An inverse base rate effect with continuously valued stimuli

It is well known that people do not always make normative use of information about relative frequencies of categories when making categorical judgments. The "inverse base rate" effect (Medin & Edelson, 1988) is a typical example of this: Subjects violate normative reasoning principles by assigning certain ambiguous stimuli as belonging to the less frequent of two categories, rather than to the more common category. This effect has been explained as being due to the shifting of attention from shared stimulus features to distinctive features during learning. When stimuli are defined by values along continuous dimensions, rather than by the presence and absence of features, then attention could shift between dimensions or between values, or both. In three experiments, base rate differences were used to determine the way in which attention is shifted during learning about stimuli with continuously valued dimensions. Simulation modeling shows that the results are consistent with the movement of attention both between and within stimulus dimensions.     

Observed attention allocation processes in category learning

In two empirical studies of attention allocation during category learning, we investigate the idea that category learners learn to allocate attention optimally across stimulus dimensions. We argue that “optimal” patterns of attention allocation are model or process specific, that human learners do not always optimize attention, and that one reason they fail to do so is that under certain conditions the cost of information retrieval or use may affect the attentional strategy adopted by learners. We empirically investigate these issues using a computer interface incorporating an “information-board” display that collects detailed information on participants' patterns of attention allocation and information search during learning trials. Experiment 1 investigated the effects on attention allocation of distributing perfectly diagnostic features across stimulus dimensions versus within one dimension. The overall pattern of viewing times supported the optimal attention allocation hypothesis, but a more detailed analysis produced evidence of instance- or category-specific attention allocation, a phenomenon not predicted by prominent computational models of category learning. Experiment 2 investigated the strategies adopted by category learners encountering redundant perfectly predictive cues. Here, the majority of participants learned to distribute attention optimally in a cost–benefit sense, allocating attention primarily to only one of the two perfectly predictive dimensions. These results suggest that learners may take situational costs and benefits into account, and they present challenges for computational models of learning that allocate attention by weighting stimulus dimensions.     

Observed attention allocation processes in category learning

In two empirical studies of attention allocation during category learning, we investigate the idea that category learners learn to allocate attention optimally across stimulus dimensions. We argue that "optimal" patterns of attention allocation are model or process specific, that human learners do not always optimize attention, and that one reason they fail to do so is that under certain conditions the cost of information retrieval or use may affect the attentional strategy adopted by learners. We empirically investigate these issues using a computer interface incorporating an "information-board" display that collects detailed information on participants' patterns of attention allocation and information search during learning trials. Experiment 1 investigated the effects on attention allocation of distributing perfectly diagnostic features across stimulus dimensions versus within one dimension. The overall pattern of viewing times supported the optimal attention allocation hypothesis, but a more detailed analysis produced evidence of instance- or category-specific attention allocation, a phenomenon not predicted by prominent computational models of category learning. Experiment 2 investigated the strategies adopted by category learners encountering redundant perfectly predictive cues. Here, the majority of participants learned to distribute attention optimally in a cost-benefit sense, allocating attention primarily to only one of the two perfectly predictive dimensions. These results suggest that learners may take situational costs and benefits into account, and they present challenges for computational models of learning that allocate attention by weighting stimulus dimensions.     

Incremental Bayesian Category Learning From Natural Language

Models of category learning have been extensively studied in cognitive science and primarily tested on perceptual abstractions or artificial stimuli. In this paper, we focus on categories acquired from natural language stimuli, that is, words (e.g., chair is a member of the furniture category). We present a Bayesian model that, unlike previous work, learns both categories and their features in a single process. We model category induction as two interrelated subproblems: (a) the acquisition of features that discriminate among categories, and (b) the grouping of concepts into categories based on those features. Our model learns categories incrementally using particle filters, a sequential Monte Carlo method commonly used for approximate probabilistic inference that sequentially integrates newly observed data and can be viewed as a plausible mechanism for human learning. Experimental results show that our incremental learner obtains meaningful categories which yield a closer fit to behavioral data compared to related models while at the same time acquiring features which characterize the learned categories. (An earlier version of this work was published in Frermann and Lapata 2014 .)     

Featural selective attention,exemplar representation,and the inverse base-rate effect

Selective attention plays a central role in theories of category learning and representation. In exemplar theory, selective attention has typically been formalized as operating uniformly across entire stimulus dimensions. Selective featural attention operating within dimensions has been recognized as a conceptual possibility, but relatively little research has focused on evaluating it. In the present research, we explored the usefulness of selective featural attention in the context of exemplar representation. We report the results of embedding the feature-to-category relations typically associated with the inverse base-rate effect—a classic and paradoxical category-learning result— perceptual category-learning task using a category structure with three multivalued feature dimensions. An exemplar model incorporating featural selective attention accurately accounted for the inverse base-rate effect that occurred but failed to do so with only dimensional attention.     

