Discriminability and preference in concept identification

Abstract

The hypothesis tested in this study was that the relative discriminability of dimensions would determine the rate of concept identification (CI) by determining the order of preference for testing the relevance of dimensions. Areaction time measure of discriminability and the Levine procedure for identifying hypotheses were employed. The results indicated that discriminability had no simple relationship to preference and that neither relative discriminability nor relative preference was predictive of CI rates. An explanation of the results in terms ofmultiple hypotheses testing is offered.The effect of selective attention variables on rate of concept identification (CI) has recently been a subject of increasing interest. The concept of selective attention implies that organisms may not attend to or respond to aß aspects of the stimulus environment but only to certain selected aspects. Observations that the rate of CI is often dependent upon the dimension that is relevant to the solution suggests that the probability of a cue being attended to will determine, in part, the ease of CI when that cue is relevant. Various investigators have postulated attention factors such as discriminability (Archer, Boume, & Brown, 1955), obviousness (Sechrest & Kaas, 1965;and Archer, 1962), and the ratio of the weight of relevant cues to the weight of aß cues (Boume & Restle, 1959; RestIe, 1962;and Bower & Trabasso, 1964).Inherent in these studies is the suggestion that the obviousness or discriminability of the dimensions of the stimuli will determine the preference of Ss for utilizing the dimensions in CI. Imai & Gamer (1965) have attempted to determine the relationship between discriminability and preference in a card-sorting task. Oiscriminability was defined as the physical distance between two levels of a dimension and preference was defined as the proportion of times Ss used a dimension as the basis of c1assification when they could employ any dimension (free c1assification). Their resuIts indicated a relationship between discriminability and preference such that the greater the discriminability, the more probable it was that the dimension would be used for c1assification. When Ss were told wh at dimension to use for c1assification (constrained c1assification), only the discriminability affected the speed of c1assification; preference had no effect on speed.. The Imai and Gamer resuIts may not be directIy applicable to a CI problem, because CI does not involve completely free nor completely constrained c1assification. However, if measures of discriminability and of original preference for dimensions during CI are taken, observation of their effects on speed of CI would be appropriate.The following study utilized areaction time (RT) measure of relative discriminability and a dimension preference measure to test the hypothesis that the relative discriminability of dimensions will determine the preference for dimensions, thus determining the order of hypothesis testing and the rate of CI. The most discriminable dimension sh...

Requiring hypotheses and the identification of unidimensional conjunctive and disjunctive concepts.

“…However, a repetition of the initial analysis after a square root transformation of the original data yielded similar outcomes. Leonard (1969) and Downing (1969). Stimulus set and dimensional preference were ignored in further analyses.…”