| dc.description.abstract | The investigation of variables controlling interresponse times (IRTs) has occasionally produced certain unexplained results. Because the IRT itself often is omitted as a variable in this investigation, the question of the effect of a prior IRT upon the subsequent IRT (the sequential IRT dependency) arises. The basic problem concerns the dual role of the IRT, i.e., as both a dependent and independent variable. To study this problem, a series of six experiments are presented. The first two experiments, dealing with the DRL, or differential reinforcement of low rates schedule (IRTs > t sec are reinforced), illustrate the difference between a sequential and nonsequential analysis of IRTs. The third experiment compares the differential performance of subjects with respect to frequency of shock on the same free-avoidance schedule. Although shock appears to be directly related to the degree of association between prior and subsequent IRTs, the results of the next study indicate that the dependency is not entirely a matter of the reinforcement event. This fourth experiment deals with the change in the probability of the subsequent IRT when the reinforcement event of a DRL schedule fails to occur in its normal relationship with a particular prior IRT. In the fifth experiment, two drl components of a mixed drl 5-sec and drl 15-sec schedule alternate in fixed sequence. The schedule is found to generate a high degree of sequential dependency in which a reinforced short IRT (or long IRT) is the occasion for a subsequent long IRT (or short IRT). When reinforcement fails to occur in association with the prior IRT, the sequential dependency vanishes. In the last experiment, a tandem drl 5-15 sec schedule, conditions are found which maintain sequential dependency even in the absence of reinforcement in the drl 5-sec component. The multiple component schedules indicate that the sequential dependency is maintained by a complex interaction of reinforcement and the length of the prior IRT. These factors are incorporated into a matrix representation which is used to distinguish both the controlling effects and the response properties of the IRT. The success of this description rests upon the segmentation of the sequential structure into non-overlapping, prior temporal states. The task requires a nonarbitrary procedure which will band the temporal continuum, a dimension capable of infinite segmentation, into appropriate class intervals. A clustering technique, based upon distance between pairs of points in n-probability space, was developed to facilitate this banding task. The points represent IRT distributions of a 26x20 transition matrix. Three non-overlapping clusters of points in then- space which are related to the properties of the mixed schedule are isolated. This clustering procedure is found to be a useful and powerful device in the study of sequential IRT dependency. | |