Effects of postresponse visual stimuli on visual discrimination learning in the rhesus monkey

To learn more about the mechanism (or mechanisms) involved with postresponse stimulus processing during discrimination learning, a series of studies was conducted with monkeys to determine why the combined relevant and irrelevant stimuli impair learning more than irrelevant stimuli appearing alone. It was found that: (a) the greater size and complexity of the combination of stimuli were not responsible for the greater deficit, while the presence of the relevant stimuli (S^D and S^Δ) within the stimulus combination apparently was; (b) the more similar the postresponse irrelevant stimuli were to the relevant stimuli the greater the deficit that resulted; and (c) monkeys that had earlier learned to discriminate the relevant and irrelevant features of a combination showed no learning impairment when this same stimulus combination was later presented after the response during a new learning problem. These results were interpreted as evidence that: (1) processes associated with learning a discrimination problem do not end with the execution of a choice response; (2) postresponse stimuli produce greater impairment in discrimination learning when they are distorted versions of the relevant stimuli; and (3) the impairment resulting from postresponse irrelevant stimuli occurs primarily when this misinformation is processed and misperceived as being relevant to learning the discrimination problem.     

Asymmetry of attentional set in rhesus monkeys learning colour and shape discriminations

We trained rhesus monkeys on six visual discrimination problems using stimuli that varied in both shape and colour. For one group of animals shape was always relevant in these six problems, and colour always irrelevant, and for the other animals vice versa. During these “intradimensional shifts” (ID) the problems were learned at equal rates by the two groups, shape-relevant and colour-relevant. We then trained three further problems in which the other dimension was now relevant (“extradimensional shifts”, ED). The animals showed slower learning when shifting from colour-relevant to shape-relevant, but not when shifting from shape-relevant to colour-relevant. These results show that monkeys' ability to selectively attend to a relevant stimulus dimension and to ignore an irrelevant dimension depends on the experimenter's choice of relevant and irrelevant dimensions.     

特征的呈现方式对类别学习中规则和相似性知识获得的影响

采用原型变异任务,通过操纵规则和相似性特征的呈现通道,探讨特征的呈现方式对类别学习中规则和相似性知识获得的影响。结果发现,在听觉-视觉条件下习得规则的人数显著多于习得相似性的人数,而在视觉-听觉和视觉-视觉条件下则不存在显著差异;并且,三种条件下习得规则的正确率都高于习得相似性的正确率。这说明特征的呈现方式影响对规则和相似性特征的习得,在听觉通道呈现规则时被试更倾向基于规则分类。     

Order Matters! Influences of Linear Order on Linguistic Category Learning

Linguistic category learning has been shown to be highly sensitive to linear order, and depending on the task, differentially sensitive to the information provided by preceding category markers (premarkers, e.g., gendered articles) or succeeding category markers (postmarkers, e.g., gendered suffixes). Given that numerous systems for marking grammatical categories exist in natural languages, it follows that a better understanding of these findings can shed light on the factors underlying this diversity. In two discriminative learning simulations and an artificial language learning experiment, we identify two factors that modulate linear order effects in linguistic category learning: category structure and the level of abstraction in a category hierarchy. Regarding category structure, we find that postmarking brings an advantage for learning category diagnostic stimulus dimensions, an effect not present when categories are non-confusable. Regarding levels of abstraction, we find that premarking of super-ordinate categories (e.g., noun class) facilitates learning of subordinate categories (e.g., nouns). We present detailed simulations using a plausible candidate mechanism for the observed effects, along with a comprehensive analysis of linear order effects within an expectation-based account of learning. Our findings indicate that linguistic category learning is differentially guided by pre- and postmarking, and that the influence of each is modulated by the specific characteristics of a given category system.     

Do chickens (Gallus gallus f. domestica) decompose visual figures?

To investigate whether learning to discriminate between visual compound stimuli depends on decomposing them into constituting features, hens were first trained to discriminate four features (red, green, horizontal, vertical) from two dimensions (colour, line orientation). After acquisition, hens were trained with compound stimuli made up from these dimensions in two ways: a separable (line on a coloured background) stimulus and an integral one (coloured line). This compound training included a reversal of reinforcement of only one of the two dimensions (half-reversal). After having achieved the compound stimulus discrimination, a second dimensional training identical to the first was performed. Finally, in the second compound training the other dimension was reversed. Two major results were found: (1) an interaction between the dimension reversed and the type of compound stimulus: in compound training with colour reversal, separable compound stimuli were discriminated worse than integral compounds and vice versa in compound training with line orientation reversed. (2) Performance in the second compound training was worse than in the first one. The first result points to a similar mode of processing for separable and integral compounds, whereas the second result shows that the whole stimulus is psychologically superior to its constituting features. Experiment 2 repeated experiment 1 using line orientation stimuli of reversed line and background brightness. Nevertheless, the results were similar to experiment 1. Results are discussed in the framework of a configural exemplar theory of discrimination that assumes the representation of the whole stimulus situation combined with transfer based on a measure of overall similarity.     

