The 47-Second Attention Span Stat Is Being Quoted Wrong
The LinkedIn post always goes the same way. Someone pastes a statistic — "the average human attention span is now 47 seconds, down from 12 minutes in 2000" — adds a line about smartphones rotting our brains, and watches the engagement pour in. The comments divide between the alarmed and the smug. Nobody asks where the number came from. Nobody mentions what the researcher who measured it actually concluded.
That researcher is Gloria Mark, a professor of informatics at UC Irvine who has spent more than two decades tracking how people allocate attention at work. She is the source of the 47-second figure. She is also the source of the finding that nobody quotes alongside it: that roughly half of all attention switches are ones you choose to make.
That second part changes everything about what the statistic means.
What Gloria Mark Actually Measured
The 47-second figure comes from a sustained program of workplace observation research. Mark and her colleagues logged the computer activity of office workers — tracking time on screen before any switch to a different window or application. No self-reporting. No surveys. Objective logging.
The number that emerged from the most recent rounds of that research was 47 seconds of median screen focus before switching. The 2016 paper, "Email Duration, Batching and Self-interruption: Patterns of Email Use on Productivity and Stress" (co-authored with Shamsi T. Iqbal, Mary Czerwinski, and others, published at ACM CHI 2016, DOI: 10.1145/2858036.2858262), draws on exactly this kind of logging data.
What the stat is measuring: time on one screen before switching. Not how long you can hold a thought. Not comprehension. Not whether your working memory has degraded. Just the gap between one click and the next.
The same lab had measured the equivalent figure at roughly 2.5 minutes in 2004, in a paper titled "'Constant, constant, multi-tasking craziness': Managing Multiple Working Spheres" (González & Mark, CHI 2004). By 2012 the figure had dropped to around 75 seconds. By 2016, 47. That trajectory is real and worth paying attention to. But real doesn't mean it tells the story most people think it does.
The Statistic Nobody Quotes Alongside It
In a 2011 paper — "Why do I keep interrupting myself?: Environment, Habit and Self-interruption" (Dabbish, Mark, and González, CHI 2011, DOI: 10.1145/1978942.1979405) — the same research group tracked 889 hours of observed task-switching behavior across 36 information workers. The finding: roughly half of all interruptions were self-initiated. Not incoming notifications. Not a colleague tapping your shoulder. You, choosing to switch.
Mark has been direct about this in interviews. "It turns out from my research that we are just as likely to interrupt ourselves as to be interrupted from some notification," she told Steelcase in 2023. The paper traces what drives self-interruption: task difficulty, boredom, curiosity, a remembered obligation (prospective memory — "I should check if that person replied"), or simply habit.
The conventional read of the 47-second stat is that something is being done to you. Your environment is fragmented, your notifications are relentless, the apps are designed for addiction, and your brain is paying the price. All of that has some truth to it. But the self-interruption data inserts you as an active participant in your own fragmentation. You are not a passive casualty. You are frequently the agent.
This distinction matters because it changes what a solution looks like. A passive-victim framing points at the phone companies and the infinite scroll designers. A self-interruption framing points at habit, at what you're doing when work gets hard or tedious, at the small behavioral patterns you've built around relief and avoidance.
Both framings have evidence. The self-interruption one just doesn't generate as many outraged shares.
Where the Goldfish Came From
The 47-second stat regularly gets paired with another claim: that humans now have shorter attention spans than goldfish (8 seconds versus 9, the story goes). Mark's research is not the source of this claim. Nobody's peer-reviewed research is.
The goldfish figure traces to a 2015 Microsoft Canada consumer insights report — a marketing document, not a study, built on proprietary survey data from 2,000 Canadians and EEG sessions with 112 people. It compared those results to a figure for goldfish attention spans that itself had no cited source. The report was never peer-reviewed. The "goldfish" comparison was a rhetorical flourish that somehow became a reference point for serious discussions of human cognition.
Mark has noted the problem directly: "The goldfish result is not exactly correct, so that shouldn't be our starting point." When a corrective researcher has to correct the misuse of her own research by distancing it from a different piece of junk science that got attached to it, something has gone wrong with how we're reading these numbers.
The Real Cost Is Physiological, Not Moral
The research that should actually concern people isn't the 47-second figure alone. It's what Mark's 2008 paper documented about recovery.
"The Cost of Interrupted Work: More Speed and Stress" (Mark, Gudith, and Klocke, CHI 2008, DOI: 10.1145/1357054.1357072) found that workers who were interrupted compensated by working significantly faster to catch up — but they also reported higher levels of frustration, time pressure, and stress. A follow-up measurement that tracked refocus time put the figure at around 23 minutes to return fully to an interrupted task.
Later work extended this into physiology. The 2019 paper "Email Makes You Sweat" (CHI 2019, DOI: 10.1145/3290605.3300898) measured skin temperature — a proxy for physiological arousal — during email interruptions and found measurable stress responses. The workers who knew they were being observed and were told email was disconnected showed lower stress markers than those in continuous-access conditions.
The cost of fragmented attention isn't cognitive decline. It's chronic low-grade stress load that accumulates across a workday. You recover your thought. You pay for it in cortisol.
The moral framing — "you need more discipline," "just focus," "put your phone in another room" — treats this as a character problem. The physiological data suggests it's a design-environment problem intersecting with a habit problem. That's a more tractable thing to work on. But you can only see it if you read past the 47-second headline.
Why the Misread Has Consequences
The standard narrative built around the 47-second stat runs like this: smartphones and social media have degraded human attention capacity, we are now cognitively impaired compared to previous generations, and the fix is some combination of digital detox and personal discipline.
That narrative does several things badly.
It frames the problem as irreversible brain change, which is not what the data shows. Mark's research measures behavior under current conditions — the same logging that found 47 seconds in 2016 also found 2.5 minutes in 2004. Those are the same kinds of brains, in different environments, with different habits and tools.
It removes agency. If your attention span is simply shorter than it used to be, the implied fix is external — wait for better product regulation, sue the platforms, let some future UX designer build apps that don't grab you. If half the problem is self-interruption driven by avoidance and habit, the implied fix involves examining what you reach for your phone to escape from.
It makes the solution feel abstract. "The attention economy is destroying us" is a true and useless statement at the individual level. "I switch away from hard tasks when I get uncomfortable, and that switch has a 23-minute recovery cost I'm not accounting for" is the same phenomenon described in a way that someone can actually do something about.
Mark's 2023 book, Attention Span, argues that sustained focus was never the goal anyway — attention has natural rhythms, and the research shows it needs recovery periods the way muscles do. The problem isn't that people switch away from tasks. It's that they switch in ways that spike stress without providing actual rest, and they've built habits around those switches that make them feel automatic.
The 47-second figure is evidence of something real. It just isn't evidence of the thing most people use it to argue.
If the narrative we've built around your attention span tells you that you're a victim of forces beyond your control, ask what it's protecting you from having to look at. The research that actually matters isn't the headline number. It's the finding that when the switch happens, there's a roughly even chance you chose it.
That's a harder thing to tweet. It's also a more honest place to start.
If you're interested in how design exploits attention at the structural level, the post on AI-generated dark patterns and accountability covers what happens when the manipulation becomes untraceable.