In a straight runway or T-maze, rats easily learn simple decreasing reward quantity sequences (e.g., 14-7-3-1-0 food pellets), running faster to initial larger rewards and slower to terminal smaller rewards. Findings suggested that rats could encode "rule", abstract relationships between items to learn the formal structure of a serial pattern. However, strong evidence has shown that formation of item-association and stimulus generalization among signals was sufficient to explain the wide range of findings, such as effects of structural complexity or phrasing, extrapolation of novel item, and so on. The dispute between rule learning and item-association is a typical one of absolute vs. relative, or concrete (sensory) vs. abstract representation of stimuli. Therefore, we cannot be sure of rats' abilities to encode abstract relationships among food stimuli. I will show some major findings of rats' reward serial learning, and then present two kind of experiments which examined discrimination learning about abstract relationships among reward quantities (less than or greater than) or reward qualities (same or different). Through these examinations, it was shown that rats had a strong tendency to respond to absolute value of food stimuli, but when great memory load was required for the rote association learning, they started to respond to abstract relationships of food stimuli.

TANIUCHI, Tohru
Department of Psychology, Faculty of Letters, Kanazawa University
Kakuma, Kanazawa 920-1192, Japan
tohruta@kenroku.kanazawa-u.ac.jp