In 2004, office workers stayed on one screen for an average of two and a half minutes before switching. By the late 2010s, Gloria Mark’s UC Irvine research put that figure at just 47 seconds — and half of all screen-attention periods lasted 40 seconds or less.

Two and a half minutes can feel almost spacious beside 47 seconds. Yet those are the bookends Gloria Mark uses to describe one of the longest-running changes her workplace research has documented: how long people remain on a screen before moving elsewhere.

In the early fieldwork, conducted in 2003 and published in 2004, office workers stayed with a screen or information resource for roughly two and a half minutes on average. In studies from about 2016 onwards, Mark says the average screen-attention period settled near 47 seconds.

The median in the later data was 40 seconds. In other words, half of the observed periods ended within 40 seconds.

The numbers are striking, but their unit is easy to lose. They describe time on one screen before a switch. They do not show that the human brain has acquired a 47-second biological limit, that nobody can read for a minute, or that every switch represents distraction.

The 2004 baseline came from people, stopwatches and 477 hours of observation

Víctor González and Mark’s 2004 CHI paper followed 14 information workers at one organisation: managers, financial analysts and software developers. Each person was observed for about three and a half days, producing 477 hours of fieldwork. Researchers recorded events, documents, devices, conversations and the projects to which they belonged.

The paper reported that people spent about three minutes on an event and “somewhat more than two minutes” using an electronic tool or paper document before switching. In later interviews, Mark has summarised this early screen baseline as approximately two and a half minutes.

The method was laborious and the sample was small. Its advantage was context. An observer could see that opening an email, speaking to a colleague and returning to a spreadsheet might all belong to one project, even though the tools changed.

It was also a study of information workers in a particular workplace, not a representative census of every adult in 2004. The number is best understood as a carefully timed field observation, not a universal norm.

Software logging produced the 40-second median

Later research replaced much of the manual shadowing with unobtrusive computer logs. A 2016 study by Mark and colleagues tracked the online activity of 40 information workers over two work weeks. Its reported median duration of online-screen focus was 40 seconds.

The researchers combined logging with surveys, sleep measurement and physiological data. Shorter focus duration was associated with lower productivity ratings at the end of the day. Some personality and stress measures were also related to switching, although an observational study cannot establish that one caused the other.

Mark later placed the average across the more recent studies at 47 seconds. In a University of California interview, she noted nearby replications of 44 and 50 seconds. She described the sequence as roughly 150 seconds in the original work, 75 seconds in 2012 and 47 seconds in the later datasets.

This is a pattern assembled across studies. It is not the same group of people followed from 2004 into the late 2010s. Participants, workplaces and measurement tools changed, so the downward line is informative without being a controlled experiment on the cause of the decline.

The average and median answer different questions

An average of 47 seconds does not mean every screen visit lasts exactly that long. A worker might spend 20 minutes editing a presentation, then make five rapid visits to a calendar, chat thread, inbox and browser. The long session and every brief check enter the distribution.

The median is the middle observation after all durations are ordered. At 40 seconds, it tells us that half of the measured periods were 40 seconds or less and half were longer. It does not say that half of the people were incapable of focusing beyond 40 seconds.

Mark calls this restless movement “kinetic attention.” The phrase captures a pattern in which attention is not simply present or absent. It is active, repeatedly redirected and often spread across several tools.

A screen switch is not necessarily a change of thought

Suppose someone writing a report moves from a document to a browser to verify a figure, then opens a spreadsheet containing the underlying data. The screen has changed twice, but the goal has remained stable. A jump from the report to an unrelated message about another deadline is a different kind of transition.

The logging measure can record both. It therefore should not be promoted as a clinical “attention span” test or compared with unsupported claims about goldfish. An American Psychological Association interview with Mark makes the narrower definition explicit: time on a screen before switching.

Nor does the research prove that one device or platform caused the historical shift. Notifications, shorter media formats and larger numbers of communication channels may contribute. People also interrupt themselves, sometimes because another task has come to mind and sometimes because their workplace has made constant checking seem prudent.

Frequent movement can carry a cost even when the switches are voluntary

Short screen periods do not automatically mean poor work. Some jobs require comparison, monitoring or fast coordination. A useful interruption can supply the answer that allows a task to continue.

The cost appears when switching forces someone to replace one set of goals and details with another, then reconstruct the first context later. Mark has reported associations between faster switching and stress, measured through both self-reports and heart-rate data. The 2016 study also linked shorter online focus with lower assessed productivity at day’s end.

A 2018 workplace experiment offers cautious evidence that the environment matters. When software blocked selected online distractions, participants rated their focus and productivity higher. The effects differed among people, and a temporary blocking tool is not proof that every worker needs the same restrictions.

Silicon Canals has previously examined why the famous 23-minute return interval after an interruption is not pure recovery time. It includes the other work people do before returning to the displaced project. The 47-second figure sits at a smaller scale, but the two findings connect: rapid digital movements can become a chain that carries the original goal out of view.

The architecture of attention is partly organisational

Advice about focus often ends with the individual: mute alerts, close tabs, batch email and resist the reflex to check. Those measures can help. They operate inside expectations that workers do not set alone.

If a delayed reply is treated as disengagement, an inbox is no longer merely a tempting distraction. If different projects use different channels, checking becomes part of the job. If every message can be urgent but nothing is labelled consistently, attention is spent repeatedly testing whether it is safe to ignore the next alert.

Teams can reduce that uncertainty by agreeing which channels are synchronous, what response times actually matter and when uninterrupted work is legitimate. Focus blocks have little value if colleagues continue to expect immediate answers throughout them.

The fall from minutes to seconds should not be read as a verdict that people have forgotten how to think. It is evidence that digital work has become organised around much faster movement. Forty-seven seconds is not a timer inside the brain. It is a measurement of the environment through which attention now has to travel.