Human preference in a concurrent-chains procedure with variable delays to reinforcement

Human subjects' preferences were observed in a concurrent-chains procedure in which two variable delays were arranged in each terminal link. Experiments 1 and 2 arranged the terminal link delays in the first condition such that the arithmetic means of the delays were equal across terminal links and the harmonic mean of the delays of one terminal link was twice the value of the second terminal link. In the second condition, the arithmetic means of the terminal links were identical to the first condition, and the two harmonic means of the terminal links were extended by a factor of two. Terminal link access was assigned dependently. In Experiment 3, the terminal link delays were arranged such that the harmonic means of the delays were equal across conditions, and the arithmetic mean of the delays of one terminal link was twice the value of the other. Terminal link entry in the first condition was assigned dependently, and in the second condition independently. Experiment 4 arranged the terminal link delays in an identical manner to the first conditions of Experiments 1 and 2. The shorter harmonic mean was arranged on the left side of the concurrent-chains procedure in the first condition, and on the right side in the second condition. Both conditions of Experiment 4 were scheduled independently. Overall, subjects were not sensitive to the relative harmonic mean of terminal link delays in Experiments 1, 2, and 4. Subjects exhibited a strong preference in the independent condition of Experiment 3 for the terminal link with the shorter arithmetic mean. A very small preference was found in the dependent condition - also for the shorter arithmetic mean terminal link. Sensitivity to the relative arithmetic mean of terminal link delays exhibited in the current study is not consistent with animal studies that have shown subjects to be sensitive to the relative harmonic mean. The moderation of preference from the independent to the dependent condition is consistent with earlier operant research employing humans as subjects