Interresponse times and the structure of choice

This paper presents some analyses of available data on interresponse-time emission on concurrent variable-interval schedules arranged either continuously or as discrete components, and in which either reinforcer rates or magnitudes were varied. Ratios of frequencies of interresponse times in a series of temporal bins varied in their sensitivity to reinforcement: Sensitivity values were least for interresponse times around 0.2-0.4s, were higher in the region 0-0.2s, and were highest at values greater than 0.4s. Relative frequencies of interresponse times were greatest around 0.2-1.6s. As a result, time allocation to alternatives were mainly determined by IRTs between 0.8 and 6.4s, and least by IRTs less than 0.8s. If inter-bout times reflect time allocation to unmeasured responses, then time spent emitting unmeasured behavior may be the main constituent of time allocation on concurrent schedules. An analysis of log survivor plots supported this suggestion, and further suggested that a full understanding of time allocation in choice requires accepting that unmeasured responses do not result from a unitary process, but from many disparate contributing processes. These other behaviors can be revealed through Gaussian log-normal analyses of interresponse-time distributions.