THEORY OF ANIMAL TIMING: MEMORY AND ASSOCIATION

前几年在这个项目上已经看到了正式的发展

Previous years on this project have seen the development of a formal theory of animal timing, Scalar Expectancy Theory (SET), applied to a wide range of timing effects in animal learning. Past work has analyzed the way in which timed delays are remembered and discriminated in a variety of animal psychophysics preparations. The recent grant period has studied memory for aggregates of variable delays to reinforcement in contrast to memory for fixed delays. A recent publication (Appendix A) summarizes much of the theoretical and empirical work on this problem. It is shown there that a variety of common findings in the literature on choice (overmatching, undermatching, matching and preference for variable delay) may be subsumed within the general rubric of SET., A major finding has been the "one-sample" rule for encoding and decoding memory for variable delays. Tests of the account contrast preference for Poisson variability, common n the laboratory (and in nature), with a unique new distribution of variable delays (Backward exponential), for which preference does not exhibit the asymmetry heretofore common. A continuation of this line of work is planned examining these sets of delays in memory utilizing a temporal generalization procedure. Techniques are developed which permit a trial-by-trial analysis of the relative contribution of memory and response variance. A second focus of the competing continuation would study the role of temporal factors in the acquisition of learned associations. Early work on this project conceived of the acquisition of an association as the perception of a better, i.e. shorter, delay to reinforcement cued by the conditioned stimulus, relative to other cues in the situation. The present proposal will address a tension developed earlier between SET and alternative, stimulus competition theories of associative learning -- prototypically, the Rescoria-Wagner theory. Competition theories of associative learning were not originally designed to address temporal predictiveness problems, but rather to understand some hallmark properties of the learning process, blocking and overshadowing. SET, in contrast, was designed to understand temporal processing, and has not seriously addressed the blocking and overshadowing phenomena. Some strong designs in the area of non-contingent versus contingent protocols (e.g. Durlach, 1983) have favored a learning account of context blocking. Preliminary work for the present proposal challenges this interpretation. The proposed studies will attempt to push the contrast between a time- based comparator theory and a competition based incremental theory of the emergence of multiple associations.