What the observation found
Mark, Gonzalez and Harris shadowed 24 information workers and recorded how their days were actually structured. The average time spent in what they called a working sphere, a coherent unit of work, was 11 minutes 4 seconds before the person switched to another one or was interrupted. Fifty-seven percent of working spheres were interrupted. The fragmentation was not occasional; it was the normal shape of the day. Interruptions came roughly evenly from outside and from the person themselves, and the internal ones were mostly personal work while the external ones mostly concerned their central projects. This is measurement of real workdays, not a laboratory estimate, which is why the numbers are worth quoting precisely.
The cost is in the return, not the interruption
The same study followed what happened after an interruption. Most interrupted work was resumed on the same day, and when it was, it took an average of 25 minutes 26 seconds to get back to it. The reason that number is large is visible in the detail: before returning, people worked in an average of 2.26 other working spheres. Attention had gone to two or three other subjects in between, and the environment that would have cued a fast return, the open document, the cursor where you left it, had usually been rebuilt around something else. Externally interrupted work came back slightly faster, at 22 minutes 37 seconds, than self-interrupted work at 29 minutes 1 second.
Faster is not free
A later experiment by Mark, Gudith and Klocke produced the result people find counterintuitive. Interrupted tasks were completed in less time, with no difference in quality. The interpretation the authors offer is that people compensate for interruption by working faster and writing less. The compensation has a price, and it showed up on the measures: participants in the interrupted conditions reported significantly higher stress, higher frustration, more time pressure and more effort. The finding to carry away is not that interruptions are harmless because the task still got done. It is that the cost moved out of the schedule and into the person, where a calendar cannot see it.
What you can do with this
Batch similar work so that the switch you do make is between related things rather than unrelated ones, because reconstructing context is the expensive part and related work reuses it. Protect a small number of long blocks rather than trying to protect the whole day, since more than half of work periods were interrupted even under observation and a plan that assumes zero interruption will fail on the first one. Leave a cue when you are pulled away, one line about where you were, so the return is cheaper. And measure your own switching before assuming these figures describe you; they describe 24 people in one study and are a starting point, not a verdict.
In short
- Observed information workers averaged 11 minutes 4 seconds in a working sphere before switching or being interrupted.
- Fifty-seven percent of working spheres were interrupted.
- Work resumed the same day took an average of 25 minutes 26 seconds to return to, after an average of 2.26 other activities in between.
- Self-interrupted work came back slower, 29 minutes 1 second, than externally interrupted work at 22 minutes 37 seconds.
- In a follow-up experiment, interrupted tasks were finished faster with equal quality but with significantly more stress, frustration, time pressure and effort.
Related guides
- Time Tracking for Freelancers: Freelancers track time for three separate reasons: to bill accurately, to support the records their tax authority expects, and to find out where the unpaid half of the week goes.
- Time Blocking: Time blocking means assigning each piece of work a specific time and place in advance instead of working from a list.
- How to Run a Time Audit: A time audit is one week of recording what you actually do, then reading the totals.