A conventional conditioning analysis of "transitive inference" in pigeons.

As an alternative to a symbolic interpretation of transitivity in the discriminative performance of pigeons, a modified reinforcement theory (value transfer theory) was proposed by Fersen, Wynne, Delius, and Staddon (1991). Its novel assumption was that the value of the negative member of a pair of stimuli with which an animal is trained is enhanced by the value of the positive member of the pair. In this article, that assumption is shown to be unnecessary. All of the transitivity data for pigeons can be simulated with a simple conditioning model developed for honeybees that retains the conventional independence assumption. In a recent experiment by Fersen, Wynne, Delius, and Staddon (1991), pigeons were trained with five stimuli (A, B, C, D, and E) presented in four pairs, with the first member of each pair reinforced (A+ B-, B+ C-, C+ D-, and D+ E-), after which evidence of transitivity was sought—and found— in tests with B and D. The possibility of a conventional reinforcement interpretation of the preference for B over D was discounted by Fersen et al. on the grounds that the two stimuli had been "equally often rewarded and nonrewarded during training" (p. 336), and they were as little inclined to the symbolic interpretations popular in the primate literature (e.g., Gillian, 1981), which seemed unsuitable for pigeons and which failed in any case to account for the ordering of the training data. What they proposed instead (both for pigeons and for primates) was a modified reinforcement interpretation called value transfer theory, the central assumption of which is that the value of the negative member of each pair of training stimuli is enhanced by the value of the positive.