Comparison of two oddity tasks with pigeons

Two oddity tasks were compared: one in which the odd stimulus could appear on any of the three stimulus positions (true oddity); the other in which the odd stimulus could only appear on the left or right but not the center (oddity-from-sample). A stimulus-configuration theory predicts faster oddity-from-sample learning, whereas a theory, based on shifts in observing over trials, predicts faster learning of true oddity. Pigeons learned the oddity-from-sample task faster, thus supporting configuration theory. Performance on the oddity-from-sample task but not the true oddity task was facilitated by increasing the number of responses required to terminate a trial. When the oddity-from-sample pigeons were shifted to the true oddity task, no evidence of positive transfer was found, suggesting that a configuration theory alone is insufficient to explain differences in acquisition. Only when a correction procedure was introduced did the true oddity pigeons perform above 50%.