Brian H. Ross

This Is Like That: The Use of Earlier Problems and the Separation of Similarity Effects

Novices attempting to solve a problem often are reminded of an earlier problem that illustrated a principle. Two experiments examined how these earlier problems are used and how this use is related to these remindings. Subjects studied four probability principles with related word problems. Test problems varied in their similarity to the study problems on story lines, objects, and correspondence of objects (variable roles). Experiment 1 tested whether remindings cue the principle or serve as the sources of detailed analogies. When the appropriate formula was provided with each test, the similarity of story lines had no effect, but object correspondences had a large effect. These results support an analogical account in which mapping is affected by the similarity of objects between study and test problems. Experiment 2 began to separate the aspects of similarity affecting the access and use of earlier problems by showing that, with confusable principles, similar story lines increased the access, but did not affect the use. The access appears to be sensitive to the relative similarity of examples because with distinctive principles, similar story lines had little effect. Discussion focuses on the further specification of the processes of noticing and analogical use of earlier problems.

Remindings and Their Effects in Learning a Cognitive Skill.

Abstract The early practice of cognitive skills includes many instances of remindings, the memory retrieval of earlier learning episodes. This paper provides experimental demonstration of remindings during the early phase of learning and examines their effect on performance. The effects of practice and difficulty are also examined. In Experiment 1, subjects learned to use a computer text editor. For each operation, two methods were taught, each with a different irrelevant context (e.g., type of textual material). At test, one of these contexts was reinstated. Overall, there was a tendency to use the method learned with similar material, but much of this “consistency” effect was due to tests in which the subjects mentioned (i.e., were reminded of) a learning exercise. In Experiments 2 and 3, subjects learned probability theory by studying abstract information and an example. The test example content might be the same as the study example content for that principle ( appropriate ), a content not shown before ( unrelated ), or a content studied with one of the other principles ( inappropriate ). In Experiment 2, the appropriate condition led to highest performance and the inappropriate condition led to lowest performance, as predicted. In Experiment 3, the same three conditions were used, but three of the principles used the same study example content, an interference manipulation. The facilitation effect found when the study and test examples had the same content was attenuated when the content was shared, as predicted. The implications of these results for theories of cognitive skill learning are discussed and a brief sketch of a theory is provided.

Category Use and Category Learning

Categorization models based on laboratory research focus on a narrower range of explanatory constructs than appears necessary for explaining the structure of natural categories. This mismatch is caused by the reliance on classification as the basis of laboratory studies. Category representations are formed in the process of interacting with category members. Thus, laboratory studies must explore a range of category uses. The authors review the effects of a variety of category uses on category learning. First, there is an extensive discussion contrasting classification with a predictive inference task that is formally equivalent to classification but leads to a very different pattern of learning. Then, research on the effects of problem solving, communication, and combining inference and classification is reviewed.

Food for Thought: Cross-Classification and Category Organization in a Complex Real-World Domain

Seven studies examined how people represent, access, and make inferences about a rich real-world category domain, foods. The representation of the category was assessed by category generation, category ratings, and item sortings. The first results indicated that the high-level category of foods was organized simultaneously by taxonomic categories for the kind of food (e.g., vegetables, meats) and script categories for the situations in which foods are eaten (e.g., breakfast foods, snacks). Sortings were dominated by the taxonomic categories, but the script categories also had an influence. The access of the categories was examined both by a similarity rating task, with and without the category labels, and by a speeded priming experiment. In both studies, the script categories showed less access than the taxonomic categories, but more than novel ad hoc categories, suggesting some intermediate level of access. Two studies on induction found that both types of categories could be used to make a wide range of inferences about food properties, but that they were differentially useful for different kinds of inferences. The results give a detailed picture of the use of cross-classification in a complex domain, demonstrating that multiple categories and ways of categorizing can be used in a single domain at one time.

