Statistically Probable Cues Bootstrap Cognitive Flexibility: Evidence from a Task-Switching Paradigm Natasha Zoe Kirkham Cornell University nzk2@cornell.edu   Task-switching paradigms investigate the ability to switch flexibly between two or more tasks. During a switch, the participant has to a) prepare to perform the alternate (appropriate) task, and b) inhibit the desire to perform the previous (now inappropriate) task. The time taken to perform these two operations is referred to as the "switch cost" (the reaction time above and beyond that required to perform the first task). This cost is deemed substantial, making task switching a difficult cognitive operation. However, in the real world, we shift constantly from one set of cognitive operations to another. These shifts are usually fast and most of the time there are no obvious external cues. How do we do this? How is the cognitive system reconfigured, from moment to moment, in a way that allows us to switch flexibly between tasks? What factors affect this flexibility? The current study investigates the role of statistical probability in task switching. Participants! were required to identify one of two objects (a red flower and a blue car) according to their color or their shape. The objects were presented sequentially at different locations on a 4 x 3 grid. The grid was divided into two halves by a bright yellow vertical line: On the left half, objects were to be identified by their color, and on the right half, objects were to be identified by their shape. The words "color" and "shape" appeared above the appropriate sides of the grids. The objects appeared on the grid in a statistically probable spatial pattern (4 triplets randomly presented). Results showed that not only did the statistically probable pattern reduce reaction times when no switch was required (an expected result based on implicit learning literature), but the pattern also reduced reaction times when a switch was required. This suggests that cognitive flexibility can be positively affected by the underlying (and, in this case, irrelevant) context statistics.   Back to main page