Task-switching in pigeons: Associative learning or executive control?

Abstract

Human performance in task-switching paradigms is seen as a hallmark of executive-control processes: switching between tasks induces switch costs (such that performance when changing from Task A to Task B is worse than on trials where the task repeats), which is generally attributed to executive control suppressing one task-set and activating the other. However, even in cases where task-sets are not employed, as well as in computational modeling of task switching, switch costs can still be found. This observation has led to the hypothesis that associative-learning processes might be responsible for all or part of the switch costs in task-switching paradigms. To test which cognitive processes contribute to the presence of task-switch costs, pigeons performed two different tasks on the same set of stimuli in rapid alternation. The pigeons showed no sign of switch costs, even though performance on Trial N was influenced by Trial N - 1, showing that they were sensitive to sequential effects. Using Pearce's (1987) model for stimulus generalization, we conclude that they learned the task associatively-in particular, a form of Pavlovian-conditioned approach was involved-and that this was responsible for the lack of any detectable switch costs. Pearce's model also allows us to make interferences about the common occurrence of switch costs in the absence of task-sets in human participants and in computational models, in that they are likely due to instrumental learning and the establishment of an equivalence between cues signaling the same task.