Reactively homogeneous compound trial-and-error learning with distributed trials and terminal reinforcement.

Abstract

Forty [male] albino rats were trained with terminal food reinforcement on a linear maze with 4 possible choices at each of 4 choice points. At any given choice point, one pathway would permit passage of the animal, the other 3 pathways being locked, though all looked exactly alike to the animal. In this expt. the correct responses at all 4 choice points were the same. This type of compound trial-and-error learning yielded late in the learning process a characteristic deviation from the goal gradient error distribution among the 4 choice points, which showed its min. number of errors at the 3d choice point rather than at the 4th. Other published studies have yielded parallel results. The explanation is that since the same act is learned at all choice points and the external stimuli are practically identical at all these points, the learning at each generalizes to all the others.