It is a short and accessible review article in an area of research—executive function in human cognition (or how we control our own minds)—which has become very active in cognitive psychology and neuroscience over the past couple of decades. The field is hungry for measurable phenomena which may index control processes, their consequences, or limitations.

The observation of substantial costs to performance (longer response times and higher error rates) at the point where someone is required to change tasks is suggestive of executive control "work" and the possibility of measuring it. The paper reviews some provocative findings on the costs of task-switching and ideas about their interpretation.

“The paper reviews some provocative findings on the costs of task-switching and ideas about their interpretation.”

It reviews new methods and discoveries from the previous decade. Although task-switch costs were first measured at least 80 years ago, research on the phenomenon was sporadic, at best, until the mid 1990s, when papers by Robert D. Rogers and myself—J. Exp. Psych: Gen, 1995; Alan Allport, Elizabeth A. Styles, and Shulan Hsieh, Attention and Performance XV, 1994; Nachshon Meiran, Learning Memory and Cognition, J. Exp. Psych: 1996; and Daniel Gopher, Cognitive Brain Research, 1996—introduced several new behavioral methods for studying switch costs, and suggested theoretical accounts for what they found.

Some of the phenomena are quite counterintuitive. For example, although having the opportunity to prepare for a task switch usually reduces switch costs, it seems to be impossible to eliminate switch costs completely, no matter how much time you have to prepare. As another example, it is often found that switching to the more familiar of two tasks incurs a greater cost than switching to the less familiar task.

Over the last few decades we have begun to discover methods for investigating the fundamental question of how our minds and brains control themselves—what 19th century philosophy and psychology called "The Will." The paper reviews one recent strand of this research. To perform any cognitive task, we have to get our brain organized in a particular way (we call this a "task-set").

The requirement to change tasks imposes two kinds of overhead. First, a process of task-set reconfiguration is needed: this takes time and, if not fully completed, can lead to a poorer quality of performance and even outright errors. Second, task-sets for recently performed tasks remain active in the mind even after the task has changed. If the present stimuli afford those now-irrelevant tasks, this can cause substantial interference with selection of the appropriate action. These overheads are but part of the more complex overhead incurred when we "multitask"—attempt to accomplish multiple tasks by interleaving them over a period of time.

I first became interested in task-switching when I was running psycholinguistic experiments at the University of Chicago. I needed participants to alternate predictably between two simple tasks, both of which involved reading a printed word. I noticed that the subjects had some difficulty performing the appropriate task even though there were plenty of cues to remind them which task they were supposed to be doing.

After moving to Cambridge, I wrote a grant proposal to investigate these switch costs as a possible source of information about control mechanisms, but it was turned down. I then suggested investigating switch costs as a project to a new Ph.D. student in my lab, Robert Rogers, and our research took off from there, especially when we discovered that colleagues elsewhere had also begun to work on switch costs, but had quite different ideas about their interpretation.

In various directions, including exploration of the neural substrate of task-set control.

An increase in the requirement to multi-task seems to be an inevitable consequence of technological development: communications technology in particular (Email, cell phones, etc.) means that we are constantly degrading our performance of one task through allowing it to be interrupted by others. It is important to understand these costs, and allow for them, especially in relation to safety-critical tasks like driving, flying, and plant control. And it is important to oppose the notion that it is somehow virtuous to multi-task: it may be necessary, but the result is almost always a substantial impairment of performance of the tasks in question.