Attention Residue: Why You Are Still Thinking About the Last Task

Short answer: attention residue is the part of your mind still working on the previous task after you have moved to the next one. It was named and measured by Sophie Leroy in 2009, and the finding that matters is narrower than the way it gets quoted. Residue is worst not when a task is merely interrupted, but when it is interrupted unfinished and under time pressure. And there is one intervention, tested and published, that reliably reduces it.

Almost everything else you have read on this topic is a chain of citations back to two papers, at least one of which is being misquoted.

What Leroy actually did

The 2009 experiments used a simple structure. Participants worked on a task, then switched to a new and unrelated one. Before and after the switch, they were given a lexical decision task, a standard tool for detecting what concepts are still active in your mind. If words from the first task were still being recognised unusually fast, part of your attention was still there.

It was. And the amount left behind depended on one thing more than any other: whether the first task felt finished.

Participants who completed their first task, or who were given the chance to plan how they would finish it, carried much less residue into the second. Participants who were pulled away mid-task under time pressure carried the most, and performed measurably worse on the task in front of them.

People need to stop thinking about one task in order to fully transition their attention and perform well on another.

Sophie Leroy, Organizational Behavior and Human Decision Processes, 2009

The mechanism is not mystical. An unfinished task with an open deadline remains an active goal. Active goals consume working memory whether or not you have consented to it.

The 23-minute figure, and where it comes from

You have seen this number. "It takes 23 minutes and 15 seconds to refocus after an interruption." It is usually attributed to Gloria Mark's CHI 2008 paper, The cost of interrupted work: more speed and stress.

That paper does not contain the figure. It reports something more interesting, and considerably less convenient for the productivity industry.

Interrupted participants finished faster, with no drop in quality. They also reported significantly more stress, more frustration and more effort. All the stress differences were significant at p < .01.

So the honest version of the finding is not "interruptions cost you 23 minutes." It is: you pay for interruptions in effort and stress, not in elapsed time. You compensate by working harder and faster, and the bill arrives as depletion rather than as a missed deadline. Which is worse in a way that is harder to see on a calendar and much easier to feel on a Thursday afternoon.

The intervention that has evidence behind it

In 2018, Leroy and Glomb published the follow-up that most articles about attention residue never reach. It tested a specific, small, almost embarrassingly simple intervention: the ready-to-resume plan.

Before switching away from an interrupted task, you take a short moment, roughly a minute, and write down where you are and what the next step is. Not a to-do list. A note to your future self about the state of the work.

Participants who did this carried less attention residue into the interrupting task and performed better on it. The mechanism follows directly from the 2009 finding: the task stops occupying working memory as an open goal because you have externalised its state. It is closed enough.

This is the rare productivity recommendation that is both trivially cheap and actually tested. Most are one or the other.

What this means for a normal working week

The unit to protect is task completion, not calendar blocks. A ninety-minute block that ends with the task half-done and no note leaves residue for the rest of the day. Two forty-minute blocks that each end at a natural boundary do not. Length is not the variable people think it is.

Write the resumption note before you switch, every time. One line. "Stopped at the pricing section, next: check whether the Q3 figures include VAT." This is the ready-to-resume plan and it is the single highest-yield item here.

Notice which interruptions carry time pressure. Leroy's residue was worst under anticipated time pressure on resumption. An interruption when you have three days left is not the same event as the same interruption when you have three hours. The clock, not the interruption, is doing much of the damage.

Batch by context where you can. Same-context and different-context interruptions produced similar effects in Mark's data, so do not over-invest in this one. But it costs nothing.

Stop measuring your day by how long you were "in flow". The stress cost of interruption does not show up as lost minutes, so a time-tracking app will show a productive day and you will still feel wrung out. That gap is the finding, not a personal failure.

We wrote a longer piece on making concentration work inside a job that does not protect it: deep work in a job that will not let you. And if your interruptions are mostly self-inflicted through a device, digital minimalism without the dogma covers the other half.

What residue looks like in an actual week

The laboratory version is a clean switch between two tasks. The version you live in is messier, and it is worth naming the specific shapes it takes.

The meeting that ends at the hour. You leave a discussion mid-decision and walk into the next one. Neither task is closed, both carry anticipated time pressure, and by the third meeting you are carrying residue from two. This is the highest-residue structure in ordinary office life and it is produced entirely by calendar defaults.

The message you read but do not answer. Opening a message loads its content into working memory and opens a goal. Deciding not to respond yet does not close it. This is why reading messages without acting on them is more expensive than it feels, and why the common advice to "check messages but only reply at set times" is the worst of both arrangements.

The task you finished but did not mark finished. Completion in Leroy's experiments was partly subjective. Work that is objectively done but sitting in an ambiguous state, unsent, unmerged, awaiting your own review, behaves like an open goal because you have not told yourself it is closed.

The thing you are dreading later today. Anticipated time pressure was the amplifier in the 2018 study. A difficult conversation at 4pm degrades the work you do at 11am, and the mechanism is the same one: an active goal you have not planned your way through.

Two things that sound like solutions and are not

Longer focus blocks, on their own. If a two-hour block ends with the task unfinished and no plan written, the residue is the same as after twenty minutes. Block length is not the operative variable in any of this research. Boundary quality is. A short session that closes cleanly beats a long one that does not.

Turning off notifications and nothing else. This reduces the number of interruptions, which is worth doing, and it does nothing about the residue from the interruptions you still take, or from your own unfinished work, which in most jobs is the larger source. Removing the trigger does not close the goals already open.

The uncomfortable implication of both is that most of what gets sold as focus advice targets interruption frequency, and the measured cost sits in interruption quality — whether the thing you left was left closed.

What the research does not say

It does not say multitasking lowers your IQ, that figure comes from a small unpublished 2005 study commissioned by a technology company and should not be in circulation.

It does not say every interruption costs 23 minutes. Nothing in the cited papers supports a fixed recovery time, and the actual measured completion times moved in the opposite direction.

It does not say open-plan offices cause attention residue. That may well be true, but these experiments were laboratory task-switching studies and cannot carry the claim.

The strength of this literature is that it is narrow and well-measured. It is worth keeping it that way rather than borrowing its authority for a broader argument it never tested.