Unsupervised category learning with integral-dimension stimuli

Despite the recent surge in research on unsupervised category learning, the majority of studies have focused on unconstrained tasks in which no instructions are provided about the underlying category structure. Relatively little research has focused on constrained tasks in which the goal is to learn predefined stimulus clusters in the absence of feedback. The few studies that have addressed this issue have focused almost exclusively on stimuli for which it is relatively easy to attend selectively to the component dimensions (i.e., separable dimensions). In the present study, we investigated the ability of participants to learn categories constructed from stimuli for which it is difficult, if not impossible, to attend selectively to the component dimensions (i.e., integral dimensions). The experiments demonstrate that individuals are capable of learning categories constructed from the integral dimensions of brightness and saturation, but this ability is generally limited to category structures requiring selective attention to brightness. As might be expected with integral dimensions, participants were often able to integrate brightness and saturation information in the absence of feedback—an ability not observed in previous studies with separable dimensions. Even so, there was a bias to weight brightness more heavily than saturation in the categorization process, suggesting a weak form of selective attention to brightness. These data present an important challenge for the development of models of unsupervised category learning.     

Unsupervised category learning with integral-dimension stimuli

Despite the recent surge in research on unsupervised category learning, the majority of studies have focused on unconstrained tasks in which no instructions are provided about the underlying category structure. Relatively little research has focused on constrained tasks in which the goal is to learn predefined stimulus clusters in the absence of feedback. The few studies that have addressed this issue have focused almost exclusively on stimuli for which it is relatively easy to attend selectively to the component dimensions (i.e., separable dimensions). In the present study, we investigated the ability of participants to learn categories constructed from stimuli for which it is difficult, if not impossible, to attend selectively to the component dimensions (i.e., integral dimensions). The experiments demonstrate that individuals are capable of learning categories constructed from the integral dimensions of brightness and saturation, but this ability is generally limited to category structures requiring selective attention to brightness. As might be expected with integral dimensions, participants were often able to integrate brightness and saturation information in the absence of feedback--an ability not observed in previous studies with separable dimensions. Even so, there was a bias to weight brightness more heavily than saturation in the categorization process, suggesting a weak form of selective attention to brightness. These data present an important challenge for the development of models of unsupervised category learning.     

Asymmetry of attentional set in rhesus monkeys learning colour and shape discriminations

We trained rhesus monkeys on six visual discrimination problems using stimuli that varied in both shape and colour. For one group of animals shape was always relevant in these six problems, and colour always irrelevant, and for the other animals vice versa. During these "intradimensional shifts" (ID) the problems were learned at equal rates by the two groups, shape-relevant and colour-relevant. We then trained three further problems in which the other dimension was now relevant ("extradimensional shifts", ED). The animals showed slower learning when shifting from colour-relevant to shape-relevant, but not when shifting from shape-relevant to colour-relevant. These results show that monkeys' ability to selectively attend to a relevant stimulus dimension and to ignore an irrelevant dimension depends on the experimenter's choice of relevant and irrelevant dimensions.     

Interaction of knowledge-driven and data-driven processing in category learning

The present paper argues that category learning is both a data-driven and a knowledge-driven process. This is described in a generic model that distinguishes between categorical knowledge, conceptual knowledge, and implicit cognitive theories. The model assumes that each of these knowledge aspects may affect the process of category learning by affecting the way similarities between objects are perceived. This central assumption of the model is tested in two experiments. The first experiment shows that the presence or absence of prior categorical and conceptual knowledge affects the psychological stimulus space by changing the saliency of the stimulus dimensions. The second experiment uses these weights to predict the distribution of errors over the stimuli and the number of trials to criterion in category learning by other participants under the same knowledge conditions. We conclude that prior categorical and conceptual knowledge affect category learning by mediation of similarity perception, and discuss the implications of these results.     

