Stimulus control

Operant stimulus control is typically established by discrimination training. For example, to make a light control a pigeon's pecks on a button, reinforcement only occurs following a peck to the button. Over a series of trials the pecking response becomes more probable in the presence of the light and less probable in its absence, and the light is said to become a discriminative stimulus or S^D.[6] Virtually any stimulus that the animal can perceive may become a discriminative stimulus, and many different schedules of reinforcement may be used to establish stimulus control. For example, a green light might be associated with a VR 10 schedule and a red light associated with a FI 20-sec schedule, in which case the green light will control a higher rate of response than the red light.

After a discriminative stimulus is established, similar stimuli are found to evoke the controlled response. This is called stimulus generalization. As the stimulus becomes less and less similar to the original discriminative stimulus, response strength declines; measurements of the response thus describe a generalization gradient.

An experiment by Hanson (1959)[7] provides an early, influential example of the many experiments that have explored the generalization phenomenon. First a group of pigeons was reinforced for pecking a disc illuminated by a light of 550 nm wavelength, and never reinforced otherwise. Reinforcement was then stopped, and a series of different wavelength lights was presented one at a time. The results showed a generalization gradient: the more the wavelength differed from the trained stimulus, the fewer responses were produced.[7]

Many factors modulate the generalization process. One is illustrated by the remainder of Hanson's study, which examined the effects of discrimination training on the shape of the generalization gradient. Birds were reinforced for pecking at a 550 nm light, which looks yellowish-green to human observers. The birds were not reinforced when they saw a wavelength more toward the red end of the spectrum. Each of four groups saw a single unreinforced wavelength, either 555, 560, 570, or 590 nm, in addition to the reinforced 550 wavelength. The birds were then tested as before, with a range of unreinforced wavelengths. This procedure yielded sharper generalization gradients than did the simple generalization procedure used in the first procedure. In addition, however, Hansen's experiment showed a new phenomenon, called the "peak shift". That is, the peak of the test gradients shifted away from the S^D, such that the birds responded more often to a wavelength they had never seen before than to the reinforced S^D. An earlier theory involving inhibitory and excitatory gradients partially explained the results,[8] A more detailed quantitative model of the effect was proposed by Blough (1975).[9] Other theories have been proposed, including the idea that the peak shift is an example of relational control; that is, the discrimination was perceived as a choice between the "greener" of two stimuli, and when a still greener stimulus was offered the pigeons responded even more rapidly to that than to the originally reinforced stimulus.[10]