Theory of learning with constant,variable, or contingent probabilities of reinforcement

The methods used in recent probabilistic learning models to generate mean curves of learning under random reinforcement are extended to the general case in which probability of reinforcement may vary in any specified manner as a function of trials and to cases in which probability of reinforcement on a given trial is contingent upon responses or outcomes of preceding trials.This paper was prepared while the writer was in residence at the Center for Advanced Study in the Behavioral Sciences, Stanford, California. The research on which it is based was supported by a faculty research grant from the Social Science Research Council.     

Effectiveness of social reinforcement as a function of contingent and noncontingent satiation

The social satiation effect is the inverse relation between the availability of a social stimulus and its subsequent efficacy in a reinforcing role. According to a suggested cognitive-interactive theory, the satiation effect is mediated by children's attributions of contingency between their own behavior and the experimenter's actions in the satiation treatments. Perceived contingencies depend, at least to an extent, on actual contingencies, and it was therefore predicted that a satiation effect would be observed only for groups presented in the satiation treatment with noncontingent social stimuli but not for groups presented with contingent stimuli. Middle-class 5- and 7-year-old children were subjected to a 10-min waiting period in which the stimulus word “Yafeh” (“good” in English) was presented 2 or 20 times, contingently or noncontingently. They were then given a 75-trial binary discrimination test: correct responses were reinforced with “Yafeh”. The hypothesis was confirmed in the analysis of variance. However, the predicted difference between the slopes of the contingency and noncontingency conditions was found clearly only in the older sample, while the younger children were more influenced by the number of social stimuli presented in the treatment (satiation) and less influenced by the method of stimulus presentation.     

Effects of delayed reinforcement in a concurrent situation

Pigeons were trained to peck either of two response keys for food reinforcement on equated aperiodic schedules. Delays of reinforcement for pecks at one key reduced the relative frequency of pecking exponentially as a function of the delay interval. Similar functions were obtained when other dependent variables were plotted against the delay interval.     

Lunchroom noise control using feedback and group contingent reinforcement

Reduction of noise levels in an elementary school lunchroom was examined as a function of feedback and feedback plus reinforcement using group contingency procedures. Feedback consisted of signals from a traffic light, with green indicating acceptable levels, yellow indicating levels approaching undesirable, and red indicating unacceptable levels. Other behaviors, including running, hitting, pushing, and kicking, were measured incidentally. Results indicate that feedback plus reinforcement was effective in reducing noise levels. Feedback only was also effective, but to a lesser degree. No response-response relationship was found to exist between noise level and the other behaviors.     

Secondary reinforcement and rate of primary reinforcement

Four pigeons were trained to peck at either of two response-keys. Pecking at either key occasionally produced a secondary reinforcer. Then, in the presence of the secondary reinforcer, further pecking occasionally produced the primary reinforcer, food. The relative rate at which each pigeon pecked to obtain a secondary reinforcer equalled the relative rate of primary reinforcement in its presence.     

Comparisons of stimulus learning and response learning in a punishment situation

These experiments use a procedure in which a rat is trained to make two topographically distinct responses on a single manipulandum, and then one of the responses is punished. Differential suppression of the punished response is taken as evidence of response learning, whereas the common suppression of both responses is attributed to stimulus learning or to general and nonassociative factors. Thus, this procedure begins the experimental separation of animals learning about the consequences of their behavior and animals learning what happens in a particular environment. A further separation is effected by using two such manipulanda; this procedure begins to distinguish between stimulus learning about the manipulandum and the more gereral factors that cause suppression. Some parameters affecting the relative importance of stimulus learning and response learning are examined.     

The effects of response contingent reinforcement and response contingent punishment upon the frequency of stuttered verbal behavior

Twelve adult subjects, 4 in each experimental condition, were exposed to 2 punishment contingencies and 1 punishment and reinforcement combined contingency, under a procedure of stimulus control. Results indicated all 3 conditions to be equally effective in the reduction of stuttering frequency. Spontaneous recovery during stimulus control segments was not observed following any of the three experimental contingencies. Implications for research and therapy are discussed.     

Discrimination learning in a foraging situation

Rats were allowed to forage in a simulated natural environment made up of eight food sources (patches) each containing a fixed number of pellets. Two of the eight contained an extra supply of peanuts. The peanut patches were signaled by an olfactory/visual cue located at the bottom of the ladder leading to the patch. In successive phases the number of sessions per day, height of the patches, and availability of peanuts were manipulated. Subjects showed evidence of discrimination learning under these conditions, although the degree of discriminatory behavior varied as a function of environmental manipulations. Assessment of behavior within foraging sessions showed that subjects systematically changed their patterns of utilization of patches across time. Sampling or exploration, as well as food reinforcement, seem implicated in these results.     

