The abstraction of form in semantic categories

Undergraduates were asked to generate a name for a hypothetical new exemplar of a category. They produced names that had the same numbers of syllables, the same endings, and the same types of word stems as existing exemplars of that category. In addition, novel exemplars, each consisting of a nonsense syllable root and a prototypical ending, were accurately assigned to categories. The data demonstrate the abstraction and use of surface properties of words.     

In search of abstraction: The varying abstraction model of categorization

A longstanding debate in the categorization literature concerns representational abstraction. Generally, when exemplar models, which assume no abstraction, have been contrasted with prototype models, which assume total abstraction, the former models have been found to be superior to the latter. Although these findings may rule out the idea that total abstraction takes place during category learning and instead suggest that no abstraction is involved, the idea of abstraction retains considerable intuitive appeal. In this article, we propose the varying abstraction model of categorization (VAM), which investigates the possibility that partial abstraction may play a role in category learning. We apply the VAM to four previously published data sets that have been used to argue that no abstraction is involved. Contrary to the previous findings, our results provide support for the idea that some form of partial abstraction can be used in people's category representations.     

The abstraction of musical ideas

Music presents information both sequentially, in the form of musical phrases, and simultaneously, in the form of chord structure. The ability to abstract musical structure presented sequentially and simultaneously was investigated using modified versions of the Bransford and Franks’ (1971) paradigm. Listeners heard subsets of musical ideas. The abstraction hypothesis predicted (1) false recognition of novel instances of the abstracted musical idea, (2) confidence of “recognition” should increase as recognition items approximate the complete musical idea, (3) correct rejection of “noncases,” which deviate from the acquired musical structure. Experiment 1 investigated sequential abstraction by using four-phrase folk melodies as musical ideas. Predictions 1 and 3 were confirmed, but the false recognition rate decreased as the number of phrases increased. Listeners were sensitive to improper combinations of phrases and to novel melodies different from melodies presented during acquisition. Experiment 2 investigated simultaneous abstraction using four-voice Bach chorales as musical ideas. Listeners spontaneously integrated choral subsets into holistic musical ideas. Musically trained listeners were better than naive listeners at identifying noncases.     

The downside of categories

One of the primary uses of categories is to draw inferences about novel objects based on their category membership. In a recent study, Lagnado and Shanks show that people make different inferences about an object depending on whether they first categorize the object at a general or specific level. Indeed, their inference changes even though they have been given no information about the object. This finding reveals limitations of category-based induction.     

The stability of social categories

One important thesis Ásta defends in Categories We Live By is that social properties and categories are somehow dependent on our thoughts, attitudes, or practices—that they are inventions of the mind, projected onto the world. Another important aspect of her view is that the social properties are related to certain base properties; an individual is placed in a category when the relevant base properties are thought to hold of them. I see the relationship between the social and the base as connected to the problem of explaining how the social properties are sufficiently stable so as to be taken seriously, both in theoretical endeavors as well as in practical matters of how we relate to each other. In this light, I identify stability constraints for an adequate account of social categories. I argue that certain distinctive aspects of Ásta's conferralist view of social categories, such as the radical contextualism in her account of gender, undermine the stability of categories and are at odds with taking social categories seriously. I end with the suggestion that a distinctive “sheltered” form of normativity might help us do justice to Ásta's insights while avoiding some of the destabilizing elements of conferralism.     

The abstraction and generalization of dot patterns

The Ss were trained on patterns representing different levels of distortion and transferred to test trials containing the old training patterns, the prototype patterns, and new distortions. Training on low levels of distortion consistently produced superior performance on all test patterns except the most distorted ones. The results also showed that the prototypes and low distortion-level test patterns of meaningful concepts were identified more often than prototypes and low distortion-level test patterns of meaningless concepts but the reverse was true for highly distorted test patterns. Training with patterns whose average dot positions corresponded to the original positions of the dots in the prototypes and to two different levels of distortion of training patterns did not facilitate transfer of any abstractions formed.     

