Association for Behavior Analysis International

Acquisition of operant behavior depends on selection from behavioral variations. As discussed by Stahlman, variations result from two main sources, direct reinforcement of variability itself and induction of variability resulting from such sources as Pavlovian relationships and extinction. Nergaard & Holth describe a novel procedure (with rats) and question the very possibility of reinforced or "operant" variability. They attribute observed variability to cycles of reinforcement and extinction of individual responses. Quite differently, Galizio, Haynes, Frye & Odum continue an extensive line of research showing that response variability is a behavioral dimension controlled by contingencies in many ways analogously to other such operant dimensions as response rate and force. They explore (with pigeons) spontaneous recovery of previously reinforced variable responding. Abreu-Rodrigues and Carmona extend research on how variability contingencies influence preferences. They explore (with pigeons) choices for different run length of varying sequences. This symposium provides innovative research and interpretations of response variability with diverse theoretical views from diverse locales -- Brazil, Norway and the US. Neuringer, as discussant, will attempt to relate the four presentations to additional basic operant-variability research and its application in education and treatment of autism.

Research by Neuringer and colleagues has reliably demonstrated that behavioral variability increases when reinforcement is contingent on it. An alternative summary of the contingencies that typically prevail in variability experiments is that several responses cycle between reinforcement and extinction. The current Experiment 1 tested this by making reinforcement contingent upon predetermined but shifting responses. Resulting U values were similar to those obtained in comparable lag schedules. Second, the details of what happens when reinforcement is contingent on variable responding may be more transparent when the descriptive operant class consists of responding to several different operanda rather than of different sequences of responses on two operanda. In Experiment 2, we also modified the lag n schedule so that reinforcement was contingent on the occurrence of a response to an operandum that differed from the n previously operated operanda rather than just the n previous responses. The results showed that (1) the rats only emitted as many different responses as the lag schedule required, (2) when the lag increased and, thus, required a response to an additional operandum, such an additional response emerged, and next (3) rather than demonstrating reinforced variability, the rats fairly consistently repeated the last reinforced response.

Choice between differing varying contingencies is affected by the variation requirement as well as by the cost of responding (i.e. switching between operanda). The present study investigated whether such choices would also be affected by within-sequence discriminative control. This possibility is supported by numerosity studies which show an inverse relation between sequence accuracy and run length, i.e., number of responses to be emitted on one operandum before switching to the other. Pigeons were exposed to a pre-training phase in which control by numerosity was established. By the end of this phase, the animals had to emit 10-response sequences according to a multiple schedule. To be reinforced, the sequence had to differ from the previous two correct ones and contain run lengths of 1 to 4 responses (Max 4-Lag 2 component) or 6 to 9 responses (Min 6-Lag 2 component). Accuracy was close to 50% in both components. Next, a concurrent-chain schedule in which the Max 4-Lag 2 and Min 6-Lag 2 contingencies operated in the terminal links. Preference for the shorter run length was observed for the pigeon that is already in this phase, despite equal reinforcer probabilities across terminal links, suggesting that discriminative control by numerosity affects choice.