Year: 2000
Type: article
Abstract: Previous research suggests that categories learned through classification focus on exemplar information, while categories learned by making predictive inferences focus on summary (i.e., prototype) information.To test this idea further, we demonstrated that it is more difficult to learn nonlinearly separable categories by making inferences than by classifying.This research also supports previous studies by indicating that different processes are likely to mediate inference and classificationIn this paper, we examine the type of categorical information people assess in the process of obtaining inductive knowledge.Specifically, we investigate the extent to which abstract summary information about a category and specific information about individual exemplars of a category are used to make feature inferences.Categories license inference in at least two ways.First, categories provide a summary representation of their members (e.g., a prototype).Given an unknown feature of a bird, for example, people may predict the value of that feature by referring to the bird prototype (Rips, 1975;Tversky & Kahneman, 1974; Yamauchi & Markman, 2000, in press).Another source of category-based induction comes from individual exemplars of a category.Many studies have shown that people classify items by retrieving information about specific exemplars from memory (Kruschke, 1992;Medin & Schaffer, 1978;Nosofsky, 1986).A similar process may be used to make feature inferences.In predicting an unknown feature of an item, people may predict characteristics of the new item based on exemplars stored in memory.Studies investigating classification have shown that exemplar information plays a crucial role in making classification judgments.Research on inductive inference, however, reveals that category-level abstract feature information (e.g., prototypes) is crucial for inference.For example, Anderson and Fincham (1996) demonstrated that people are capable of predicting the value for one feature given the value of another, based on the overall correlation between the features in the study phase of the experiment, rather than on the basis of seeing those specific values during the study phase.Yamauchi and Markman (2000) further showed that varying the appearance of exemplars during learning disrupts classification, but not inference.These findings suggest that, while classification and inference may be formally equivalent, they make use of different kinds of information in practice (Yamauchi & Markman, 1998).In this paper, we extend this hypothesis and examine the idea that category-level summary information provides a basis for inference (e.g., prototypes), while exemplar information plays a major role in classification.In the following sections, we describe the inference and classification tasks that were employed in our experiments.Then, we examine the role of exemplar and prototype information in two experiments.
Authors: Takashi Yamauchi, Arthur B. Markman
Language: English
Subfield: Experimental and Cognitive Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
Funder: National Science Foundation
OpenAlex ID: W2564072317