Cognitive theories of timing and conditioned reinforcement provide two different theoretical perspectives on choice between delayed rewards. The primary objective of this research was to identify the process that generates choice in the concurrent-chains procedure and to characterize its relationship with temporal control. Experiments 1-3 investigated the relationship between the dynamics of pigeons’ preference and temporal control in concurrent chains using an arrangement in which the delays to reinforcement changed unpredictably across sessions. To obtain convergent measures of choice and timing behavior, occasional ‘no-food’ terminal links lasted longer than the schedule values and ended without reinforcement. Measures of choice (log initial-link response ratios) and timing (start and stop times from no-food terminal links) stabilized within individual sessions. Sensitivity of log response ratios to relative immediacy increased as initial-link duration decreased or absolute terminal-link delays increased, but absolute initial- and terminal-link duration did not affect temporal control. Residual covariation analyses of log response ratios with log start and stop time ratios confirmed that measures of choice and timing were interdependent. Experiments 4 and 5 used concurrent-chains procedures in which immediacy, magnitude (and probability, in Experiment 5) ratios for left and right keys were 2:1 or 1:2, determined across sessions by independent, random series. Experiment 6 was a concurrent schedule in which relative reinforcement rate and magnitude were 2:1 or 1:2, determined the same way. Multiple regression analyses showed that pigeons’ response allocation in Experiments 4-6 was sensitive to multiple dimensions of reinforcement. Levels of preference within individual sessions and initial links or interfood intervals was more extreme when all dimensions favored the same key than when at least one dimension favored each key, consistent with assumptions of the generalized matching law. Within individual sessions, changes in response allocation in all experiments tended to be abrupt, consistent with the assumptions of Rate Estimation Theory (Gallistel & Gibbon, 2000). A decision model that posits a comparison between delayed outcomes with a criterion delay (Grace & McLean, 2006) described initial-link responding in Experiments 1-3. A modified decision model in which outcome expectancy is compared to an expectancy criterion described responding in Experiments 4-6.

The purpose of this research is to determine if loss aversion is replicable as an overt behavioral response to potential gains and losses in complex, recurring, uncertain, and risky choice with real gains and losses of money and food. Cognitive methods used to determine the effect of loss have primarily measured verbal response to hypothetical choice scenarios in which participants cognitively predict their behavior in a series of bets or situations involving imagined monetary gains and losses. Less has been done using behavioral methods that measure overt behavioral response to gains and losses of actual commodities. The present study uses the experimental analysis of behavior to measure the asymmetrical effect of loss in multiple choice domains. A series of four experiments investigated four factors likely to affect the expression and degree of loss aversion: (a) learning and experience with consequences of choice; (b) real gains and losses instead of hypothetical quantities or imagined commodities; (c) gains and losses of a non-quantitative, primary reinforcer (food); and (d) the menstrual cycle. Participants played one of two computer games in which they earned or lost coins or food tokens exchanged for real food. Participants (N = 27, 15 women) played several 18-minute sessions in gains-only conditions and 16 sessions in 36-minute gains+punishment conditions. Recurring, complex, uncertain, and risky choice was simulated in the games by using 6-ply interdependent concurrent variable interval schedules of reinforcement (gains) and punishment (losses). Choice behavior with real gains and losses of money and food was modeled using the generalized matching law, allowing for the quantification of the effects of potential loss, relative to gains, as a change in bias and sensitivity. Loss aversion was operationalized as gain-loss asymmetry ratios derived from bias estimates produced in unpunished and punished choice conditions. Gain-loss asymmetry was replicated in both women and men in complex, recurring, uncertain, and risky choice with potential gains and losses of real money and food. Average gain-loss asymmetry ratios were 3 to 6 times greater in choice with money and 4 to 16 times greater in choice with food than those reported in the cognitive and behavioral literature. Although individual differences in response to loss were striking, the asymmetrically larger behavioral effects of loss, relative to gains, were nearly ubiquitous. Marked disruption in sensitivity to reinforcement was observed in punished choice for most participants, but for 33% of participants in choice with money and 42% in choice with food, sensitivity to reinforcers increased. No evidence was found for behavioral choice varying with the menstrual cycle.