Summary
Mark, Gudith, and Klocke studied information workers and found that after an interruption, people compensated by working faster — but at the cost of higher error rates and significantly increased stress. The average time to resume a task after interruption was 23 minutes. The key finding: interruptions aren’t just a time cost; they degrade the quality of subsequent work. Even “quick” interruptions carry a cognitive tax that compounds over a workday.
What it means for our work
This paper is the research foundation for Biff’s pull-based design and Vox’s opt-in model. Biff uses /mesg n to block interruptions and /read to check messages on your schedule — you decide when to context-switch. Vox defaults to silent and only speaks when explicitly enabled. Both tools are designed around Mark’s finding: the cost of an interruption isn’t the interruption itself, it’s the 23 minutes of degraded work that follows. Pull-based communication respects cognitive flow.