Weak stimulus generalization using sinusoidal gratings: A cautionary note in animal psychophysics

Psychophysical techniques involving operant and classical conditioning are used commonly to determine the visual thresholds of nonverbal subjects. Typically, subjects are conditioned to respond to a particular stimulus, and once trained sufficiently, the subject’s sensitivity to this and similar stimuli can be determined by decreasing the stimulus intensity until the animal no longer responds. However, this assumes that the animal’s tendency to respond correlates with its sensory abilities. Stimulus generalization to sinusoidal gratings of different spatial frequencies was examined in goldfish. Subjects were classically conditioned to suppress respiration upon presentation of a sinusoidal grating. Animals’ learning curves to each spatial frequency grating were compared in order to determine whether the animals generalized across stimuli. Results indicate that fish show weak stimulus generalization across spatial frequencies. Also, the shape of the contrast sensitivity function, a common measure of the animal’s visual capabilities, is distorted by insufficient training. We conclude that when the goal is to obtain accurate data on visual capabilities, either subjects should be trained to each stimulus to be tested or their generalization gradients should be measured.