Constraints on the process of interresponse-time reinforcement as the explanation of variable-interval performance

Computer modelling was used to investigate the adequacy of the principle of reinforcement of single interresponse-times (IRTs) as an explanation of variable-interval performance. Variations in three components of models employing IRT reinforcement were simulated; these were (i) rules for selecting previously-reinforced IRTs from the IRT memory as "candidates" for output (acceptance rules), (ii) processes involving the transformation of previously-reinforced IRTs to generate behavioural variability, and (iii) the role of the minimum IRT value permitted by the models. In general, simulation output resembled experimental data only when (i) acceptance rules discriminated strongly against long, and in favour of short, IRTs; and (ii) when IRTs were transformed by distributions in which standard deviation was a large fraction (50-125%) of the mean, thus generating substantial IRT variability. Simulation results were analyzed by means of Herrnstein's hyperbolic equation, and in general, it was found that the output of IRT reinforcement models was insensitive to the rate of reinforcement received, as well as conforming less well to Herrnstein's equation than did experimental data. Overall, therefore, it was concluded that models of the reinforcement of single IRTs could provide an adequate simulation of VI performance only when the processes within them operated within narrow constraints. Even then, models involving the reinforcement of single IRTs could adequately simulate only those VI data in which response rate was insensitive to reinforcement rate.