Stimulus Control Glenn Mason-Riseborough (15/9/1997) Stimulus control is an area of research which studies the relationship between antecedent stimuli and the consequent behaviours. These behaviours will include such characteristics as vigour, rate of occurrence and temporal patterning. Stimulus control may be seen as a measurement of the amount of generalisation exhibited in the behaviour of a subject when it responds to different stimuli. This can be seen as a continuum such that generalisation occurs when the subject responds to new stimuli in the same manner as it did to stimuli it has previously encountered, whereas discrimination occurs when certain behaviours occur only over a small range of stimuli. Three types of training procedures for stimulus control are considered; these are single stimulus, interdimensional and intradimensional discrimination training. Single stimulus training uses only one stimulus (S+) with some schedule of reinforcement (typically VI) to train the subject. Both interdimensional and intradimensional discrimination training use a second stimulus (S- ) in addition to S+. S- is used to train the subject’s response to that stimulus to extinction, thus allowing the subject to form some basis for discrimination between S+ and S-. Interdimensional discrimination uses an S- stimulus that is not on the same dimension as S+ (e.g. S+ is light and S- is tone) whereas intradimensional discrimination uses an S- stimulus on the same dimension as S+ (e.g. S+ is 560 nm and S- is 600 nm wavelength of light). Stimulus control is measured using a generalisation test. Typically this test will be a transient generalisation-test procedure which is carried out in extinction. This test will consist of testing the subject using a number of stimuli with a range of values (both greater and less than S+) on the same dimension as S+. The responses are then graphed as a function of the stimulus to show the generalisation gradient. Generalisation gradients may be either absolute or relative. Absolute generalisation gradients are based on the total number of responses, whereas relative generalisation gradients are based on the percentage of responses (either total responses or only S+ responses). Flat generalisation gradients show that there is no discrimination between the stimuli on that dimension, whereas a decremental generalisation gradient (peaked at S+) shows that the subject learned discrimination. Peak shift is seen on a generalisation gradient when the maximum (minimum) number of responses does not occur at S+ (S-). Positive peak shift is indicated by a movement of the peak away from S+ opposite to S-; negative peak shift is indicated by a movement of the trough away from S- opposite from S+. A number of experiments show evidence of peak shift (e.g. Thomas & Williams, 1963 cited in Beale et. al., 1990). Behavioural contrast occurs when a change in one schedule of reinforcement alters the response rate for another schedule of reinforcement. A positive behavioural contrast occurs when there is an increase in the response rate of S+ due to a decrease in rate of reinforcers of S-. Conversely, negative behavioural contrast occurs when an increase in reinforcer rate of S+ causes a decrease in response rate of S-. Evidence for behavioural contrast can be seen in a number of experiments (e.g. Gutman, 1977 cited in Mazur, 1994) There are two conflicting theories regarding what an animal learns when discrimination between stimuli occurs. An absolute view suggests that the animal learns the particulars of each stimulus, and regardless of the situation and accompanying stimuli, the animal will always respond in the same way to that stimulus. On the other hand, the relational view predicts that it is the relationship between the stimuli that is learned (e.g. choose the brighter of the two light sources), this is often called transposition. Spence (1937, cited in Davidson & Jones, 1997) developed an absolute theory to account for peak shift and transposition (both relational stimulus control). This theory proposed the existence of both excitatory and inhibitory gradients which peak at S+ and S- respectively. Subtracting the inhibitory gradient from the excitatory gradient results in a curve corresponding to the generalisation gradient. Thus, Spence’s theory predicts that the generalisation gradient from interdimensional discrimination training will be similar to the generalisation gradient from single stimulus training (because S- is on a different dimension). Conversely, the generalisation gradient from intradimensional discrimination training will exhibit positive peak shift (because of the presence of the inhibitory gradient on the same dimension). Spence’s theory also predicts that the maximum height of the absolute generalisation gradient of the intradimensional discrimination training should be less than the maximum height of the absolute generalisation gradient of the interdimensional discrimination training. Finally, Spence’s theory predicts that the generalisation gradient of the intradimensional discrimination training is steeper between S+ and S- than the generalisation gradient of the interdimensional discrimination training. The present experiment tested these predictions with a visual intensity continuum and used a black square on green for S- on the intradimensional training. The experiment consisted of four parts plus two generalisation tests. The first two parts trained for interdimensional discrimination training, whilst the second two trained for intradimensional discrimination training. Parts one and three used multiple VI 30s VI 30s schedules, whilst parts two and four used multiple VI 30s EXT schedules. The two generalisation tests were conducted in extinction after the second and fourth parts respectively. If Spence’s theory holds, then we should expect that the generalisation gradients from both the generalisation tests to be different. Specifically, we would expect that the generalisation gradient from the second generalisation test to exhibit peak shift (whereas the generalisation gradient from the first generalisation test would not), to have a lower maximum peak than the generalisation gradient from the first generalisation test, and to be steeper than the generalisation gradient from the first generalisation test. References: Beale, I., Davison, M, Alsop, B., & Elliffe, D. (1990). Reinforcement and stimulus control. Davidson, M., & Jones, M. (1997). Psychology 461.250 SC Animal learning and behaviour course notes 1997. Unpublished Manuscript, University of Auckland. Ernst, A. J., Engberg, L., & Thomas, D. R. (1971). On the form of stimulus generalization curves for visual intensity. Journal of the Experimental Analysis of Behavior, 16, 177- 180. Guttman, N., & Kalish, H. I. (1956). Discriminability and stimulus generalization. Journal of Experimental Psychology, 51, 79-88. Mazur, J. (1994). Learning and behaviour (3rd ed.). Englewood Cliffs, NJ: Prentice-Hall. Terrace, H. S. (1968). Discrimination learning, the peak shift, and behavioral contrast. Journal of the Experimental Analysis of Behavior, 11, 727-741.