Case Study 9.1: Gloria Mark and the 23-Minute Recovery — The Science of Attention Interruption

The Number That Changed How We Think About Interruption

Twenty-three minutes and fifteen seconds. That figure — the average time a knowledge worker takes to return to their original task after an interruption — has become one of the most widely cited statistics in discussions of workplace productivity, digital wellbeing, and the attention economy. It appears in articles about smartphone addiction, corporate productivity consulting reports, and school district policies on phone use. It has been quoted by senators, cited in judicial proceedings, and referenced in dozens of subsequent academic papers.

The researcher behind the finding, Gloria Mark, is both credited with it and, in a certain sense, burdened by it. The 23-minute statistic has taken on a life of its own, repeatedly simplified and overgeneralized in ways that Mark herself has had to publicly clarify. Understanding what she actually found — and what she didn't — requires going back to the research itself.

The Original Studies: Watching People Work

In 2004, Gloria Mark was a professor in the Department of Informatics at the University of California, Irvine, when she and colleagues Daniela Gudith and Ulrich Klocke conducted what would become her most influential study. The methodology was deliberately low-tech: they sent trained observers into two companies and had them watch employees work. Not through monitoring software or self-report surveys — literally watching. Observers followed individual workers throughout their workdays, recording on paper every time the worker switched their attention from one task to another, and timing how long each activity lasted before the next switch.

What Mark and her colleagues were studying was the natural ecology of interruption in knowledge work — who interrupted whom, how often, and what happened afterward. The companies involved were a financial services firm and a software development company, both with workforces doing cognitively demanding, sustained-attention work: analysis, writing, programming, planning.

The findings were striking. Workers were interrupted, on average, every three minutes and five seconds. The vast majority of interruptions came from other people — colleagues stopping by desks, phone calls, impromptu meetings. A smaller but significant proportion came from electronic sources: email notifications, instant messages, phone calls. And the recovery time — the time between when a worker was interrupted and when they resumed the original task at full attention — averaged 23 minutes and 15 seconds.

The 23-minute figure was not a measure of how long workers were derailed from work entirely. Workers resumed doing something productive fairly quickly after most interruptions. What the figure measured was how long it took workers to return to the same task they had been doing when interrupted, at equivalent depth of engagement. The gap included the time spent dealing with the interruption, plus the time spent on other tasks that arose during the recovery period (what Mark called "resumption lag"), plus the time required to rebuild the cognitive state necessary to re-engage at the prior depth.

This distinction matters because the 23-minute figure is often misread as evidence that interruptions cause extended periods of non-work. What it actually captures is something more subtle and more consequential: the cognitive reconstruction cost of rebuilding context and depth after it has been disrupted.

Why 23 Minutes? The Cognitive Architecture of Deep Work

To understand why re-engagement is so time-consuming, it helps to understand what is lost in a deep-work interruption.

When a person is deeply engaged in a complex task, they maintain in working memory a rich structure of context: the problem they're solving, the approach they're taking, the decisions already made, the next steps planned, the constraints on their options. This mental model is not retrieved instantaneously from long-term memory — it is rebuilt and maintained actively, consuming cognitive resources and requiring sustained attention to keep intact.

Interruption collapses this structure. The contents of working memory that constituted the active task context are displaced by the interruption content. When the interruption ends, the original task must be re-approached essentially from scratch — the worker must re-read their notes, re-examine their earlier work, and patiently reconstruct the mental model that they had just spent focused time building. This reconstruction process, not the interruption itself, is what takes 23 minutes.

This is why, in Mark's research, workers who were interrupted while composing a report could not simply sit back down and resume typing. They had to figure out where they were, what they had been thinking, what came next, and why. The recovery was cognitive archaeology — excavating the state of mind that had been present before the interruption.

For some task types, this reconstruction is relatively fast. If you are interrupted while replying to a routine email, returning to that task takes little reconstruction — the context is simple and the next steps are obvious. For complex analytical or creative work, reconstruction takes much longer, and the 23-minute average captures the center of this distribution.

Translating to the Smartphone Era

When Mark's 2004 research was published, smartphones were not yet a widespread phenomenon. The interruptions she was studying came primarily from colleagues, phone calls, and email. The study's implications for smartphone notification behavior became apparent only as smartphone adoption accelerated between 2009 and 2014.

Mark herself turned her attention to this translation. In subsequent work, she and colleagues examined how email notifications specifically affected worker focus, finding in a 2012 study that workers who had their email turned off for five days showed not just lower levels of stress and lower heart rates (measured via portable monitors), but also reported being significantly more focused and productive. When email was restored, stress and multitasking frequency quickly rebounded. The experiment demonstrated that the interruption effects she had documented in 2004 were, in fact, causally produced by notification-driven communication rather than being an artifact of open-plan office culture or other confounds.

In subsequent years, Mark extended her research to smartphones, social media, and what she began calling "the ecology of attention" in digital environments. Her findings, synthesized in her 2023 book Attention Span: A Groundbreaking Way to Restore Balance, Happiness and Productivity, painted a picture of compounding interruption effects.

Several findings from her smartphone-era work are particularly relevant to the notification context:

Attention spans are shortening, but not in the way popular accounts suggest. Mark found that average consecutive time spent on a single screen before switching dropped from an average of 2.5 minutes in 2004 to 47 seconds by the mid-2010s. This is not evidence that humans are becoming cognitively incapable of sustained attention — it is evidence that the digital environment has been engineered to produce constant switching. The capacity for deep attention appears intact; what has changed is the environment's tolerance for it.