Generalizing From the Use of Earlier Examples in Problem Solving

Novices often are reminded of earlier examples during problem solving. Four experiments examine the hypothesis that the use of earlier examples promotes generalizations about problem types, thereby influencing what is learned about the domain. Subjects studied four probability principles with word problems and then tried to solve two tests problems for each principle. For half of the first tests, cues indicated which study problem might be used. All second tests were uncued. Discussion focuses on the distinction about how problem comparisons are used in learning and the implications of the view that remindings lead to generalizations

CONTENT EFFECTS IN PROBLEM CATEGORIZATION AND PROBLEM SOLVING

The goal of this article is to examine how the specific content of a problem may affect the problem solving of experienced solvers. In most domains, there is an empirical correlation between problem types and problem contents. We argue that experienced problem solvers learn to make use of these formally irrelevant, but empirically predictive, contents in accessing and applying their relevant knowledge. Understanding these content effects is important not only because content affects performance, but also because these effects provide clues as to how experienced solvers represent and use this relevant knowledge. We first provide some background on problem content and expertise and then return to this issue.

The Roles of Automatic and Strategic Processing in Sensitivity to Superordinate and Property Frequency

Four experiments with 183 undergraduates examined the ability to estimate frequency for 2 kinds of nonpresented information automatically activated by presented items. Exp I showed that Ss were sensitive to how many presented items belonged to the same superordinate category (superordinate frequency) but were not sensitive to how many items shared the same context-independent property (property frequency). Exp II indicated that sensitivity to superordinate frequency did not result from strategically preparing for free recall. Exp III demonstrated that (1) conflicting strategic processing reduced but did not eliminate sensitivity to superordinate frequency and (2) appropriate strategic processing enabled sensitivity to property frequency. Findings of Exp IV show that experimental training made strategic processing unnecessary for sensitivity to property frequency and suggest that automatic processing only results in frequency sensitivity for nonpresented information whose well-established representations are of a certain type. It is concluded that strategic processing is not necessary for sensitivity to the frequency of nonpresented information and that sensitivity is reduced when strategic processing conflicts with well-established representations and becomes possible in the absence of well-established representations when appropriate strategic processing is used.

Reminding-Based Category Learning.

Abstract Four experiments examine a reminding-based generalization view of category learning. According to this proposal, when subjects categorize a new instance by the use of an earlier instance, they learn about the aspects common to the two instances and this knowledge may be used in later categorizations. Experiment 1 demonstrates that which earlier instance is used to categorize a new instance affects performance on a later category test. Two assumptions of this view are then tested. Experiment 2 provides evidence for the necessity of distinct instance representations to obtain this effect. Experiment 3 shows that the reminding effect is due to the presence of the common aspects and not simply to the increased availability of the instance of which one is reminded. The final experiment extends the reminding-based learning idea to a situation in which feature relevance is provided. This study shows that the reminding determines whether this relevance has an effect later on categorization time for individual features. Discussion focuses on the implications of these results for theories of category learning.

Category-based predictions: Influence of uncertainty and feature associations

Four experiments examined how people make inductive inferences using categories. Subjects read stories in which 2 categories were mentioned as possible identities of an object. The less likely category was varied to determine if people were using it, as well as the most likely category, in making predictions about the object. Experiment 1 showed that even when categorization uncertainty was emphasized, subjects used only 1 category as the basis for their prediction. Experiments 2-4 examined whether people would use multiple categories for making predictions when the feature to be predicted was associated to the less likely category. Multiple categories were used in this case, but only in limited circumstances; furthermore, using multiple categories in 1 prediction did not cause subjects to use them for subsequent predictions. The results increase the understanding of how categories are used in inductive inference.

Predicting Features for Members of Natural Categories When Categorization Is Uncertain

An important function of concepts to allow prediction of unseen features. A Bayesian account of feature prediction suggests that people will consider all the categories an object could belong to when they judge the likelihood that the object has a feature. The judgment and decision literature suggests that they may instead use a simpler heuristic in which they consider only the most likely category. In 3 experiments, no evidence was found that participants took into account alternative categories as well as the most likely one when they judged feature probabilities for familiar objects in meaningful contexts. These results, in conjunction with those of Murphy and Ross (1994), suggest that although people may consider alternative categories in certain limited situations, they often do not. Reasons for why the use of alternative categories may be relatively rare are discussed, and conditions under which people may take alternative categories into account are outlined.