The abstraction of visual prototypes by children

The present studies demonstrated that children as young as second graders could form prototypical representations from a set of highly similar stimuli. The prototype-plus-transformation model of this performance (Franks & Bransford, 1971) was contrasted with an attribute-frequency model (Neumann, 1974). Results supported the latter model when trigram stimuli were used to test differential predictions from the models.     

Case-like categories in children: The actor and some related categories

Case categories were investigated using a method in which pictures are presented accompanied by a sentence describing the scene: Preschool children learn to put tokens on the objects in the pictures according to the role each object plays in the scene as described. Generalization trials explore what the children include within the various roles. Children rapidly learn to associate tokens with case-like categories during the training phase, and readily transfer to new stimuli in generalization trials. The experiments demonstrate that children have a broad Actor category. The Actor “does” the action, and Actors are not necessarily animate. However, animacy is a class property of stimuli to which the children were very sensitive. There was evidence that User and Instrument are subcategories of Actor that are differentiated only when a sentence contains both. The data also suggest a Locative category, and Patient as a category of the grammatical object. Finally, the children appeared to treat subjects of predicate adjectives as a separate category, here called the Subject of Attribution, distinct from Actor, and functionally independent of objects of transitive verbs.     

The cognitive structure of social categories

Support for the prototype theory of categorization was found in a study of the structure of social categories. Though occupational terms such as DOCTOR are socially defined, they do not have the classical structure their clear definitional origins would predict. Conceptions of social categories are richer and more complex than those of physical object categories and subjects agree upon them. Comparison of various instructions for eliciting attributes of categories showed that whether subjects are asked to define a term, give characteristics, or describe ways they recognize members of categories, the attributes they list contribute to a prototype structure. These data provide evidence against the view that prototype structure is relevant only to an identification procedure and not to the core of concepts, as has been suggested.     

The representational basis of syllable categories

Two hypotheses concerning the relation between phoneme and syllable categories are contrasted. The phonological hypothesis holds that phoneme categories are perceptually basic, whereas syllable categories are abstract and specify the phonological rules operating on phoneme categories. The integrative routine hypothesis suggests that whereas phoneme categories are basic, syllable categories are not entirely abstract but are represented at least partially on the basis of higher order perceptual information, specifying the dynamics of coarticulating phoneme categories. Three experiments used a primed naming task in which subjects had to name the entire syllable (consonant-vowel and vowel-consonant; Experiment 1), the syllable-initial phoneme category (Experiment 2), or the syllable-final phoneme category (Experiment 3). The results supported the integrative routine hypothesis. S.D. is a postdoctoral fellow of the Belgian Fund for Scientific Research.     

The conceptual basis of graded categories

Two implications of the Conceptual Base View of Categorization (Honeck, Kibler, & Sugar, 1985) were tested. First, to the extent that subjects form similar schemas or conceptual bases, their judgments of the exemplariness of various events should agree, and, second, whether a conceptual base is framed from verbal or pictorial inputs should not matter. To test these implications, subjects were provided different reference stimuli as guides for ranking a set of sentences in terms of how well they illustrated the deeper meaning of the reference stimuli. These stimuli were an excellent verbal instance of the figurative meaning of a proverb, a poor verbal instance of this meaning, an abstract picture representing this meaning, or nothing (control group). The results yielded four lines of evidence that were consistent with the Conceptual Base View.     

Cue abstraction and ideal prototype abstraction in estimation tasks

This study investigates abstracted processes and introduces a new prototype abstraction model adapted to estimation tasks. This prototype abstraction model assumes that the processing of whole exemplar patterns supports the detection of the underlying statistics necessary for the abstraction of two extreme prototypes on the continuous criterion dimension of the task. The prototypes are stored in memory as valid reference points for future similarity-based judgments. This prototype model was compared with the cue abstraction model, which assumes that people abstract cue weights in learning and add the cue information from exemplars to infer their criterion values varying on the continuous dimension. This study hypothesises that the training mode and the number of exemplars in training interact and affect subsequent model performance at test. The results from an experiment confirmed this hypothesis and showed that observational training supports an efficient prototype abstraction and feedback training supports an efficient cue abstraction.     

