"focus completely and without distraction on a single research task. (He typically divides the writing of a scholarly paper into three discrete tasks: analyzing the data, writing a full draft, and editing the draft into something publishable.) During these periods, which can last up to three or four days, he'll often put an out-of-office auto-responder on his e-mail so correspondents will know not to expect a response. "It sometimes confuses my colleagues," he told me. "They say, 'You're not out of office, I see you in your office right now!'" But to Grant, it's important to enforce strict isolation until he completes the task at hand."