Induction and category coherence

In studies of category formation, subjects rarely construct family resemblance categories. Instead, they divide objects into categories using a single dimension. This is a puzzling result given the widely accepted view that natural categories are organized in terms of a family resemblance principle. The observation that natural categories support inductive inferences is used here to test the hypothesis that family resemblance categories would be constructed if stimuli were first used to generate inductive inferences. In two experiments, subjects answered either induction questions, which made interproperty relationships more salient, or frequency questions, which required information only about individual properties, before they performed a sorting task. Subjects were likely to produce family resemblance sorts if they had first answered induction questions but not if they had answered frequency questions.     

A Two-Stage Model of Category Construction

The current consensus is that most natural categories are not organized around strict definitions (a list of singly necessary and jointly sufficient features) but rather according to a family resemblance (FR) principle: Objects belong to the same category because they are similar to each other and dissimilar to objects in contrast categories. A number of computational models of category construction have been developed to provide an account of how and why people create FR categories (Anderson, 1990; Fisher, 1987). Surprisingly, however, only a few experiments on category construction or free sorting have been run and they suggest that people do not sort examples by the FR principle. We report several new experiments and a two-stage model for category construction. This model is contrasted with a variety of other models with respect to their ability to account for when FR sorting will and will not occur. The experiments serve to identify one basis for FR sorting and to support the two-stage model. The distinctive property of the two-stage model is that it assumes that people impose more structure than the examples support in the first stage and that the second stage adjusts for this difference between preferred and perceived structure. We speculate that people do not simply assimilate probabilistic structures but rather organize them in terms of discrete structures plus noise.     

先前知识对初中学生学习家族相似性类别的影响研究

采用特征—主题的类别学习范式,考察了先前知识对初中学生学习具有家族相似性结构人工类别的影响。结果表明:基于先前知识的主题关系对初中学生的类别学习有显著促进;初中学生在基于已有知识的类别学习中更多地获得了知识关联特征的信息,同时也能对机械特征保持敏感。     

Are rating-based procedures reliable for derivation of typicality judgments from children?

The aim of the present study was to examine whether a rating-based procedure that has already been used by other investigators can be used for derivation of typicality ratings from children. The evidence reported in the study establishes that such a procedure cannot be reliably used for this purpose. The results show that children rated category items in terms of personal preferences rather than as a function of how representative they considered the items to be of their superordinate category. On the basis of these findings, an alternative method based on the family resemblance scores of the category members was proposed in order to derive typicality ratings from young children. This family resemblance method of obtaining typicality judgments may be useful to investigators interested in assessing how children process categorical information.     

Classification of exemplars with single- and multiple-feature manifestations: the effects of relevant dimension variation and category structure

Most classification research focuses on cases in which each abstract feature has the same surface manifestation whenever it is presented. Previous research finds that people have difficulty learning to classify when each abstract feature has multiple surface manifestations. These studies created multiple manifestations by varying aspects of the stimuli irrelevant to the abstract feature dimension. In this article, multiple manifestations were created by varying aspects of the stimuli relevant to the abstract feature dimension. People given categories with the family resemblance category structure often used in psychology experiments had difficulty learning to classify when multiple manifestations were present, even though the variation was relevant. This effect was reversed when a family resemblance structure with nondiagnostic values was used.     

A further investigation of category learning by inference

Categories are learned in many ways besides by classification, for example, by making inferences about classified items. One hypothesis is that classifications lead to the learning of features that distinguish categories, whereas inferences promote the learning of the internal structure of categories, such as the typical features. Experiment 1 included single-feature and full-feature classification tests following either classification or inference learning. Consistent with predictions, inference learners did better on the single tests but worse on the full tests. Experiment 2 further showed that inference learners, unlike classification learners, were no better at classifying items that they had seen at study compared with equally typical items they had not seen at study. Experiment 3 showed that features queried about during inference learning were classified better than ones not queried about, although even the latter features showed some learning on single-feature tests. The discussion focuses on how different types of category learning lead to different category representations.     

两种学习模式下类别学习的结果：原型和样例

利用“学习－迁移”的任务范式和单一特征类别判断技术,探讨了分类和推理两种类别学习模式的结果,比较了两种学习模式的效果和策略。研究表明：两种学习模式产生了不同的结果,分类学习的结果是样例,推理学习的结果是原型;在学习效果方面,分类学习比推理学习在达标比例上更高,但在进度上差异不显著;在策略运用方面,分类学习比推理学习更快地使用单维度策略,而在高水平策略的运用上,两者差异不显著     

Classification as diagnostic reasoning

An ongoing goal in the field of categorization has been to determine how objects’ features provide evidence of membership in one category versus another. Well-known findings include that feature diagnosticity is a function of how often the feature appears in category members versus nonmembers, their perceptual salience, how features are used in support of inferences, and how observable features are related to other observable features. We tested how diagnosticity is affected by causal relations between observable and unobserved features. Consistent with our view of classification as diagnostic reasoning, we found that observable features are more diagnostic to the extent that they are caused by underlying features that define category membership, because the presence of the latter can be (causally) inferred from the former. Implications of these results for current views of conceptual structure and models of categorization are discussed.     

