Switching between tasks costs more than the interruption itself

A cost that shows up in the laboratory

Task switching has been studied in tightly controlled conditions for decades, using designs where people alternate between two simple rules and their response times are compared with blocks where the rule stays fixed. Switching trials are reliably slower and produce more errors. This is one of the sturdier findings in cognitive psychology, precisely because the tasks are trivial and the measurement is direct.

The size of the cost in those experiments is measured in fractions of a second, which sounds negligible and is not the point. What the design establishes is that a switch is an operation with a cost, not a free redirection of attention, and that the cost persists even when the person knows the switch is coming and has time to prepare.

Why the real-world version is larger

The laboratory case understates it because the tasks are simple and hold no state. Real work involves a partly built structure in your head — where you were, what you had ruled out, what you were about to check — and that structure is not preserved across a switch. Rebuilding it’s most of the cost, and it scales with how complicated the work was.

There is also a documented tendency for attention to remain partly on the previous task after moving to a new one, particularly when the first was left unfinished. The practical consequence is that the second task is performed with less than full capacity for some period, so the loss isn’t confined to the moment of transition.

Extrapolating precise figures from the laboratory to a working day is not defensible, and published claims about how many minutes it takes to recover focus should be treated with suspicion unless the source is clear. The direction of the effect is well established; the magnitude in real conditions is not.

Interruption and self-interruption behave differently

Observational studies of office work have found that a large share of interruptions are self-generated: people leave a task without any external prompt, often when it becomes difficult or when a natural pause occurs. This matters because the two kinds call for different responses. External interruptions are a scheduling and environment problem. Self-interruption is a habit, triggered by a cue that is usually identifiable.

The cue is frequently the moment of friction — the point where the work stops flowing and something else becomes appealing. Noticing that specific moment is more useful than any general resolution to concentrate, because it is the one point where a different response is actually available.

Batching is the structural fix

If a switch costs something, then the number of switches is a variable you can manage. Grouping similar work together means the mental structure built for one item is partly reusable for the next, and the transitions happen between blocks rather than within them. This is why a day with three long stretches produces different results from one with twelve fragments, even if the total minutes are identical.

The same logic applies to the things that generate interruptions. Checking messages at set times rather than continuously does not reduce the number of messages, and it substantially reduces the number of transitions, which is the quantity that was costing you.

What multitasking actually is

Genuine simultaneous performance of two demanding tasks is rare and, where it occurs, usually involves one task being sufficiently automatic to run without attention. What is normally called multitasking is rapid alternation, paying the switch cost repeatedly, and it feels productive because activity is high and each individual switch is imperceptible.

The feeling of competence at it appears to be poorly related to actual performance, and some studies have found that people who report doing it most aren’t the ones who do it best. That is a correlational finding from specific samples rather than a general law, but it is at least consistent with the general pattern that fluency and accuracy come apart.

A realistic accommodation

Nobody works in uninterrupted blocks all day, and a policy that requires it will break on contact with any job involving other people. The achievable version is to protect a smaller number of stretches for the work that has the most state to rebuild, and to accept fragmentation for the work that does not — correspondence, small decisions, anything that carries no structure between sessions.

The other accommodation is to leave a marker when a switch is forced on you. A single line about where you were and what you intended next reduces the rebuilding cost when you return, which is a small habit with a return out of all proportion to the effort of writing it.