An evaluation of the assumptions of the random walk model to account for rats' learning of reward patterns in a radial-arm maze

Suzan Alptekin, Purdue University

Abstract

The Random-Walk Model, which was developed to explain underlying mechanisms of multiple pattern learning behavior, was evaluated in seven radial-arm maze experiments. In all experiments but one, there was only one arm that contained large reward, all other arms containing small reward. The location of large reward was different from trial to trial. Findings of all seven experiments indicate that rats did not associate trial number (with the possible exception of Trial 1) with the location of large reward. Similar results were obtained regardless of (a) amount of Proactive Interference (PI) accumulation within and between trials, (b) sample size, (c) whether amount of training was doubled, (d) whether incentive value of large reward was increased, (e) steps taken to reduce precipitous arm-entry decisions and (f) the addition of salient intra-trial cues. Reliable patterns of responding were observed when only half of the arms in a radial-arm mate could contain large reward over the course of eight trials. Under those conditions rats learned to enter arms that had a higher probability of containing large reward earlier in the trial. These findings led to the conclusion that rats did not use a trial-number strategy in the present experiments, but they did not seem to use an item-item strategy either. Several relevant studies were discussed in this light.

Degree

Ph.D.

Advisors

Capaldi, Purdue University.

Subject Area

Psychology|Experiments

COinS