Probability-Matching in the Fish

The experiments reported in the present paper were designed to provide further information on the course of probabilitylearning in the mouthbreeder and the conditions under which matching occurs.

Abstract

The only animal other than man which has yielded unequivocal evidence of probability-matching is the African mouthbreeder, Tilapia macrocephala.1 In the first of two exploratory investigations reported several years ago, a small number of these fish were trained on a simultaneous horizontal-vertical discrimination with response to one of the stimuli reinforced on a random 70% of trials and response to the other reinforced on the remaining 30% of trials. This distribution of reinforcements was ensured by the use of a 'guidance' procedure: if on any trial the unreinforced stimulus was chosen initially, it was removed, and S was permitted to earn a reinforcement for response to the other. Under these conditions, each animal in the group developed a stable tendency to choose the more frequently reinforced stimulus approximately 70% of the time. In the second experiment, the same animals were trained with two gray stimuli on a 70:30 spatial problem. Again, matching developed, although the preference for the more frequently reinforced position shifted rapidly from 70% to 100% when guidance was abandoned in favor of a pure noncorrectional method. Analogous experiments with rats and with monkeys have yielded no indication of matching. These animals maximize; that is, they tend to choose the more frequently reinforced stimulus almost 100% of the time.2 The experiments reported in the present paper were designed to provide further information on the course of probabilitylearning in the mouthbreeder and the conditions under which matching occurs.