Generalization gradients and stimulus control in delayed matching-to-sample.

Neurological patients were subjects in delayed visual matching-to-sample. The sample and choice stimuli were ellipses of varying size. By measuring the difference in size between the sample on a given trial and the ellipse the subject chose on that trial, gradients of differences between samples and choice stimuli could be plotted. These difference gradients broadened with increasing delays. Sharp gradients were controlled by the samples. Flat gradients were controlled by features of the choice display, independently of the samples. Intermediate gradients reflected combined control by the samples and by the choice displays.