注意和工作记忆对交错呈现优势的影响

以家族相似性图案为材料,让被试在单任务或双任务条件下以集中呈现或交错呈现的方式进行观察（实验1）或反馈（实验2）学习,记录眼动,探究注意对交错呈现优势的影响,以及工作记忆在其中的作用。发现当进行观察学习时,注意影响交错呈现优势,结果支持区别对比理论和注意衰减理论;当进行反馈学习时,注意的影响还有待进一步探究。同时,工作记忆影响交错呈现优势,但工作记忆并非完全通过影响注意从而影响交错呈现优势。     

ALCOVE: an exemplar-based connectionist model of category learning.

ALCOVE (attention learning covering map) is a connectionist model of category learning that incorporates an exemplar-based representation (Medin & Schaffer, 1978; Nosofsky, 1986) with error-driven learning (Gluck & Bower, 1988; Rumelhart, Hinton, & Williams, 1986). Alcove selectively attends to relevant stimulus dimensions, is sensitive to correlated dimensions, can account for a form of base-rate neglect, does not suffer catastrophic forgetting, and can exhibit 3-stage (U-shaped) learning of high-frequency exceptions to rules, whereas such effects are not easily accounted for by models using other combinations of representation and learning method.     

Interference effects on preschool children's learning sets

Four experiments investigated interference in 5-year-olds' learning sets. Experiment 1 manipulated intertrial and interproblem intervals, showing that increased intertrial intervals retarded performance. Long interproblem intervals interacted with short intertrial intervals to produce performance facilitation. Experiment 2 investigated interpolations within problems with either similar or dissimilar stimulus presentations. Similar stimuli caused greater performance decrements than dissimilar. Experiment 3 hypothesized that the within-subjects design of Experiment 1 masked interproblem interval effects. Testing between-subjects established that increased interproblem delays facilitate performance. In Experiment 4 interproblem delays were further tested in a series of shifts within and between dimensions. Increased delays more greatly benefited shifts within dimensions. It is concluded that intertrial and interproblem intervals affect learning set performance differentially. Intertrial delays result in performance deterioration due in part to interference. Interproblem delays facilitate performance at least in part by providing the opportunity for suppression of proactive interference, and perhaps in part by providing the opportunity for constructive cognitive processing.     

The prioritization of perceptual processing in categorization

In three experiments, the effects of selective attention on perceptual processes in a complex multidimensional object categorization task were investigated. In each experiment, participants completed a perceptual-matching task to gain estimates of the perceptual salience of each stimulus dimension, then a categorization task using the same stimuli. In Experiments 1 and 2, the perceptual processing of stimulus dimensions was faster when dimensions were more diagnostic of category membership, regardless of their perceptual salience. Experiment 3 demonstrated that this prioritization of perceptual processing was evident even when stimuli were presented in unpredictable locations during categorization, indicating that the physical characteristics of the stimulus guide selective attention to diagnostic stimulus dimensions.     

The effects of relational structure on analogical learning

Relational structure is important for various cognitive tasks, such as analogical transfer, but its role in learning of new relational concepts is poorly understood. This article reports two experiments testing people’s ability to learn new relational categories as a function of their relational structure. In Experiment 1, each stimulus consisted of 4 objects varying on 2 dimensions. Each category was defined by two binary relations between pairs of objects. The manner in which the relations were linked (i.e., by operating on shared objects) varied between subjects, producing 3 logically different conditions. In Experiment 2, each stimulus consisted of 4 objects varying on 3 dimensions. Categories were defined by three binary relations, leading to six logically different conditions. Various learning models were compared to the behavioral data, based on the theory of schema refinement. The results highlight several shortcomings of schema refinement as a model of relational learning: (1) it can make unreasonable demands on working memory, (2) it does not allow schemas to grow in complexity, and (3) it incorrectly predicts learning is insensitive to relational structure. We propose schema elaboration as an additional mechanism that provides a more complete account, and we relate this mechanism to previous proposals regarding interactions between analogy and representation construction. The current findings may advance understanding of the cognitive mechanisms involved in learning and representing relational concepts.     

选择性注意对听觉内隐学习的影响

内隐学习被认为是人类无意识、无目的获得复杂规则的自动化过程。已有研究表明, 在人工语法学习范式下, 视觉内隐学习的发生需要选择性注意。为了考察选择性注意对内隐学习的影响是否具有通道特异性, 本研究以90名大学生为被试, 以人工语法为学习任务, 采用双耳分听技术, 在听觉通道同时呈现具有不同规则的字母序列和数字序列, 考查被试在听觉刺激下对注意序列和未注意序列构成规则的习得情况。结果发现：只有选择注意的序列规则被习得, 未选择注意的序列规则未能被习得。研究表明：在人工语法学习范式下, 只有选择注意的刺激维度能够发生内隐学习。选择性注意对内隐学习的影响具有跨通道的适用性, 不仅适用于视觉刺激, 也同样适用于听觉刺激。