Expectancy mediation of concurrent conditional discriminations.

Pigeons learned a pair of two-choice conditional discriminations concurrently. One was red/green identity matching. The other also employed red and green sample stimuli, but the choice alternatives were vertical vs. horizontal lines, with vertical being correct following red samples and horizontal being correct following green. Correct choices produced either food reward or a brief feedback tone. For the "congruent" group, a given sample reliably predicted a particular outcome (e.g., red-food, green-tone), regardless of whether the animal chose between colors or lines. For the "incongruent" group, the outcome that followed a particular sample was one thing on color-choice trials and the other thing on line-choice trials. For example, if a subject in the incongruent group received food for red matches and tone for green matches, then it received tone for correct vertical choices and food for correct horizontal choices. The congruent group reached significantly higher levels of correct performance on both problems than did the incongruent group. Moreover, these relative levels of performance were reversed when birds trained on the congruent procedure were switched to the incongruent procedure and vice versa, even though the stimulus-response requirements of both problems remained unchanged. The experiment provides further evidence that the outcome expectancies acquired in learning situations can be significant sources of stimulus control of instrumental behavior.