conducted. Over 30 studies have shown that the exemplar-based generalized context model (GCM) usually provides a better quantitative account of 5–4 learning data as compared with the prototype model. However, J. D. Smith and J. P. Minda (2000) argued that the GCM is a psychologically implausible account of 5–4 learning because it implies suboptimal attention weights. To test this claim, the authors recorded undergraduates ’ eye movements while the students learned the 5–4 category structure. Eye fixations matched the attention weights estimated by the GCM but not those of the prototype model. This result confirms that the GCM is a realistic model of the processes involved in learning the 5–4 structure and that learners do not always optimize attention, as commonly supposed. The conditions under which learners are likely to optimize attention during category learning are discussed.

Research has shown that learning a concept via standard supervised classification leads to a focus on diagnostic features, whereas learning by inferring missing features promotes the acquisition of withincategory information. Accordingly, we predicted that classification learning would produce a deficit in people’s ability to draw novel contrasts—distinctions that were not part of training—compared with feature inference learning. Two experiments confirmed that classification learners were at a disadvantage at making novel distinctions. Eye movement data indicated that this conceptual inflexibility was due to (a) a narrower attention profile that reduces the encoding of many category features and (b) learned inattention that inhibits the reallocation of attention to newly relevant information. Implications of these costs of supervised classification learning for views of conceptual structure are discussed.

Several studies propose that exemplar retrieval contributes to multi-attribute decisions. The authors have proposed a process theory enabling a priori predictions of what cognitive representations people use as input to their judgment process (Sigma, for “summation”; P. Juslin, L. Karlsson, & H. Olsson, 2008). According to Sigma, exemplar retrieval is a back-up system when the task does not allow for additive and linear abstraction and integration of cue-criterion knowledge (e.g., when the task is non-additive). An important question is to what extent such shifts occur spontaneously as part of automatic procedures, such as error-minimization with the Delta rule, or if they are controlled strategy shifts contingent on the ability to identify a sufficiently successful judgment strategy. In this article data are reviewed that demonstrate a shift between exemplar memory and cue abstraction, as well as data where the expected shift does not occur. In contrast to a common assumption of previous models, these results suggest a controlled and contingent strategy shift. Keywords: exemplar memory, cue abstraction, strategy shifts, multi-attribute decisions, Sigma. 1

In addition to supervised classification learning, people can also learn categories by predicting the features of category members. One account of feature inference learning is that it induces a prototype representation of categories. Another is that it results in a set of category-to-feature rules. Because neither model provides an adequate account of existing data, we propose instead that inference learning induces an anticipatory learning strategy in which learners attend to aspects of training items they think will be needed in the future, and by so doing incidentally encode information about the category’s internal structure. The proposal is formalized by an exemplar fragment model (EFM) that represents partial exemplars, namely, those parts that are attended during training. EFM’s attention weights are approximated by eyetracking data, resulting in fewer free parameters as compared to competing theories. When people classify objects, problem solve, describe concepts, or infer missing information, they must access conceptual knowledge. Thus, the question of how people learn and represent concepts has been central to the overall mission of cognitive psychology. Researchers have developed sophisticated formal theories that explain many aspects of concept acquisition. These theories are largely based on supervised classification learning in which subjects classify items whose category membership is unknown and receive immediate feedback. Recently, to understand the interplay between how categorical knowledge is used and the concept acquired, researchers have begun to investigate a wider range of learning tasks

Reprints and permission:

learning use a gradient-based learning method, which assumes that locally-optimal changes are made to model parameters on each learning trial. This method tends to underpredict variability in individual-level cognitive processes. In addition many recent models of human category learning have been criticized for not being able to replicate rapid changes in categorization accuracy and attention processes observed in empirical studies. In this paper we introduce stochastic learning algorithms for NN models of human category learning and show that use of the algorithms can result in (a) rapid changes in accuracy and attention allocation, and (b) different learning trajectories and more realistic variability at the individual-level.

Abstract — Shepard et al. made empirical and theoretical investigation of the difficulties of different kinds of classifications using both learning and memory tasks [Shepard et al., 1961]. As the difficulty rank mirrors the number of feature dimensions relevant

In addition to traditional supervised classification learning, people can also learn categories by predicting the features of category members. It has been proposed that feature inference learning promotes the learning of more within-category information and a prototype representation of the category, as compared to classification learning that promotes learning of diagnostic information. We tracked learners ' eye movements during inference learning and found (Expt. 1) that they indeed fixated other features (even though those features were not necessary to predict the missing feature), providing the opportunity to extract within-category information. But those fixations were limited to only those features that needed to be predicted on future trials (Expt. 2). In other words, inference learning promotes the acquisition of within-category information not because participants are motivated to learn that information, but rather because of the anticipatory learning it induces. Whenever a person classifies an object, describes a concept verbally, engages in problem solving, or infers missing information, they must access their conceptual knowledge. As a result, the study of concept acquisition has been a critical part of understanding how people experience the world and how they interact with it in appropriate ways. Concept researchers have developed sophisticated formal theories that explain certain aspects of concept acquisition. These theories are largely based on the study of what has come to be known as standard supervised classification—a task that occupies the majority of experimental research in this area (Solomon, Medin, & Lynch, 1999). However, an emerging literature is focused on expanding the range of tasks that can be used to inform our models of concept acquisition. By studying different learning tasks we can understand other aspects of concept acquisition, including the interplay between category use and the type of concept

Two experiments tested the effect of prior knowledge on attention allocation in category learning. Using eyetracking, we found that (a) knowledge affects dimensional attention allocation, with knowledge-relevant features being fixated more often than irrelevant ones, (b) this effect was not due to initial attention bias to the relevant dimensions but rather gradually emerged in response to observing category members, and (c) the effect grew even after the last error trial, that is, in the absence of error. These results pose challenge to current models of knowledge-based category learning. Because of the importance of categories for human cognition, the manner in which people learn categories has received intensive study. Among many procedures, supervised classification learning has been popular, and a

Research has shown that category learning is affected by (a) attention, which selects which aspects of stimuli are available for further processing, and (b) the existing semantic knowledge that learners bring to the task. However, little is known about how knowledge affects what is attended. Using eyetracking, we found that (a) knowledge indeed changes what features are attended, with knowledgerelevant features being fixated more often than irrelevant ones, (b) this effect was not due to an initial attentional bias toward relevant dimensions but rather emerged as a result of observing category members, and (c) this effect grew even after a learning criterion was reached, that is, despite the absence of error feedback. We argue that models of knowledge-based learning will remain incomplete until they include mechanisms that dynamically select prior knowledge in response to observed category members and which then directs attention to knowledge-relevant dimensions and away from irrelevant ones. Knowledge and Attention in Category Learning 3 How Prior Knowledge Affects Selective Attention During Category Learning: