University of Canterbury.
J Exp Anal Behav. 2019 Sep;112(2):192-209. doi: 10.1002/jeab.544. Epub 2019 Aug 12.
Four pigeons responded in a three-component concurrent chains procedure in which the terminal-link schedules were either both fixed-interval (FI FI), both variable-interval (VI VI), both mixed-interval (MI MI) or variable-interval fixed-interval (VI FI). Across components within sessions, overall terminal-link duration was varied while schedule types were varied across conditions. For the conditions with homogeneous schedules, the strongest preference was obtained with FI FI, intermediate with MI MI, and weakest with VI VI. Preference increased with overall terminal-link duration for all schedules, but the increase was more rapid for VI FI. The hyperbolic value-added model (HVA) and cumulative decision model (CDM) were fitted to the choice data and accounted for 83.9% and 76.1% of the variance, respectively. However, deviations from both models' predictions were systematic. A modification of the CDM, which assumed that reinforcers delivered after variable delays were more effective, improved the fit of the model and substantially reduced the systematic deviations in residuals. Comparable modifications of HVA produced only limited improvement. Results show that preference in concurrent chains with homogeneous terminal links depends on the degree of variability in reinforcer delays, and that the CDM can provide an excellent account of results across different terminal-link schedules.
四只鸽子在三成分同时链程序中做出反应，其中终端链接的时间表要么都是固定间隔（FI FI），要么都是变化间隔（VI VI），要么都是混合间隔（MI MI），要么是变化间隔和固定间隔（VI FI）。在各阶段内的各成分中，在改变条件的同时，总体终端链接持续时间发生变化。对于同类型时间表的条件，FI FI 得到了最强的偏好，MI MI 是中等偏好，而 VI VI 是最不受欢迎的。对于所有时间表，偏好都随总体终端链接持续时间的增加而增加，但 VI FI 的增加更为迅速。双曲线增值模型（HVA）和累积决策模型（CDM）拟合了选择数据，分别解释了 83.9%和 76.1%的方差。然而，两个模型的预测都存在系统偏差。CDM 的一个修正假设，即在变化延迟后提供的强化物更有效，这改善了模型的拟合度，并大大减少了残差的系统偏差。HVA 的类似修正仅产生了有限的改进。结果表明，具有同类型终端链接的同时链中的偏好取决于强化物延迟的可变性程度，而 CDM 可以很好地解释不同终端链接时间表下的结果。
