Journal ol Experimental Psychology 1966, Vol. 72, No. 5, 649-654 CONCEPT IDENTIFICATION AS A FUNCTION OF PROBABILITY OF POSITIVE INSTANCES AND NUMBER OF RELEVANT DIMENSIONS 1 ROGER W. SCHVANEVELDT University of Wisconsin Probability of positive instances (.125, .250, or .500), the number of relevant stimulus dimensions (1, 2, 3, or 4), and the sex of S were factorially combined in a concept-identification task. Trials to cri- terion were found to increase linearly with the number of relevant stimulus dimensions and loga of the probability of positive instances. The number of positive instances to criterion increased linearly with the number of relevant stimulus dimensions, and the slope of the function increased as the probability of positive instances increased. The results indicate that negative instances serve an important role in the process of identifying the concept. The results are discussed in terms of the way in which Ss process information in the task. In the typical concept-identification task, 6" is required to learn to classify a series of stimuli according to certain rules (concepts) chosen by E. The classification is often based on two cate- gories, one representing examples of the concept (positive instances) and the other category representing non- examples (negative instances). Par- ticular concepts may differ in the type of rule (e.g., conjunctive, disjunctive, or relational) that is used to combine the relevant stimulus attributes to form the concept. The exact proportion of positive instances (P) of a conjunctive concept in a set of stimuli composed of all combinations of the attributes on binary stimulus dimensions is given by P = 2- R , where R is the number of relevant stimulus dimensions. Thus with conjunctive concepts, the relative size of the subset of positive instances decreases as the number of relevant dimensions increases. 1 This research was supported in part by Predoctoral Fellowship No. MH-20,485 from the National Institutes of Health. The author is indebted to E. J. Archer and R. C. Calfee for their assistance in the preparation of the manuscript. Considerable attention has been given to the role of positive and nega- tive instances in the concept-identifica- tion task. Smoke (1933) concluded that the information conveyed by posi- tive instances is easier for 5 to use in arriving at a solution. Whitman and Garner (1963) argue that the primary function of negative instances is to define the total set of stimuli from which the particular subset (of positive instances) is drawn, and that, if the subset itself defines the larger set, then negative instances are of no use to S, and they may make the problem more difficult by adding "noise" to the stimu- lus situation. Hovland and Weiss (1953) controlled the amount of infor- mation in various stimulus series and found that 5 is more likely to discover a concept conveyed by all positive in- stances than by all negative instances or by mixed positive and negative instances. In the situation where 6" is presented with the stimuli successively and in a random order, it is a nearly impossible task to store all the information in each stimulus for comparison with subsequent stimuli. In such a situa- 649