Two new procedures-a four-key choice procedure and a four-ply multiple fixed ratio schedule procedure-were used to train pigeons to categorize color slides depicting natural (cat, person, flower) and human-made (car, chair) objects. In Experiments IA, IB, 2A, and 2fl, 16 pigeons trained with 10 slides from each of four categories reliably classified novel examples from these categories. However, performance was more accurate on training than on novel stimuli. In Experiment 3, 8 pigeons learned to classify 2,000 nonrepeating slides. Thus, repetitive training with a limited number of stimuli is not necessary for pigeons to learn a four-category classification task. In Experiment 4, 4 pigeons were trained with a set of repeating slides while concurrently being trained with novel stimuli. As in Experiments IA, IB, 2A, and 2e, performance here was more discriminative on repeatedly seen stimuli than on novel ones. Thus, repetition facilitates categorization, whether or not the pigeons are concurrently exposed to novel stimuli. The implications of these results for models of categorization are discussed. We conclude that the conceptual abilities of pigeons are more advanced than hitherto suspected.

This article is based on a dissertation submitted to the Graduate College of the State University of Iowa in partial fulfillment of the requirements for the Ph.D. degree. The author is indebted to Kenneth W. Spence who directed the investigation.

Let H be a normal subgroup of a finite group G, and let ζ be an (absolutely) irreducible character of H. In [7], Clifford studied the irreducible characters X of G whose restrictions to H contain ζ as a constituent. First he reduced this question to the same question in the so-called inertial subgroup S of ζ in G, and secondly he described the situation in S in terms of certain projective characters of S/H. In section 8 of [10], Mackey generalized these results to the situation where all the characters concerned are projective.

No abstract

A reversed PRE was obtained for choice behavior following T-maze training in which CRF and PR Ss were "yoked" for correct and incorrect responses. Conventional acquisition procedures yielded differences in acquisition performance and a conventional PRE during extinction. Speed data yielded conventional PREs during extinction in all cases. ProblemStudies of partial reinforcement extinction effects (PREs) in selective learning situations typically have failed to insure equivalence of behavior of CRF and PR groups during acquisition. A typical finding is that, while PR Ss make more correct responses during extinction than do CRF Ss, CRF Ss typically make more correct r esponses during acquisition (see Brunswik, 1939; Cotton et al.,1959) . Recently, Pavlik & Born (1962) reported that following a procedure which insured that CRF and PR Ss made equal numbers of correct and incorrect responses during acquisition, the CRF group showed greater resistance to extinction of correct responses than did the PR group . This finding suggests that the conventional PRE observed in selective learning situations, rather than being a direct effect ofthe schedules, may result from differences between the CRFand PR groups in numbers of correct and incorrect responses made during acquisition. The present study sheds furthe:f light on this possible confounding by comparing the effects upon the PRE of acquisition procedures which either permitted CRF and PR Ss to perform differently during acquisition or which required CRF and PR Ss to make the same numbers of correct and incorrect responses during acquisition. ~IethodThe apparatus was a single unit T maze, consisting of a 10-in start box, 24-in stem,lS-in arms, and 12-in goal boxes. Plexiglas doors at the choice point were used to prevent entry into or retracing from an arm of the maze. Photoelectric circuits provided measures of starting time, running speed in the stem, and running speed in each of the arms.Forty naive male Sprague-Dawley rats, 120-150 days old, were randomly divided into four groups. Group 100 and Group 50 received all free trials during acquisition. Group 100 Ss were reinforced for each correct response, and Group 50 Ss for an irregular 50% of correct responses. Group 100X and 50X were yoked to Group 50 during acquisition. Each S in Groups 100X and 50X was matched to a S in Group 50 for number of correct and incorrect responses during each block of 10 acquisition