The effect of linguistic abstraction on interpersonal distance

It is well known that people describe positive behaviors of others close to them (e.g., in‐group member, friend) in abstract terms, but with concrete terms in the case of people who they are not close to (e.g., out‐group member, enemy). In contrast, negative behaviors of people who we are close to are described in concrete terms, but in abstract terms for people who are distant. However, the communicative impact of such subtle differences in language use on a receiver who is also the actor of the behavior being described has never been addressed. We hypothesized and found that a positive abstract message compared to a positive concrete message leads to perceived proximity to the sender, while a negative abstract message compared to a negative concrete message leads to perceived distance. The implications of this study, which is the first to show the communicative impact of biased language use, are discussed. Copyright © 2006 John Wiley & Sons, Ltd.     

The empirical case for role-governed categories

Most theories of categorization posit feature-based representations. Markman and Stilwell (2001) argued that many natural categories name roles in relational systems and therefore they are role-governed categories. There is little extant empirical evidence to support the existence of role-governed categories. Three experiments examine predictions for ways that role-governed categories should differ from feature-based categories. Experiment 1 shows that our knowledge of role-governed categories, in contrast to feature-based categories, is largely about properties extrinsic to category members. Experiment 2 shows that role-governed categories have more prominent ideals than feature-based categories. Experiment 3 demonstrates that novel role-governed categories are licensed by the instantiation of novel relational structures. We then discuss broader implications for the study of categories and concepts.     

Classical physical abstraction

An informal theory is set forth of relations between abstract entities, includingcolors, physical quantities, times, andplaces in space, and the concrete things thathave them, or areat orin them, based on the assumption that there are close analogies between these relations and relations between abstractsets and the concrete things that aremembers of them. It is suggested that even standard scientific usage of these abstractions presupposes principles that are analogous to postulates of abstraction, identity, and other fundamental principles of set theory. Also discussed is the significance of important disanalogies between sets and physical abstractions, including especiallymodal andtemporal aspects of physical abstractions, which is related to the problem of the characterizingconstancy, of colors, physical attributes, and locations in space.     

The stability and flexibility of spatial categories

Four experiments examined the flexibility and stability with which children and adults organize locations into categories based on their spatiotemporal experience with locations. Seven-, 9-, 11-year-olds, and adults learned the locations of 20 objects in an open, square box. During learning, participants experienced the locations in four spatiotemporally defined groups (i.e., four sets of nearby locations learned together in time). At test, participants attempted to place the objects in the correct locations without the aid of the dots marking the locations. Children and adults displaced the objects toward the corners of the box consistent with the organization they experienced during learning, suggesting that they used spatiotemporal experience to organize the locations into groups. Importantly, the pattern of organization remained the same following a long delay for all four age groups, demonstrating stability. For adults, this organization shifted after a new pattern of spatiotemporal experience was introduced, suggesting that adults' categories based on spatiotemporal experience are quite flexible. Children only exhibited flexibility when the new pattern of spatiotemporal organization was consistent with available perceptual cues, demonstrating that the flexibility with which children organize locations into categories is intimately tied to both remembered and perceptual sources of information.     

The development of geographic categories and biases

Children and university students (N=58) estimated the locations of major cities in North America. At age 9, a distinct home region was apparent, but no differentiation between northern US and Canadian cities. At 11, four developments were observed: Children divided North America into regions that were not based solely on national boundaries but were the same as university students' regions; psychological border zones between regions exaggerated distances between them; children used new location information to update their estimates for all cities in a seeded region and in adjacent and nonadjacent regions; children preserved the ordinal structure of their initial location estimates for cities in their home region but relied on regional prototype locations to adjust estimates in less familiar regions. The updating methods reflect fundamentally different mechanisms. Theoretical and educational implications are discussed.