A cumulative decision model for three-alternative choice in concurrent chains

Traditional models for choice in the concurrent-chains procedure have assumed that terminal-link stimuli acquire value as conditioned reinforcers, and that 2-alternative choice provides a measure of relative value according to the matching law. By contrast, the cumulative decision model (CDM; Christensen & Grace, 2010) explains choice as the aggregate effect of comparing delays to a criterion on initial-link responding, not conditioned reinforcement. Here we test whether the CDM can account for choice in 3-alternative concurrent-chains (3ACC) and compare it with the hyperbolic value-added model (HVA; Mazur, 2001), which assumes that choice depends on the increase in conditioned reinforcement value signaled by terminal-link stimuli and has been successful in previous 3ACC research (Mazur, 2000). In Experiment 1, 4 pigeons responded in 3ACC in which the terminal links were fixed-interval schedules, and parameter estimates from fits of CDM and HVA were used to calculate predictions for conditions with variable-interval terminal links. The predictions of CDM were more accurate than those of HVA. In Experiment 2, 7 pigeons responded in 3ACC in which the terminal links were fixed-interval schedules. Although both models described the data well, residuals from HVA fits showed a systematic pattern predicted by CDM, characterized by a third-order polynomial with a negative cubic coefficient. Finally, we conducted a residual meta-analysis (Sutton, Grace, McLean, & Baum, 2008) of data from prior 3ACC studies. HVA residuals showed the same negative cubic pattern as in Experiment 2, whereas no systematic pattern was found in the CDM residuals. Overall, results support the CDM and suggest that the same principles which describe binary choice in concurrent chains generalize without modification to 3-alternative choice. (PsycINFO Database Record