BACKGROUND: Studies examining preference in concurrent schedules and concurrent- chains procedures have primarily involved steady-state designs with repeated exposure to unchanging reinforcer contingencies. More recently, the acquisition of preference has been studied using rapid-acquisition designs in which reinforcer contingencies change unpredictably across sessions. Grace and colleagues have proposed the cumulative decision model (CDM; Christensen & Grace, 2010; Grace & McLean, 2006; see Grace, 2016 for review) which accounts for acquisition of choice in concurrent chains as well as at steady state. However, the effects of conditioned reinforcement (i.e., terminal-link entries) are not considered in the CDM, and few studies have examined how relative terminal-link entry rate affects acquisition of choice.

AIMS: The first aim was to how preference is affected by rapidly changing reinforcer contingencies in concurrent schedules and concurrent chains. The second aim of this research was to develop a model of choice which could account for preference exhibited in concurrent schedules under rapid-acquisition conditions and use this model to extend the CDM to account for effects of conditioned reinforcement.

METHODS: Experiment 1 was conducted using concurrent schedules and examined preference in terms of sensitivity to reinforcer ratio when the distribution of log reinforcer ratios was uniform or bimodal. Experiment 2 examined sensitivity to terminal-link entry ratio when left and right terminal-link durations were both short (4 s) or long (16 s). Experiment 3 examined sensitivity to terminal-link entry ratio when initial-link durations were both short (4 s) or long (16 s). Experiment 4 examined sensitivity to terminal-link immediacy ratio when terminal links signalled which initial-link alternative resulted in terminal-link entry (cued) and when they did not (uncued). Left and right terminal links were of unequal durations (4 s and 16 s) or (16 s and 4 s). Experiment 5 examined sensitivity to terminal-link entry ratio when terminal links were cued and uncued but left and right terminal-link durations were equal.

RESULTS: In our concurrent schedules experiment, preference was stronger when the distribution of reinforcers was bimodal (approximate matching) and weaker when its distribution was uniform (severe under matching). In Experiment 2, short terminal-link durations resulted in approximate matching while long terminal links produced severe under matching. Bias was strong in pigeons in Experiment 3 but not systematic in terms of order or initial-link duration. Preference between short and long initial links was also approximately equal. However, there was a decrease in preference during the last 12 sessions for short initial links. In Experiment 4, preference was much stronger in cued terminal links (severe over matching) compared to uncued (matching). In Experiment 5, when left and right terminal-link durations were equal, preference was approximately equal between cued and uncued terminal links (severe under matching). However, preference was stronger when birds had prior exposure to uncued terminal links and preference was pooled across cued and uncued conditions (approximate matching). The decision model we developed was able to describe the difference in preference in Experiment 1 but was unable to explain it. This model was also able to account for observed preference in our concurrent-chains experiments, with the exception of Experiment 3. Only experiments using concurrent chains resulted in strong biases. Biases were strongest when terminal links were long and there was no prior exposure to short terminal links. Bias was weakest during short terminal links but only when there was prior exposure to long terminal links. When left and right terminal links were of equal durations, bias was stronger with prior exposure to cued terminal links and weaker with prior exposure to uncued terminal links. Bias was strongest when terminal links were uncued with no prior exposure to cued terminal links. Bias was weakest when terminal links were uncued but with prior exposure to cued terminal links.

CONCLUSION: Reinforcer contingencies such as terminal-link duration and signalled terminal links have been observed in previous research. Although the effect of initial-link duration on preference was not observed, there was a decrease in preference at the end of training for short initial links. This indicates that preference under rapid-acquisition conditions could be similar to that observed in steady-state research if pigeons had repeated exposure to unchanging terminal-link entry ratios. Results indicate that preference acquisition under rapidly changing terminal-link entry and immediacy ratios is similar to that observed in steady state studies. The strong biases observed are inversely related to the strength of preference. Although we offer one plausible explanation for this, further scrutiny of this relationship is required to understand it. In terms of predictive accuracy, the decision model we developed was able to predict observed preference and accounted for an adequate amount of variance in the data, with the exception of Experiment 5. The model's accounts of preference at the individual trial level were largely successful. However, when left and right terminal-link durations were equal and did not change, there was a more noticeable effect of the preceding session's terminal-link entry rates on current responding which indicates a molar component of responding prevalent under these conditions in Experiment 5, More research is required to improve our model's ability to account for this molar component. Moreover, additional empirical work is required to allow the model to explain preference acquisition in concurrent schedules under rapid acquisition conditions while satisfying parameter invariance.