Modification of extreme social isolation by contingent social reinforcement

Two socially isolated patients were placed on a program where social reinforcement from staff members was made dependent on social interaction with other patients or staff members. This procedure was imposed in a setting where only limited control was possible. The level of social interaction and a concomitant alternate behavior in each patient was increased when the contingency for social reinforcement was imposed. The study provides another example of the efficacy of social reinforcement where there is little control over other reinforcers. Implications for use of similar procedures to increase generalization in the community are discussed.     

Instrumental and contingent saccharin licking in rats: Response deprivation and reinforcement

Instrumental licking of .4% saccharin solution was increased by the contingent opportunity to lick a less-preferred saccharin solution when the contingency schedule deprived the subject of the contingent response, but not otherwise. Schedules that imposed comparable amounts of response deprivation produced comparable increases in instrumental responding. The results support the hypothesis that instrumental responding will increase if and only if the contingency schedule deprives the subject of the contingent response. They also support the implication that the predicted increase will occur even if the contingent response has a lower operant level than the instrumental response.     

Competition between noncontingent and contingent reinforcement schedules during response acquisition

We examined the extent to which noncontingent reinforcement (NCR), when used as treatment to reduce problem behavior, might interfere with differential reinforcement contingencies designed to strengthen alternative behavior. After conducting a functional analysis to identify the reinforcers maintaining 2 participants' self-injurious behavior (SIB), we delivered those reinforcers under dense NCR schedules. We delivered the same reinforcers concurrently under differential-reinforcement-of-alternative-behavior (DRA) contingencies in an attempt to strengthen replacement behaviors (mands). Results showed that the NCR plus DRA intervention was associated with a decrease in SIB but little or no increase in appropriate mands. In a subsequent phase, when the NCR schedule was thinned while the DRA schedule remained unchanged, SIB remained low and mands increased. These results suggest that dense NCR schedules may alter establishing operations that result in not only suppression of problem behavior but also interference with the acquisition of appropriate behavior. Thus, the strengthening of socially appropriate behaviors as replacements for problem behavior during NCR interventions might best be achieved if the NCR schedule is first thinned.     

Response rate and reinforcement rate in Pavlovian conditioning

Four experiments used delay conditioning of magazine approach in rats to investigate the relationship between the rate of responding, R, to a conditioned stimulus (CS) and the rate, r, at which the CS is reinforced with the unconditioned stimulus (US). Rats were concurrently trained with four variable-duration CSs with different rs, either as a result of differences in the mean CS-US interval or in the proportion of CS presentations that ended with the US. In each case, R was systematically related to r, and the relationship was very accurately characterized by a hyperbolic function, R = Ar/(r +c). Accordingly, the reciprocal of these two variables-response interval, I (= 1/R), and CS-US interval, i (= 1/r) - were related by a simple affine (straight line) transformation, I = mi+b. This latter relationship shows that each increment in the time that the rats had to wait for food produced a linear increment in the time they waited between magazine entries. We discuss the close agreement between our findings and the Matching Law (Herrnstein, 1970) and consider their implications for both associative theories (e.g., Rescorla & Wagner, 1972) and nonassociative theories (Gallistel & Gibbon, 2000) of conditioning. (PsycINFO Database Record (c) 2011 APA, all rights reserved).     

The effects of contingent music and differential reinforcement on infantile colic

Infantile Colic is a behavioral syndrome characterized by paroxysms of excessive crying and increased motor activity, hypertonicity of the musculature, excessive flatus and erratic sleeping and feeding patterns. Ten to 40% of all infants are diagnosed as having Colic. The medical evidence to date does not justify any conclusions regarding the etiology or treatment of Colic. Typically, parents are advised to simply wait until Colic has run its course which is often 3-4 months. The behavioral program reported here was designed to increase behavior that competed with crying by reinforcing quiet alertness with music and parental attention. Concurrently, it attempted to inhibit excessive crying by a brief time-out procedure. A group of 8 infants diagnosed as Infant Colic were included in this study. The dependent variable, crying, was measured through direct observations based on hourly samples involving 30 observations of 2-min intervals. The independent variable consisted of a behavioral treatment package. A within-S reversal design was used to assess the functional properties of the treatment. The results show that across all 8 infants the introduction of the treatment package led to a substantial decrease in excessive crying of about 75% of the initial baseline. Further, a functional relationship was identified between the treatment and excessive crying behavior: crying decreased when the treatment was initially introduced, it resumed when the treatment was withdrawn and decreased again when the treatment was reinstated.