Abstract
Learning by 60 hooded rats of 12 patterns in a six-unit multiple T-maze was measured. Each pattern contained one response in the direction opposite to the other five responses, differing only in the location of the odd response. The most and least difficult patterns (errors to criterion) had the odd response in the middle and the first two units of the maze, respectively. The distribution of errors over choice-points varied with maze patterns, which was reflected in a significant interaction of patterns and choice-points. Within a maze pattern the most difficult choice-point was the odd response; perseverative errors were fewer than anticipatory errors, and the latter occurred mostly in the unit preceding the odd response. The applicability of bi-directional or goal gradients to account for the results is discussed.