Effects of Manipulation of Attributes on Efficiency of Concept Formation

In the typical concept formation experiment, instances of what a concept is and/or of what it is not are presented to Ss, and Ss' task is to determine the defining attributes of the concept. Ss may be required to make a response after each instance, indicating whether or not the instance represented the concept. Often several concepts are acquired simultaneously and Ss indicate after each instance which of the concepts has been presented by pressing an appropriate button (e.g., Archer, Bourne, & Brown, 1955) or giving an appropriate nonsense syllable (e.g., Heidbreder, 1947). In some studies no responses are made during a series of instances that define a single concept. At the conclusion of the series Ss attempt to define the concept (Hovland & Weiss, 1953; Glanzer & Huctenlocher, 1960). A major variation from these procedures with potential for increasing our understanding of inductive reasoning processes was introduced in a series of experiments carried out by Bruner, Goodnow, and Austin (1956). These authors studied the acquisition of single concepts and used a type of response that differed from those cited above. An entire set of instances containing the permutations of values for the attributes used was displayed to Ss. Ss selected instances and received information as to whether or not these were instances of the concept. Since Ss' responses, i.e., selection of instances, determined the succession of instances presented, their strategies for taking in information could be examined. While using a procedure similar to that of Bruner, Goodnow, and Austin to investigate che development of inductive reasoning in children, the present author noted that Ss of a given age seemed to have more difficulty acquiring a concept if they manipulated the attributes than if they were presented a series of instances. This observation was not made under controlled conditions and was confounded by the fact that manipulations carried out by Ss often involved redundant steps whereas those presented by E did not. The present study is designed to discover whether solution of inductive reasoning problems may be adversely affected when Ss conduct their own "experiment" of locating the defining attributes of concepts. METHOD