Association for Behavior Analysis International

Temporal discrimination performance is typically assessed in discrete-trial procedures where the subject is asked to track how much time has elapsed since the presentation of some exteroceptive stimulus such as a brief key-colour flash. The typical finding is that following a period of training where reinforcement is consistently delivered at a fixed time (t) since stimulus presentation, response maxima (peak choice) in trials where no reinforcement is presented occurs approximately at t. Thus, it has been shown that pigeons and a variety of other species can track both the passage of time and changes over time in the probability of reinforcement. The current symposium extends these findings to free-operant procedures and furthermore explores the effects of changes in relative probabilities of food within inter-reinforcer intervals. The questions addressed by these papers include: 1) Can time since food delivery act as a discriminative stimulus for the future likelihood and key-location of food? 2) Is control by time since food delivery enhanced when changes in the contingencies of food occur rapidly or gradually? Does trial length and relative reinforcer rate affect time discrimination? The data that will be discussed relate to theories and models of choice, timing and reinforcement.

Pigeons pecked two concurrently available keys for food that was arranged according to an overall arithmetic VI 30-s schedule. At some time since food delivery, the probability of obtaining food on a key changed for a short period of time. For example, in Condition 3, 15 s after the last food delivery, the probability of food being arranged on the left key, should it be arranged by the VI schedule, became .0001 (and .9999 on the right key) for 5 s. These changes were unsignaled. Food probabilities at other times were constant and selected so that there was an overall equal food frequency on the two keys. We investigated the effects on local choice of the extremity and direction of the food peak, and the effects of periods of unsignaled extinction prior to the commencement of the food-peak. Choice was affected more when the food peak occurred sooner after prior food, and was generally unaffected by unsignaled extinction.

In a modified concurrent VI VI schedule, contingencies were arranged such that the not-just-productive key was most likely to produce a food delivery sooner than the just-productive key. The last-food location, and the time since the last food delivery, were both necessary to track changes in the local food ratio across time since food. When the last-food location was unsignaled, control by changes in the local food ratio was weak and transient, and preference stabilized at indifference after approximately 10 s into the inter-food interval. When the last-food location was signaled by a red keylight for the duration of the inter-food interval, choice closely followed changes in the local food ratio for the duration of the inter-food interval. When the last-food location was signaled for only part of the inter-food interval, changes in choice occurred shortly prior to the onset of the signal, and were also present for a short time after the signal offset. Thus, control by the local food ratio was enhanced by "reminder" stimuli, but the effect of these stimuli extended beyond the time at which they were present.

A 2-alternative free-operant psychophysical choice procedure was used to examine the effect of trial duration and relative rate of reinforcement on discimination. There were two trial types 100-s trials and 30-s trials. At the start of a trial both left or right keys were illuminated red (on long trials) or green (on short trials), respectively. Pigeons were required to discriminate between the 1st and 2nd halves of a trial. Left- and right-key responses were reinforced according to independent Variable Interval schedules. Left-key responses were only during the 1st half of a trial, and right-key responses were only reinforced in the 2nd half of a trial. The VI schedule pairs were (VI30s/VI30s, VI18s/VI90s, VI90s/VI18s, VI180s/VI180s, VI270s/VI54s & VI270s/VI54s). Psychometric functions relating proportion of right-key responses to time since trial onset were sigmoidal, and when plotted as a function of relative time did not reliably superpose. When overall rate of reinforcement was greater, and when trials were longer rather than shorter, the distribution of responses shifted from the left to the right key earlier in a trial. The data will be discussed in relation to models of timing and choice.