Interrupted office workers finished faster but more stressed

A lab study built around a realistic office task

Gloria Mark, Daniela Gudith and Ulrich Klocke's paper “The Cost of Interrupted Work: More Speed and Stress” appeared in the ACM CHI 2008 proceedings, in April 2008. The full paper, confirmed on Mark's own publication list, describes a lab experiment with 48 participants, mostly German university students, role-playing a human resources manager answering a folder of twelve emails as quickly, correctly and politely as possible. Each participant worked through the task under three conditions: no interruptions, interruptions from a “supervisor” on the same topic as the email task, and interruptions on an unrelated topic, with interruptions arriving roughly every two minutes by phone or instant message.

What was measured, and the counter-intuitive result

The researchers expected interruption context to matter and expected interrupted work to take longer. Neither happened. There was no significant difference between same-topic and different-topic interruptions, and, surprisingly, the uninterrupted baseline condition took longer (22.8 minutes) than either interrupted condition (about 20.3 to 20.6 minutes). Email quality, measured by errors and politeness, did not differ significantly across conditions. What did differ was self-reported workload: participants in both interrupted conditions rated significantly higher stress, frustration, time pressure and effort than the uninterrupted group, measured on a standard NASA workload scale after each block of emails.

What this does and does not show

The paper's own interpretation is that people facing frequent interruptions compensate by working faster and writing shorter emails, not that interruptions are costless. That compensation carries a measured price in subjective strain, recorded after only about 20 minutes of interrupted work; the paper explicitly says it cannot show whether people adapt to that strain over longer periods or whether it accumulates. The sample was small, mostly psychology and science students in a simulated task, not a longitudinal study of working professionals, and the “interruption” was a scripted question from an experimenter rather than the varied real interruptions of an actual office. Applying the finding to, say, a chat-heavy remote job requires treating it as a mechanism demonstration, not a direct measurement of that job.

Questions to ask before you adopt it

• Is a policy aimed at protecting completion speed, output quality, or the person's subjective workload, since this study found these move independently?
• Does the interruption volume in your setting resemble roughly one every two minutes, or is it far less frequent?
• Would a short lab session capture the accumulated cost of stress over a full working day or week?

The finding that people compensate for interruptions by rushing, at a cost to how they feel rather than to what they produce, is one of the more precise and durable results in this literature. It is still a single lab session with a scripted task, and the honest reading is about that mechanism, not a general verdict on every open-plan office or every notification.