The visual discrimination of shape by Octopus: The effects of stimulus size

Octopuses were trained in a successive situation to discriminate between vertical and horizontal rectangles: Group S was trained with rectangles of side length 5 × 1 cm., Group M with 10 × 2 cm. rectangles, and Group L with 20 × 4 cm. rectangles. The larger the shapes used, the more readily were they discriminated, both in terms of speed of learning and of the asymptote of performance. After training, each group was given transfer tests with the two pairs of rectangles not used in training and with two further pairs of 2.5 × 0.5 cm. and 40 × 8 cm. The results can be summarized in three generalizations: (1) When the size of a shape is changed, performance is worse than on the original training shape. (2) The bigger the change in proportionate size, the less transfer is shown. (3) For corresponding changes of proportionate size, there is better transfer to larger shapes than to smaller. These generalizations are supported by data from earlier experiments on the question of transfer to different sized shapes: some of these data were reworked and are presented in detail here. The theoretical implications of the results are discussed.