Serial reversal learning in Gray Squirrels: learning efficiency as a function of learning and change of tactics

Abstract

Learning allows individuals to adapt their behaviors flexibly to a changing environment. When the same change recurs repeatedly, acquiring relevant tactics may increase learning efficiency. We examined this relationship, along with the effects of proactive interference and other interference information, in a serial spatial reversal task with 5 gray squirrels (Sciurus carolinensis). Squirrels completed an acquisition and 11 reversal phases with a poke box in which 2 of 4 possible reward locations were baited diagonally in a square array. In this situation, an efficient tactic is to locate the diagonally related locations consecutively (integrative search tactic) instead of searching rewards in a clockwise or counterclockwise direction (sequential search tactic). All squirrels formed a learning set, acquiring successive reversals in fewer trials. Although 4 squirrels gradually employed more integrative tactics in locating the rewards both within and between phases, sequential tactics were used in the first trial of each phase. This suggests the integrative tactic did not depend on an association between the rewarded locations but was learned as a spatial pattern and/or by use of extra-apparatus cues to locate individual rewards. Generalized estimating equation models showed that learning efficiency increased with experience and tactic change. Although tactic change partially mediated the effect of learning on learning efficiency, learning was an independent contribution to improved efficiency. Squirrels that used more integrative tactics made fewer total errors than squirrels that used less integrative tactics, suggesting that learning a task-relevant tactic using spatial cues can provide direct benefits in maximizing rewards and minimizing time costs.