My prefered MO usually involves a fairly hefty amount of layered multitasking, often drawing comments from more focused types. The heart of this NYT article on multitasking isn't so much that it's all bad, as that there's an optimum point after which the rate of returns on layering further goes way down. The "main" task(s) will usually suffer, so one needs to be able to adjust one's concurrent project load depending on the heaviest-focus item. I'm not sure knowledge-work interrupts represent quite the same thing as intentional layered multitasking though - I think interrupts that can be controlled differ greatly from those that can't, like checking new email versus having someone pop by your office with a hot crisis. I wonder if heavy multitaskers have any difference in their context-switching abilities, since that seems to be the key element?