The cost of context switching
A didactic simulation of two task streams: each switch is charged a switch cost and a resumption lag, both shown as ranges, and the day's total is compared with the same work done in batches.
What the simulation does
The panel above runs a deliberately simple model. You choose how often you switch between two streams of work, how complex the work is, and how long your day is. For each switch the model charges two costs, both as a range: a switch cost for reconfiguring to the other task, and a resumption lag for finding your place again. It then adds the ranges up over the day and compares the result with the same workload done in batches.
What this shows / what it does not show
Shows: the shape of the argument for batching: how small per-switch costs accumulate, how they grow with complexity, and how much of the day batching gives back under the same assumptions.
Does not show: measured productivity. The per-switch ranges are illustrative: 0.2–1.5 s for a simple switch, longer for complex re-orientation. They are scaled from laboratory tasks lasting seconds to office work lasting hours, which is an inference the source studies do not make. Treat the outputs as a didactic approximation, not a measurement.
Where the switch cost comes from
Switching between two simple tasks reliably costs time and accuracy. In Rogers and Monsell's experiments, participants who alternated predictably between tasks were a few hundred milliseconds slower on switch trials, and part of that cost survived even when they had time to prepare A[2]. Monsell's review summarises two decades of such work: the cost reflects a mixture of active reconfiguration and interference from the task set you just left A[1]. Rubinstein, Meyer and Evans showed that the cost grows when tasks are more complex or less familiar, and shrinks with practice and cues A[3]. That is why the complexity control scales both ranges in the model.
These are seconds-long laboratory tasks. Nothing in them tells you how many minutes an office worker loses per day; the model's daily total is our extrapolation, and the ranges are wide on purpose.
Resumption lag: the second bill
When you leave a task mid-way, you later have to recover the goal you suspended. Altmann and Trafton's memory-for-goals model treats that goal as a memory trace whose activation decays, so resuming takes longer the longer and the more demanding the interruption was B[5]. Monk, Trafton and Boehm-Davis tested the prediction and found resumption lags of a few seconds that increased with interruption duration and demand B[4]. In a later procedural task, interruptions of under three seconds roughly doubled sequence errors and interruptions of about four and a half seconds roughly tripled them B[6]. Errors are a cost the simulation does not model at all, which is one reason its totals are conservative.
Field work adds a twist: Mark and colleagues found that interrupted participants compensated by working faster, finishing in less time but reporting higher stress and effort B[7]. "Lost minutes" is therefore only one of several currencies in which switching is paid.
Batched versus interleaved
The comparison bars keep the workload identical and change only the number of switches. Batching down to one change per hour removes most of the per-switch charges. Note what the model does not credit to batching: longer stretches of continuity may also reduce errors and the sense of time pressure, but we have no defensible numbers for those, so the bars show only the time charge.
Media multitasking: suggestive, not settled
A separate literature asks whether people who habitually juggle many media streams differ in attention. Uncapher and Wagner's review finds that heavier media multitaskers tend to perform worse on some measures of sustained attention and working memory, while noting that nearly all studies are correlational and that results are mixed B[8]. A replication effort and meta-analysis found the pooled effect small, with considerable heterogeneity and no reliable heavy-versus-light difference in several tests B[9]. Two cautions follow: the direction of any effect is unknown (people with weaker attention control may simply multitask more), and the effects are far smaller than popular writing suggests. This page treats media multitasking as a reason to be curious about your own habits, not as evidence of harm.
Practical reading of the model
- The complexity setting matters more than the switch count. A dozen trivial switches cost little in the model; a handful of deep re-orientations cost a lot.
- Batching helps most where re-orientation is expensive: design, analysis, writing, debugging.
- If you cannot batch, shorten the resumption: leave a one-line note of where you were before switching, which attacks the lag term directly.
- See how focus works in the brain for the attention side of the same story, and the focus planner to lay out batched blocks around your meetings.
Frequently asked questions
Is it true that multitasking costs 40% of productive time?
That figure is a popularisation and does not appear in the research cited here. Laboratory switch costs are measured in fractions of a second per switch; converting them into a percentage of a workday requires assumptions the studies do not make. Our model shows ranges for exactly that reason.
Why does the model show a range and not one number?
Because the inputs are uncertain. Switch costs depend on preparation time, practice and task similarity, and resumption lags depend on how long and how demanding the interruption was. A single number would imply a precision the evidence does not have.
Does batching always help?
In the model it always reduces the time charge, because it only removes switches. In practice some work needs quick turn-taking (support, collaboration), and very long single-task blocks bring their own fatigue. Use batching where re-orientation is expensive.
Does media multitasking damage attention?
The evidence is correlational and the pooled effect is small. People who multitask heavily with media score somewhat lower on some attention measures, but the direction of cause is unknown and replications are mixed.
Sources and evidence
- Task switching — Monsell S (2003), review A doi.org/10.1016/S1364-6613(03)00028-7 (retrieved 2026-11-09)
- Costs of a predictable switch between simple cognitive tasks — Rogers RD, Monsell S (1995), lab experiments A doi.org/10.1037/0096-3445.124.2.207 (retrieved 2026-11-09)
- Executive control of cognitive processes in task switching — Rubinstein JS, Meyer DE, Evans JE (2001), lab experiments A doi.org/10.1037/0096-1523.27.4.763 (retrieved 2026-11-09)
- The effect of interruption duration and demand on resuming suspended goals — Monk CA, Trafton JG, Boehm-Davis DA (2008), lab experiment B doi.org/10.1037/a0014402 (retrieved 2026-11-09)
- Memory for goals: an activation-based model — Altmann EM, Trafton JG (2002), theoretical model B doi.org/10.1207/s15516709cog2601_2 (retrieved 2026-11-09)
- Momentary interruptions can derail the train of thought — Altmann EM, Trafton JG, Hambrick DZ (2014), lab experiment B doi.org/10.1037/a0030986 (retrieved 2026-11-09)
- The cost of interrupted work: more speed and stress — Mark G, Gudith D, Klocke U (2008), lab experiment (CHI conference paper) B doi.org/10.1145/1357054.1357072 (retrieved 2026-11-09)
- Minds and brains of media multitaskers: Current findings and future directions — Uncapher MR, Wagner AD (2018), review B doi.org/10.1073/pnas.1611612115 (retrieved 2026-11-09)
- Cognitive control in media multitaskers: Two replication studies and a meta-analysis — Wiradhany W, Nieuwenstein MR (2017), meta-analysis B doi.org/10.3758/s13414-017-1408-4 (retrieved 2026-11-09)