The usual explanation for lost attention at work is that other people keep taking it. Messages arrive, meetings begin, colleagues ask questions and someone taps you on the shoulder.

That explanation is partly correct. The other part is more surprising and may be more useful.

Research shows that people are almost as likely to interrupt themselves as they are to be interrupted by something around them. Of the interruptions measured, 56% came from external events and 44% were self-interruptions.

That means nobody sent a message and no notification appeared. You were working on something, then found yourself checking something else without an external event prompting the change.

This finding can initially sound as though the problem is personal. It is not.

Self-interruption is often a habit developed in an environment that repeatedly rewards checking. Habits can be changed more easily than colleagues, meetings or other external demands.

This is why Unit 2 gives so much attention to the 44%. External interruptions also have a cost, but you may have limited control over them. Self-interruptions are the part you can influence more directly.

What does an interruption actually cost? The interruption itself is only part of it. Much of the cost appears when you return to the original task.

You need to reconstruct where you were, what you had already decided, which options you had ruled out and what you intended to do next. This rebuilding is difficult to see, but it happens after every interruption.

That explains a familiar type of day: one filled with small interruptions that seems to produce very little. None of the interruptions lasted long, but repeatedly rebuilding your train of thought consumed much of the available time.

The cost also depends on the type of work you were doing.

An interruption while formatting a document may cost very little because there was not much information to hold in mind. An interruption while solving a difficult problem may cause you to lose the entire thread, especially if you had not written down what you were thinking.

The same two-minute interruption can therefore be inexpensive during a routine task and costly during focused work. This is why both of these statements can be true: I only had a few interruptions and I could not get anything done.

Another cost can appear before the interruption happens. Once you know that a period of time may be interrupted, part of your attention begins waiting for it.

If you have a meeting in 20 minutes, you may avoid starting anything substantial because you are already aware of the approaching stop. The problem is not only the meeting itself. It is also the boundary you are monitoring before it begins.

This is why a fragmented calendar can cost more attention than the total duration of its individual events suggests.