Multitasking, and what it actually costs

People who describe themselves as good at multitasking are generally describing rapid switching, and the research on switching is fairly unambiguous.

What is actually happening

Tasks requiring central attention cannot be performed simultaneously. What occurs is alternation — attention moving between them.

Task switching studies consistently find switch costs: responses are slower and less accurate on the trial after a switch than on a repeat trial.

The costs are measured in fractions of a second in laboratory tasks. In real work, where tasks are complex and the reorientation involves reconstructing context, they are considerably larger.

What can be combined

Worth being precise, because the blanket claim that multitasking is impossible is too strong.

Tasks can be combined where they use different processing resources and where at least one is automatic.

Walking and talking. Driving a familiar route and holding a conversation — with the important caveat that the driving degrades measurably, which is why hands-free phone use is associated with impairment despite the hands being free.

Listening to instrumental music while doing routine work.

What cannot be combined is two tasks both requiring language processing, or both requiring decisions, or both requiring working memory — which describes most knowledge work.

Which is why listening to a podcast while writing does not work, and why people who believe it does are generally not attending to one of them.

The self-assessment problem

A robust and slightly uncomfortable finding.

Studies of self-described heavy media multitaskers have generally found them performing worse on measures of attention filtering and task switching than people who report multitasking less.

The causal direction is not established — it may be that people with more distractible attention gravitate toward multitasking rather than the reverse — and the correlation with confidence is the striking part.

People are poor judges of their own performance while switching, because the experience of being busy is easily mistaken for the experience of being productive.

Where the cost lands

Time. The most obvious and the least important.

Errors. More significant. Switch costs show up in accuracy as well as speed, and in work where errors are expensive this dominates.

Depth. Frequent switching prevents the sustained attention that complex problems require. A problem worked on in five-minute fragments is not being worked on.

Memory. Attention at encoding predicts later recall. Material processed while switching is retained poorly, which matters for anything involving learning.

Practical responses

Batch by type. Grouping similar tasks reduces switch cost because the mental context is shared.

Answering all email in one block is cheaper than answering it throughout the day, by more than the time saved.

Single-window working. Physically removing the possibility rather than resolving against it.

Close the tabs. Each visible tab is an invitation, and self-interruption is a large share of total interruption.

Capture rather than act. When an unrelated thought arrives, write it down and continue. The cost of the note is seconds; the cost of following it is minutes.

Finish something. Incomplete tasks occupy attention. Completing a small one before starting a large one frees capacity.

The organisational version

Individuals are frequently blamed for a problem the workplace created.

A culture expecting immediate response to messages guarantees switching. A tool that notifies on every message guarantees it. A workload requiring parallel projects guarantees it.

None of that is fixed by a person deciding to concentrate harder, and the interventions that work at that level are agreements about response times and notification defaults rather than exhortations.

The honest exception

Some work genuinely is coordination, and switching between threads is the job rather than an obstacle to it.

For those roles the goal is not elimination but protecting a portion of the day from it, so that the work requiring depth has somewhere to happen.

An hour is usually enough, and most people currently have none.

Where it is genuinely fine

The prohibition should not be absolute.

Combining an automatic physical task with something demanding is usually harmless — walking while thinking, washing up while listening.

The costs appear when two activities compete for the same resource, which in practice means anything involving language at the same time as anything else involving language.

That single rule covers most of the practical cases without requiring a general theory.