Salience Not Status: How Category Labels Influence Feature Inference

Two main uses of categories are classification and feature inference, and category labels have been widely shown to play a dominant role in feature inference. However, the nature of this influence remains unclear, and we evaluate two contrasting hypotheses formalized as mathematical models: the label special‐mechanism hypothesis and the label super‐salience hypothesis. The special‐mechanism hypothesis is that category labels, unlike other features, trigger inference decision making in reference to the category prototypes. This results in a tendency for prototype‐compatible inferences because the labels trigger a special mechanism rather than because of any influences they have on similarity evaluation. The super‐salience hypothesis assumes that the large label influence is due to their high salience and corresponding impact on similarity without any need for a special mechanism. Application of the two models to a feature inference task based on a family resemblance category structure yields strong support for the label super‐salience hypothesis and in particular does not support the need for a special mechanism based on prototypes.     

Differential brain responses when applying criterion attribute versus family resemblance rule learning

Subsystems of category learning have been identified on the basis of general domains of content (e.g., tools, faces). The present study examined categories from the standpoint of internal structure and determined brain topography associated with expressing two fundamentally different category rule structures (criterion attribute, CA, and family resemblance, FR). CA category learning involves processing stimuli by isolated features and classifying by properties held by all members. FR learning involves processing stimuli by integral wholes and classifying on overall similarity among members without sharing identical features. fMRI BOLD response to CA and FR categorization was measured with pseudowords as stimuli. Category knowledge for both tasks was mastered prior to brain imaging. Areas of activation emerged unique to the structure of each category and followed from the nature of the rule abstraction procedure. CA categorization was implemented by strong target monitoring and expectation (medial parietal), rule maintenance in working memory, feature selection processes (inferior frontal), and a sensitivity to high frequency components of the stimulus such as isolated features (anterior temporal). FR categorization, consistent with its multi-featural nature, involved word-level processing (left extrastriate) that evoked articulatory rehearsal (medial cerebellar). The data suggest category structure is an important determinant of brain response during categorization. For instance, anterior temporal structures may help attune visual processing systems to high frequency components to support the learning of criterial, highly predictive rules.     

类别学习中两种学习模式的比较研究：分类学习与推理学习

采用学习迁移任务范式,使用基于单一特征的类别判断技术,比较了非线性分离结构下,分类学习和推理学习的学习效率、学习过程与策略和学习结果。结果表明：在学习效率上,分类学习比推理学习更好地习得了含有较多样例的类别知识,分类学习的速度上显著快于推理学习。在学习的过程与策略上,推理学习比分类学习更为关注类别内不同特征的相关,但在分类策略的运用上不如分类学习灵活。在学习的结果上,推理学习倾向于原型记忆,分类学习倾向于进行样例记忆,分类学习比推理学习更好地掌握了类别原型     

Wittgensteinian Semantics

Wittgenstein emphasizes two points concerning his notion of family resemblance. One is that the use of a family resemblance expression resists characterization by certain kinds of rules; the other is that due to the prevalence of family resemblance in the philosophical lexicon, philosophical inquiry must in many cases proceed differently from how it traditionally has. This paper develops an interpretation of family resemblance that seeks to do justice to these claims. I argue that what is characteristic about family resemblance expressions is not that they exhibit a basic semantic feature unique to themselves, but that they combine a number of semantic properties that happen not to be coinstantiated elsewhere. These features include (1) content variability (also a property of ambiguous expressions, polysemes, and standard indexicals), (2) a feature I call "topicality" (which is also a characteristic of polysemes), and (3) "semantic openness" (a feature of many ordinary indexicals). The notions of topicality and semantic openness are explained, and certain terms of natural language are shown to be family resemblance expressions. I conclude by indicating some of the potential philosophical ramifications of these results.     

Features of graded category structure: Generalizing the family resemblance and polymorphous concept models

Many real-world categories contain graded structure: certain category members are rated as more typical or representative of the category than others. Research has shown that this graded structure can be well predicted by the degree of commonality across the feature sets of category members. We demonstrate that two prominent feature-based models of graded structure, the family resemblance (Rosch & Mervis, 1975) and polymorphous concept models (Hampton, 1979), can be generalized via the contrast model (Tversky, 1977) to include both common and distinctive feature information, and apply the models to the prediction of typicality in 11 semantic categories. The results indicate that both types of feature information play a role in the prediction of typicality, with common features weighted more heavily for within-category predictions, and distinctive features weighted more heavily for contrast-category predictions. The same pattern of results was found in additional analyses employing rated goodness and exemplar generation frequency. It is suggested that these findings provide insight into the processes underlying category formation and representation.     

