Why Context Switching Costs So Much Attention
Context switching feels cheap because the switch itself takes two seconds. The cost lands afterwards, in the minutes your mind spends finishing the previous task while your hands are already on the next one. Two separate research lines measured this: Sophie Leroy's work on attention residue, and Gloria Mark's observational studies of interruption in office work. Together they explain why a day full of small switches leaves you tired and unusually unproductive.
A switch has two costs, not one
The first cost is residue. Leroy's 2009 research found that when you leave task A for task B, part of your attention stays with A, particularly when A is incomplete or unresolved. You are physically in B while cognitively still finishing A, and the harder A is to put down, the worse B performs.
The second cost is reorientation. To restart an interrupted task you have to rebuild your place in it: which paragraph you were on, which hypothesis you were testing, which file you had open. That rebuilding is invisible in a calendar and easy to forget in your own accounting of the day.
What the interruption studies measured
Mark and colleagues observed information workers directly rather than asking them to recall their day, and found people moving between tasks every few minutes, with self-interruption accounting for a large share of the switches. In one 2008 study the average time to return to the original task after an interruption was about 23 minutes, and the interrupted work was completed faster than uninterrupted work, at the cost of higher stress, more effort and more frustration. You do not lose the time only. You pay for it in strain.
Two details stop that 23-minute figure from being misused. It is an average across tasks and people, so a short reply to a colleague is not the same as being pulled into a problem. And it measures the return to the original task, not the return to work, which is why the number is bigger than most people expect.
Not every switch costs the same
Switch cost asymmetry is real: moving from a demanding task to an easy one is cheaper than the reverse, and coming back to a task you had barely started is cheaper than returning to one you were deep inside. Switches you chose cost less than switches that arrived, and a switch during a natural break in the work costs less than one mid-sentence.
That gives you levers. Cluster similar work so the reorientation is shared, keep the hardest task in the same slot each day so you rebuild context only once, and end blocks at a finished sentence instead of in the middle of a thought.
There is also a sizing rule you can use without counting anything: if a switch cannot be described in one short sentence when you come back, it was too expensive to take on a whim.
- Chosen beats imposed - Choosing when to check messages is cheaper than being pulled. Even the same interruption costs less when you scheduled its slot.
- Hard to easy is cheap - Order the day so your most demanding block comes before the shallow stretch, not after it.
- Break points absorb switches - A switch at a natural stopping point costs a fraction of one mid-paragraph. Stop deliberately and often.
Four ways to cut the tax
First, close loops in writing. Before you leave a task, spend 30 seconds writing the next concrete action, for example 'add the third failure case to the test file'. That single line gives your mind somewhere to park the task and measurably reduces residue.
Second, capture stray thoughts instead of acting on them. One notes file, one line per thought, and back to the task. Third, batch communication into three checkpoints rather than leaving the inbox open. Fourth, make blocking cheap: a status message that says when you are unavailable removes the need for people to interrupt you in order to find out.
The switches you cannot remove
Some jobs are made of switching. On-call engineers, support agents, teachers, nurses and parents working from home are interrupted as a condition of the role, and advice to close the inbox is not available to them. For those roles the workable goal is not zero interruptions but scheduled recovery: knowing which hour afterwards is protected for the interrupted task, and never scheduling that hour full.
Everywhere else, the honest target is reducing the self-inflicted switches, which are usually the majority and are the ones you control. Count them for two days. If you are the one pulling yourself off task 20 or 30 times a morning, notifications are running your calendar.