The Properties of Reinforcement: Strengthening Versus Signalling

Abstract

The properties of reinforcement are currently being debated in operant psychology. The traditional view is that reinforcers strengthen the behaviour that precedes it, whereas a modern view treats reinforcers as discriminative stimuli that an organism uses to determine which behaviour will produce the next reinforcer. This thesis approached the issue as if both properties exist, and used local analyses of choice procedures to isolate each property for observation. In Experiment 1, the signalling property was isolated by creating an environment where the previously reinforced response alternative was removed and replaced with a previously unavailable response alternative. This prevented the subjects from making the response that had just been reinforced and therefore limited the usefulness of strengthening based explanations for local response patterns. The subjects responded to those environmental contingencies by showing preference for the locally richer alternative immediately after reinforcement. When Experiment 1 was altered so that subjects could respond to the just-reinforced alternative, there was a small increase in preference for the just-reinforced alternative, but only when that alternative could produce reinforcers. In Experiment 2, the strengthening property was isolated by removing the relevance of any discriminative stimulus that was produced by reinforcement which meant subjects could use not use an individual reinforcer to guide them towards the next reinforcer. Subjects responded by showing preference for the richest alternative immediately after reinforcement rather than preferring the location of the last reinforcer, as has been observed in local analyses of previous concurrent VI VI procedures. Data from Experiment 2 were used in two additional analyses to provide support for the conclusions reached during that experiment. The first analysis involved computer simulations of responding to ensure that response patterns were caused by the effects of reinforcement, and not due to artefacts created by analysis procedures. The second analysis fitted alternative choice models to test the stability of the response data. Overall, at local levels of analysis there was more evidence to support signalling properties of reinforcement than strengthening properties, although this does not mean that strengthening properties of reinforcement do not exist.