Transitional choice behavior in concurrent-chain schedules

Reprint of "Acquisition of choice in concurrent chains: Assessing the cumulative decision model"

Concurrent chains is widely used to study pigeons' choice between terminal links that can vary in delay, magnitude, or probability of reinforcement. We review research on the acquisition of choice in this procedure. Acquisition has been studied with a variety of research designs, and some studies have incorporated no-food trials to allow for timing and choice to be observed concurrently. Results show that: Choice can be acquired rapidly within sessions when terminal links change unpredictably; under steady-state conditions, acquisition depends on both initial- and terminal-link schedules; and initial-link responding is mediated by learning about the terminal-link stimulus-reinforcer relations. The cumulative decision model (CDM) proposed by Christensen and Grace (2010) and Grace and McLean (2006, 2015) provides a good description of within-session acquisition, and correctly predicts the effects of initial and terminal-link schedules in steady-state designs (Grace, 2002a). Questions for future research include how abrupt shifts in preference within individual sessions and temporal control of terminal-link responding can be modeled.

Sensitivity to changing contingencies in an impulsivity task

Using a video-game-based escalating interest task, participants repeatedly encountered a reward that gradually increased in value over a 10-second interval. Responding early in the interval netted less immediate reward than responding later in the interval. Each participant experienced four different reward contingencies for waiting. These contingencies were changed three times as the experiment proceeded. Behavior tracked these changing contingencies, but wait times reflected long-term carryover from the previously assigned contingencies. Both the tendency to respond slowly and the optimality of behavior were affected by the order of contingencies experienced. Demographic variables only weakly predicted behavior, and delay discounting rate in a hypothetical money choice task predicted choice only when the contingencies in the game were weaker.

Response allocation in concurrent chains when terminal-link delays follow an ascending and descending series

Eight pigeons were trained in a concurrent-chains procedure in which the terminal-link immediacy ratio followed an ascending or descending series. Across sessions, one terminal-link delay changed from 2 s to 32 s to 2 s or from 32 s to 2 s to 32 s, while the other was always 8 s. For all pigeons, response allocation tracked changes in delay and was biased towards the 8-s alternative on the descending series, indicating a hysteresis effect, and was more sensitive to changes in the terminal-link delay ratio for relatively long (> 8 s) than short (< 8 s) delays. Both the hysteresis and effect of delay duration were predicted by an extended version of Grace and McLean's (2006) decision model. The extended decision model provided an overall better account of the results than a simple linear-operator model (Grace, 2002), and holds promise for an integrated account of choice in concurrent chains for both acquisition and steady-state conditions.

Initial-link duration and acquisition of preference in concurrent chains

Initial-link response allocation in concurrent chains becomes less extreme as the absolute duration of the initial links increases (Fantino, 1969). The present study asked whether initial-link duration affected how quickly response allocation reached asymptote (i.e., acquisition of preference). Six pigeons were trained on a concurrent-chains procedure in which the terminal links were fixed-interval (FI) 8 sec FI 16 sec or FI 16 sec FI 8 sec and were reversed every 20 sessions. Across conditions, all possible combinations of transitions between variable-interval (VI) 8-sec (short) and VI 24-sec (long) initial-link schedules were studied. Overall, the rate of acquisition was faster when the durations of the initial links preceding the reversal were short rather than long, and when the durations of the initial links following the reversal were long rather than short. By contrast, initial-link duration had no effect on acquisition or asymptotic measures of temporal control of terminal-link responding. These results support the core principle of delay-reduction theory (Fantino, 1969) that the impact of a conditioned reinforcer varies directly with initial-link duration, but also suggest that temporal learning during the terminal links proceeds independently ofinitial-link duration.

Effects of experience on preference between forced and free choice

Preference between forced choice and free choice in concurrent-chain schedules of reinforcement was investigated in pigeons after exposure to particular combinations of terminal links. In Experiment 1, in which terminal links always ended with reinforcers, one of three pairs of terminal links was arranged as preexposure: (a) both terminal links had only one key (forced choice), (b) both terminal links had two keys (free choice), or (c) a combination of forced and free choice was arranged across sessions. In test sessions following the preexposure, pigeons' preferences rapidly shifted to the terminal links with which they had no recent experience. In Experiment 2, the same procedure was repeated except that each terminal link ended intermittently with reinforcers with a probability of .5 and there was no terminal-link arrangement with a combination of free and forced choice. Pigeons showed the same preference changes as in Experiment 1, but the preference changes did not appear immediately at the beginning of test sessions. These data suggest that recent previous experience was a more important determinant of preference than the difference between forced-choice and free-choice terminal links.

Rapid acquisition of preference in concurrent chains

We report two experiments using a concurrent-chains procedure in which one terminal-link schedule was fixed-interval 8 s and the alternative schedule changed randomly from day to day. In Experiment 1, the alternative schedule varied between 4 s and 16 s according to a pseudorandom binary sequence similar to the one used by Hunter and Davison (1985). Similar to results with concurrent schedules, pigeons' response allocation in the initial link was most sensitive to the schedules arranged in the current session, although some effect of prior history was evident. Overall sensitivity was lower than for comparable data from steady-state research. In Experiment 2, a unique value between 2 s and 32 s was used for the alternative-schedule delay in each session. Sensitivity levels were similar to Experiment 1 and remained unchanged across 61 sessions of training. For all subjects, sensitivity was greater when the alternative-schedule delay was greater than 8 s compared with when it was less than 8 s. Generalized-matching plots revealed evidence of clustering of data points into two groups for some pigeons, suggesting that a process similar to a categorical discrimination may have at least partly determined response allocation. Overall, this research shows that pigeons' initial-link response allocation can adjust rapidly to frequent changes in the terminal links.

Concurrent-chain performance in transition: effects of terminal-link duration and individual reinforcers

Pigeons responded on concurrent-chain schedules with variable-interval initial links and equal delays as terminal links. The terminal-link delays were 1 sec in some conditions and 20 sec in other conditions. The percentages of reinforcers delivered for responses on the left key were 10%, 30%, 70%, or 90%, and this percentage was switched every five to nine sessions. The rate of change in the pigeons' response percentages after a switch was the same whether the terminal-link delays were 1 sec or 20 sec. Analysis of the effects of individual reinforcers showed that after a response on one key had been reinforced, response percentages on that key were higher for at least the next 100 responses. Small effects of individual reinforcers were evident after eight or nine additional reinforcers had been delivered. The effects of individual reinforcers were about equally large during times of transition and during periods in which overall response percentages were relatively stable.