Groups of six cats were trained, with noncorrection or with guidance, for 1200 trials on each of four ratios of reinforcement: 70:30, 60:40, 40:60 and 30:70, in a spatial probability learning experiment. Both groups eventually learned to choose the more frequently rewarded stimulus on almost every trial, the noncorrection Ss learning to do so much sooner than the guidanc e Ss. The results of this experiment with cats tend to agree with previous observations that rats and monkeys maximize on spatial probability problems. The available evidence, however, indicates that cats tested with guidance on spatial probability problems learn to maximize considerably more slowly than rhesus monkeys and that monkeys are, in turn , inferior to rats. Bitterman (1965a, b) maintains that vertebrate classes differ qualitatively in performance on probability learning tasks. This view implies that there should be no major qualitative differences among species of the same class. The available data are, however, too limited to provide a realistic estimate of intraclass variability; the mammals, for example, are represented only by rats and monkeys. The purpose of the present experiment was to determine whether the choice behavior of cats on a spatial probability problem is qualitatively similar to that of rats and monkeys tested under analogous conditions. This investigation was motivated by the observation of rather impressive differences between the performance of cats and monkeys in visual probability learning (Warren & Beck, 1966).
MethodTwelve mature cats, nine mongrel and three Siamese, were studied. All of the Ss had previously served in learning set experiments, and all had been trained on visual probability problems, but none had participated in any experiment on spatial probability learning before this experiment.The cats were tested in the WGTA. The manipulanda were identical 2.5 x 4.0 rectangles cut from 3/4 in. lumber which were painted dark gray, and presented on a medium gray test tray, containing two food wells 12 in. apart.The Ss were tested 1200 trials on each of four ratios of reinforcement in the following sequence: 70:30, 60:40, 40:60, and 30:70, with 50 trialS' being given in each test session and five sessions per week. The 12 Ss were assigned to two groups of six animals each, the Noncorrection and Guidance groups. The Noncorrection group was trained with a strict non correction procedure throughout the experiment; S was allowed only a single response on each trial. For Guidance Ss each trial ended with a reinforced response. When the cat's initial response on a given trial was incorrect, the opaque screen of the WGT A was lowered, the object on the nonrewarded side of the tray was positioned behind the empty food well, and the baited food well remained covered with its object. The test tray was then presented to S, and the trial ended only when it displaced the object covering the rewarded food well and secured the reinforcement. Only initial responses were considered in analyzing the results.For half the Ss...