The Multitasking Myth: What Dual-Task Games Reveal About Your Brain
Almost everyone believes they are good at multitasking. Almost nobody is. Decades of dual-task research point to one uncomfortable conclusion: for any two activities that each require attention, the human brain does not run them in parallel. It alternates — rapidly, imperfectly, and at a measurable cost that most people never notice because they have nothing to compare it against.
The bottleneck in your head
The clearest demonstration is a lab paradigm called the psychological refractory period. Give someone two simple decisions in quick succession — classify a tone, then classify a light — and the second response is delayed in direct proportion to how closely it follows the first. The delay persists after thousands of practice trials. The interpretation: decision-making passes through a central bottleneck that handles one response selection at a time. Perception can overlap, movement can overlap — but the choosing stage is strictly serial.
Task-switching experiments add the second cost. Every time you switch from task A to task B, performance on B suffers a switch cost — typically a few hundred milliseconds of slowing plus a burst of errors — because your brain must tear down one task set (rules, goals, response mappings) and load another. Worse, the abandoned task leaves attention residue: part of your mind keeps processing it, quietly degrading whatever you switched to.
Why heavy multitaskers perform worst
The most counter-intuitive finding in this literature comes from Stanford's studies of chronic media multitaskers — people who habitually juggle streams, chats and feeds. Compared with light multitaskers, heavy multitaskers performed worse on nearly every component skill: they were more distractible, worse at filtering irrelevant information, and — the twist — slower at task-switching itself. Practice at multitasking does not build multitasking ability; it appears to erode the filtering that focused work depends on. Self-rated multitasking skill, meanwhile, correlates with actual performance close to zero, and in some studies negatively.
Real dual-tasking does exist — narrowly
The honest caveat: genuine parallel performance is possible when at most one of the tasks needs central decision-making. Walking while talking works because walking is automatised. Experienced pianists can sight-read while shadowing speech because years of training pushed reading-to-fingers below the level of conscious selection. The rule of thumb from the research: you can combine one attention-demanding task with any number of fully automatic ones — but never two attention-demanding tasks without cost. The skill worth training is therefore not "doing two things at once" but two adjacent abilities: automatising components until they stop needing attention, and switching between tasks cleanly when switching is unavoidable.
Practical rules from the lab
- Serialise by default. Batch messages, close tabs, finish the thought before answering the ping. Every avoided switch is a few hundred milliseconds and an error risk saved.
- Finish or park before switching. Attention residue is worst for unfinished tasks. Writing a one-line "next step" note before switching measurably reduces the carry-over.
- Protect deep work from choice. The bottleneck is in decision-making — so remove decisions from the environment (notifications, open feeds) rather than relying on willpower to ignore them.
- Train switching as a skill. Switch costs shrink (though never vanish) with practice on rapid alternation tasks — useful for jobs where interruption is structural, like support, trading or on-call work.