刺激材料的整体性知觉对类别建构的影响

采用标准－匹配的实验程序,操纵刺激材料的空间整合性和知觉的整体性水平,探讨逐个呈现刺激材料时影响被试类别建构策略的根本原因。报告了3个实验,结果证明：逐个呈现并不一定会导致被试倾向于家族相似性归类;刺激材料的空间整合性不一定会影响被试的类别建构策略,空间整合和空间分离都可能出现家族相似性和单维归类倾向;刺激材料的整体性知觉水平对类别建构影响明显：知觉为分离则倾向于家族相似性分类;知觉为整体则倾向于单维分类。被试在实验可能采用分析性策略,然而材料的整体性知觉影响了分析的侧重点,从而对类别建构产生了非常重要的影响     

Children's category‐based inferences affect classification

Children learn many new categories and make inferences about these categories. Much work has examined how children make inferences on the basis of category knowledge. However, inferences may also affect what is learned about a category. Four experiments examine whether category‐based inferences during category learning influence category knowledge and thereby affect later classifications for 5‐ to 7‐year‐olds. The children learned to classify pictures of new types of creatures on the basis of a salient feature (colour) and then answered a question that required them to make an inference on the basis of other features. At test, children classified pictures that included only some features (without colour). Experiment 1 showed that the features relevant to the inference during learning led to better classification than did features irrelevant to the inference. Experiment 2 replicated this finding even when the relevant features were physically close to the irrelevant features. Experiments 3 and 4 found this effect even when the classification was learned prior to the inference task and even when no mention was made of the categories during inference learning. Taken together, these results show that making inferences during category learning can influence category knowledge and suggest a need to integrate the work on category learning and category‐based inferences.     

Family resemblances: Studies in the internal structure of categories

Six experiments explored the hypothesis that the members of categories which are considered most prototypical are those with most attributes in common with other members of the category and least attributes in common with other categories. In probabilistic terms, the hypothesis is that prototypicality is a function of the total cue validity of the attributes of items. In Experiments 1 and 3, subjects listed attributes for members of semantic categories which had been previously rated for degree of prototypicality. High positive correlations were obtained between those ratings and the extent of distribution of an item's attributes among the other items of the category. In Experiments 2 and 4, subjects listed superordinates of category members and listed attributes of members of contrasting categories. Negative correlations were obtained between prototypicality and superordinates other than the category in question and between prototypicality and an item's possession of attributes possessed by members of contrasting categories. Experiments 5 and 6 used artificial categories and showed that family resemblance within categories and lack of overlap of elements with contrasting categories were correlated with ease of learning, reaction time in identifying an item after learning, and rating of prototypicality of an item. It is argued that family resemblance offers an alternative to criterial features in defining categories.     

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.     

Do Additional Features Help or Hurt Category Learning? The Curse of Dimensionality in Human Learners

The curse of dimensionality, which has been widely studied in statistics and machine learning, occurs when additional features cause the size of the feature space to grow so quickly that learning classification rules becomes increasingly difficult. How do people overcome the curse of dimensionality when acquiring real‐world categories that have many different features? Here we investigate the possibility that the structure of categories can help. We show that when categories follow a family resemblance structure, people are unaffected by the presence of additional features in learning. However, when categories are based on a single feature, they fall prey to the curse, and having additional irrelevant features hurts performance. We compare and contrast these results to three different computational models to show that a model with limited computational capacity best captures human performance across almost all of the conditions in both experiments.     

Learning about the internal structure of categories through classification and feature inference

Previous research on category learning has found that classification tasks produce representations that are skewed toward diagnostic feature dimensions, whereas feature inference tasks lead to richer representations of within-category structure. Yet, prior studies often measure category knowledge through tasks that involve identifying only the typical features of a category. This neglects an important aspect of a category's internal structure: how typical and atypical features are distributed within a category. The present experiments tested the hypothesis that inference learning results in richer knowledge of internal category structure than classification learning. We introduced several new measures to probe learners' representations of within-category structure. Experiment 1 found that participants in the inference condition learned and used a wider range of feature dimensions than classification learners. Classification learners, however, were more sensitive to the presence of atypical features within categories. Experiment 2 provided converging evidence that classification learners were more likely to incorporate atypical features into their representations. Inference learners were less likely to encode atypical category features, even in a “partial inference” condition that focused learners' attention on the feature dimensions relevant to classification. Overall, these results are contrary to the hypothesis that inference learning produces superior knowledge of within-category structure. Although inference learning promoted representations that included a broad range of category-typical features, classification learning promoted greater sensitivity to the distribution of typical and atypical features within categories.