Self-interruption has become more common than external interruption. By her later research, Mark found that more interruptions in digital work environments came from the individual themselves — self-initiated task switches, checking social media, opening a new browser tab — than from external sources. This mirrors the "anticipatory checking" described in Chapter 9: the conditioned habit of checking has become so robust that it occurs independently of any external trigger. People interrupt themselves because they have been conditioned to do so, and the conditioning has become self-sustaining.

Negative emotional states predict higher rates of self-interruption. Mark's research found that workers who were frustrated, bored, or stressed were more likely to self-interrupt with phone or social media checks than workers in positive or neutral emotional states. This creates a particularly troubling loop: notification checking, which is itself mildly stressful (the uncertain anticipation of the notification, the emotional processing of its content), may increase the probability of subsequent self-interruption, which in turn increases stress. The cycle is self-amplifying.

Adaptation, but at a Cost

One of the more nuanced findings in Mark's later work is that workers adapt to high-interruption environments — but the adaptation has costs.

In her long-term observation studies, Mark found that workers in chronically high-interruption environments learned to work in shorter bursts, completing self-contained tasks rapidly rather than sustaining long focused sessions. They became, in a sense, functionally adapted to the fragmented environment. This adaptation reduced the apparent work disruption from individual interruptions, because workers had already accommodated to the expectation of interruption and structured their work around it.

But this adaptation came with a documented cost: elevated stress and reduced work quality on tasks that required sustained deep engagement. For tasks like writing a complex analysis or solving a multi-step problem — precisely the tasks that young students and knowledge workers most need to do well — the fragmented-environment adaptation was counterproductive. You can adapt to doing short tasks in a highly interrupted environment. You cannot fully adapt to doing deep work in one.

This is why Mark has been careful not to let her findings about adaptation soften her concerns about chronic interruption. The finding that workers adapt is not reassuring — it is a description of a coping mechanism that mitigates some costs of a problematic environment while leaving others intact and adding new ones (stress, reduced depth).

What Mark's Work Means for Maya

Maya does not work in the environments Mark studied. She is a teenager in Austin trying to complete homework. But the structural conditions are identical to those Mark documented in knowledge workers: she is attempting cognitively demanding tasks (writing essays, solving math problems, analyzing history) in an environment engineered for constant interruption.

If the 23-minute recovery figure applies even approximately to Maya's study context — and the available evidence suggests it does — the implications are severe. A two-hour study session interrupted four times (a conservative estimate for a teenager with notifications active) may contain less than 30 minutes of genuinely focused cognitive engagement, with the remaining 90 minutes spent in various stages of attention recovery and anticipatory checking.

Mark herself has spoken about adolescent phone use in interviews and public talks. In a 2019 interview with The Atlantic, she described being troubled by the normalization of constant connectivity among young people and the difficulty of doing genuinely focused intellectual work in such an environment. "We have an epidemic of distraction," she said, "and we're allowing it to continue unchecked in the environments where people most need to be able to think."

The 23-minute recovery number is, in this light, not primarily a statistic about workplace productivity. It is a measure of what is being lost every time a notification succeeds in doing its job.

The Limits and the Enduring Point

It is worth acknowledging the legitimate criticisms of Mark's research. The 23-minute figure comes from specific workplace settings with specific task types. It has not been precisely replicated in adolescent homework or creative work contexts, and there is real individual variation in recovery time. The figure applies to deep, cognitively demanding tasks — for simpler tasks, recovery is faster.

But the directional conclusion is robust and has been replicated across numerous subsequent studies using different methodologies: interruptions impose cognitive costs beyond their own duration, because the process of rebuilding focused attention is itself time-consuming and effortful. The precise number of minutes is less important than the principle, and the principle is well-established.

Gloria Mark's contribution is to have made measurable and concrete a cost that we all intuitively sense but rarely quantify. When we feel frustrated that we "never got anything done" despite being busy all day, what we are often experiencing is the cumulative weight of attention recovery time. When Maya finishes a two-hour homework session feeling vaguely unproductive, it may be because a significant portion of those two hours were spent in the cognitive aftermath of notification-triggered interruptions rather than in the focused engagement those hours were intended for.

The notification that interrupts is not a minor inconvenience. It is the beginning of a 23-minute clock.

Key Findings Summary

• Gloria Mark's 2004 research with Gudith and Klocke established the 23-minute attention recovery baseline in knowledge worker contexts, using naturalistic observation methodology.
• The 23 minutes measures the time to return to equivalent depth of engagement with the interrupted task, not total inability to work.
• Mark's subsequent research found that average single-screen focus time dropped from 2.5 minutes (2004) to approximately 47 seconds by the mid-2010s.
• Self-interruption became more prevalent than external interruption in later research, consistent with the conditioned checking described in Chapter 9.
• Workers adapt to high-interruption environments, but at the cost of increased stress and reduced capacity for deep work on complex tasks.
• Mark has explicitly connected her research to concerns about adolescent phone use in digital learning environments.

For further reading on Gloria Mark's research, see her 2023 book Attention Span and her publication archive at the University of California, Irvine. Her original 2004 CHI paper with Gudith and Klocke is available through the ACM Digital Library.