Children progressed through three levels of understanding: from rigid (level 1), to abstract (level 2) to broad (including non‐prototypical category exemplars) (level 3) understanding of spatial category membership. This developmental pattern was common to all spatial categories, and the ages at which children reached each level varied across categories, in line with the order in which category representations emerge in infancy.     

Semantics of the transitive construction: prototype effects and developmental comparisons

This paper investigates whether an abstract linguistic construction shows the kind of prototype effects characteristic of non-linguistic categories, in both adults and young children. Adapting the prototype-plus-distortion methodology of Franks and Bransford (1971), we found that whereas adults were lured toward false-positive recognition of sentences with prototypical transitive semantics, young children showed no such effect. We examined two main implications of the results. First, it adds a novel data point to a growing body of research in cognitive linguistics and construction grammar that shows abstract linguistic categories can behave in similar ways to non-linguistic categories, for example, by showing graded membership of a category. Thus, the findings lend psychological validity to the existing cross-linguistic evidence for prototypical transitive semantics. Second, we discuss a possible explanation for the fact that prototypical sentences were processed differently in adults and children, namely, that children's transitive semantic network is not as interconnected or cognitively coherent as adults'.     

What's behind different kinds of kinds: effects of statistical density on learning and representation of categories

This research examined how differences in category structure affect category learning and category representation across points of development. The authors specifically focused on category density--or the proportion of category-relevant variance to the total variance. Results of Experiments 1-3 showed a clear dissociation between dense and sparse categories: Whereas dense categories were readily learned without supervision, learning of sparse categories required supervision. There were also developmental differences in how statistical density affected category representation. Although children represented both dense and sparse categories on the basis of the overall similarity (Experiment 4A), adults represented dense categories on the basis of similarity and represented sparse categories on the basis of the inclusion rule (Experiment 4B). The results support the notion that statistical structure interacts with the learning regime in their effects on category learning. In addition, these results elucidate important developmental differences in how categories are represented, which presents interesting challenges for theories of categorization.     

On the Interaction of Theory and Data in Concept Learning

Standard models of concept learning generally focus on deriving statistical properties of a category based on data (i.e., category members and the features that describe them) but fail to give appropriate weight to the contact between people's intuitive theories and these data. Two experiments explored the role of people's prior knowledge or intuitive theories on category learning by manipulating the labels associated with the category. Learning differed dramatically when categories of children's drawings were meaningfully labeled (e.g., “done by creative children”) compared to when they were labeled in a neutral manner. When categories are meaningfully labeled, people bring intuitive theories to the learning context. Learning then involves a process in which people search for evidence in the data that supports abstract features or hypotheses that have been activated by the intuitive theories. In contrast, when categories are labeled in a neutral manner, people search for simple features that distinguish one category from another. Importantly, the final study suggests that learning involves an interaction of people's intuitive theories with data, in which theories and data mutually influence each other. The results strongly suggest that straight-forward, relatively modular ways of incorporating prior knowledge into models of category learning are inadequate. More telling, the results suggest that standard models may have fundamental limitations. We outline a speculative model of learning in which the interaction of theory and data is tightly coupled. The article concludes by comparing the results to recent artificial intelligence systems that use prior knowledge during learning.     

Order Algebras as Models of Linear Logic

The starting point of the present study is the interpretation of intuitionistic linear logic in Petri nets proposed by U. Engberg and G. Winskel. We show that several categories of order algebras provide equivalent interpretations of this logic, and identify the category of the so called strongly coherent quantales arising in these interpretations. The equivalence of the interpretations is intimately related to the categorical facts that the aforementioned categories are connected with each other via adjunctions, and the compositions of the connecting functors with co-domain the category of strongly coherent quantales are dense. In particular, each quantale canonically induces a Petri net, and this association gives rise to an adjunction between the category of quantales and a category whose objects are all Petri nets.     

Korean deaf adolescents' recognition of written words for taxonomic categories of different levels

Deaf college students seem to have relatively stronger associations from words for taxonomic categories of basic (e.g., snake) to those of super-ordinate (e.g., reptiles) level than vice versa compared with hearing students in word association (Marschark, Convertino, McEvoy & Masteller, 2004). In deciding whether two sequentially presented words for taxonomic categories of different levels are conceptually related, deaf adolescents might therefore have a poorer performance when they see a category name before than when they see it after one of the corresponding exemplar words. Deaf Korean adolescents were found to recognize words for taxonomic categories of super-ordinate level with lower efficiencies than those of basic level. Their accuracy seemed to reflect a reversed typicality effect when they decided that first-presented words for taxonomic categories of basic level were conceptually related to second-presented words for those of super-ordinate level. It was argued that deaf Korean adolescents went through a temporary stage of having iconic representations of several exemplars of the category aroused in working memory before the abstract semantic representation was fully activated when they saw the word for a taxonomic category of super-ordinate level.     

Distinct and common cortical activations for multimodal semantic categories

If semantic representations are based on particular types of perceptual features, then category knowledge that arises from multimodal sensory experiences should rely on distinct and common sensory brain regions depending on the features involved. Using a similarity-based generation-andcomparison task, we found that semantic categories activated cortical areas associated with taste and smell, biological motion, and visual processing. Fruit names specifically activated medial orbitofrontal regions associated with taste and smell. Labels for body parts and clothing activated lateral temporal occipitoparietal areas associated with perceiving the human body. More generally, visually biased categories activated ventral temporal regions typically ascribed to visual object recognition, whereas functional categories activated lateral frontotemporal areas previously associated with the representation of usage properties. These results indicate that semantic categories that are distinguished by particular perceptual properties rely on distinct cortical regions, whereas semantic categories that rely on similar types of features depend on common brain areas.