Leroy — Attention Residue


Author	Sophie Leroy
Work	”Why is it so hard to do my work? The challenge of attention residue when switching between work tasks”
Year	2009
Venue	Organizational Behavior and Human Decision Processes, 109(2), pp. 168–181
Type	Peer-reviewed paper
AM axis	Cognitive ergonomics (candidate lens)
Tags	attention-residue, task-switching, context-switch, focus, interruption
Actionable	5 / 5

Leroy named and measured the thing every developer running several agents at once feels but rarely quantifies: when you switch tasks, part of your attention stays stuck on the last one. She called it attention residue, and showed it degrades performance on the task you’ve just moved to.

Summary

Switching leaves residue. When you move from Task A to Task B, your attention does not follow cleanly. A residue stays with A — especially if A was left unfinished or under time pressure — and it measurably impairs your performance on B.
Unfinished tasks leave the most residue. Interrupting a task before a natural stopping point, or switching under deadline, leaves more residue than completing it does. The cost is worst exactly when you’re busiest and switching most.
The cost is cognitive, not just temporal. This is not the seconds lost re-orienting. It is degraded reasoning on the new task while part of your mind is still resolving the old one — a quality cost, not only a speed cost.

The one reusable technique

Reach a stopping point before you switch — and batch, don’t interleave. Leroy’s remedy for residue is cognitive closure: finish a bounded unit of work (or explicitly park it with a note capturing where you were) before turning to the next. Applied to focus: batch similar work and switch at natural boundaries rather than interleaving many open threads, because every mid-task switch imports residue into the next task.

Representative quote

“People need to stop thinking about one task in order to fully transition their attention and perform well on another. Yet, results indicate it is difficult for people to transition their attention away from an unfinished task.”

— Leroy, Why is it so hard to do my work? (2009)

What this means for agentic engineering

Attention residue is the direct cost of context-switching between concurrent agent sessions — and the clearest reason a candidate cognitive-ergonomics lens belongs in AM. Orchestration culture treats running many agents at once as pure throughput gain. Leroy shows the hidden tax: each hop between sessions imports residue, and your judgment on the session you switch to is quietly degraded while you’re still holding the one you left.

Every session hop has a cognitive price. Five agents running in parallel is not five times the output if you are round-robining between them — the residue from each unfinished thread degrades your review of the next. This is why “how many agents can I run?” is the wrong question; “how many can I supervise without my judgment degrading?” is the right one.
It maps to a measurable signal. The workspace toolkit surfaces attention metrics — context_switch_rate (project changes between consecutive prompts, per active hour) and interrupts_per_hour (prompts that hop to a different session) — that are, in effect, attention-residue proxies. See the cognitive-load lens.
Batch and drain at boundaries. The remedy — reach closure before switching — is exactly what the deep-work playbook operationalizes: batch interruptions into an asks queue and drain them at task boundaries rather than servicing them mid-task.