There is no single human attention span number, and the 8-second figure has no traceable study. Microsoft Canada printed it in 2015 and credited a website called Statistic Brain, whose listed sources had no record of it. Your attention varies with the task, and office screen studies measure how often you switch.
The myth has outlived its debunking. In a King's College London survey of 2,093 UK adults, run in September 2021 and published in 2022, 50% said the 8-second claim was true and 25% called it false. I wanted to see where the number came from, so I followed its citations through two archived web pages and a BBC investigation.
The claim runs like this: the average attention span fell from 12 seconds in 2000 to 8 seconds, one second short of a goldfish. You'll find it on page 6 of Attention spans, a Spring 2015 report from Microsoft Canada's Consumer Insights team. The page gives 12 seconds for 2000, 8 seconds for 2013 and 9 seconds for a goldfish, and credits one source, Statistic Brain. Microsoft's original link now returns a 404, so I read the copy the Wayback Machine saved in May 2015.
Microsoft's own research measured no attention span in seconds. Its team surveyed 2,000 Canadians online in late 2014 and recorded the brain activity of 112 people with EEG that December (pages 11 and 12). On page 9 the authors explain that they split attention into three types because they doubted attention could be summed up as how long people concentrate. Microsoft borrowed the report's most quoted number from someone else.
Statistic Brain's site no longer loads, but the Wayback Machine kept its attention span page from February 2012, three years before Microsoft's report. The 8 February 2012 capture labels 8 seconds as the average attention span in 2012, next to 12 seconds for 2000 and 9 seconds for a goldfish, and credits the Associated Press. In the 19 March 2015 capture, the same 8 seconds belongs to 2013, and the credit line adds the National Center for Biotechnology Information at the US National Library of Medicine. The number held still while its year moved.
Simon Maybin checked those credits for the BBC in 2017, in a More or Less episode on the BBC World Service and an accompanying article. He contacted the National Center for Biotechnology Information and the Associated Press, and neither could find any research behind the figures. His attempts to reach Statistic Brain went nowhere. The attention researchers he interviewed had no idea where the numbers came from.
The trail ends there. I can't trace the 8 seconds past Statistic Brain, and the BBC couldn't either, so I treat the figure as folklore. The King's College London report traces the myth to Microsoft on page 11, though the archive puts the figure online three years before Microsoft printed it.
I found no study that gives goldfish a 9-second memory or attention span. Professor Felicity Huntingford, who had spent almost half a century studying fish behavior, told the BBC in 2017 that goldfish can manage the kinds of learning described in mammals and birds. She added that scientists use goldfish as a model for studying how memories form, because they learn and remember.
One primary study shows how far that memory reaches. In Gee, Stephenson and Wright (1994), eight goldfish learned to press a lever for food. The researchers narrowed the paying window to 1 hour in every 24, at the same time each day, and held that schedule for 4 weeks. The fish began pressing ahead of feeding time. After the researchers switched the feeders off, the anticipation persisted for a limited number of days. A goldfish that times a once-a-day meal carries information across 24 hours, far past 9 seconds.
The best measured number I found for knowledge work is 47 seconds, and it measures something narrower than attention span. Gloria Mark of UC Irvine, with colleagues from Microsoft Research and MIT, logged the computers of 40 information workers at a large US tech company for about 12 business days. In their CHI 2016 paper, the mean time on one screen before a switch was 47.0 seconds (sd 21.4), and the median was 40.2 seconds.
The breakdown by app reads like a normal workday. People stayed in their email client for 61.8 seconds on average, and in Word, Excel, PowerPoint and similar software for 64.0 seconds. Chat got 42.5 seconds; in 2016 that meant Skype, instant messaging and Lync, the Slack of the day. Each person switched between apps about 273 times a day, plus 132 switches inside apps, between documents or websites (Table 1 of the same paper).
Read the method before you quote the figure. Mark's team divided each person's logged computer time by their number of screen switches. The result tells you how often people changed windows. It can't tell you how long they could have focused with nothing pulling at them, and the authors present switching as a stand-in for attention on a computer.
You can move that number by changing the setup. In a CHI 2012 study, Mark, Stephen Voida and Armand Cardello cut off email for 13 information workers for five workdays. Mean time in a window rose from 75.5 seconds with email to 131.9 seconds without it, and window switches fell from 37.1 to 18.2 an hour. The people and their jobs stayed the same, and the inbox was gone. If that inbox-and-chat shuffle sounds familiar, I wrote about it in the fragmentation problem, and about the cost of each switch in how long it takes to refocus after an interruption.
Mark's studies show time per screen falling. In 2004, Victor González and Mark shadowed 14 information workers, logged each activity with time stamps, and found people spent 2 minutes 11 seconds with any device or paper before switching. The 2012 email study logged 75.5 seconds per window at baseline, and the 2016 study logged 47.0 seconds per screen. On the page for her book, Mark describes the 47 seconds as a decline in attention span.
Those three numbers come from three methods. The 2004 figure came from human observers and includes phones and paper. The later two came from software logs, calculated two different ways, on 13 and 40 people. I read the series as a trend in switching, measured on small samples, with no test of capacity in it.
The King's College London researchers reach the same place on page 5 of their report: without long-term studies, they can't tell whether attention spans have declined. Their respondents feel the decline anyway, with 49% saying their attention span is shorter than it used to be and 23% disagreeing (page 6).
Collective attention has sped up, and that has a primary number. Philipp Lorenz-Spreen and colleagues analyzed hashtags from 43 billion tweets. In their 2019 Nature Communications paper, a hashtag in the hourly global top 50 stayed there for 17.5 hours on average in 2013 and 11.9 hours in 2016. They found the same steeper rise and fall in books, films, Google searches and Reddit threads. Their model attributes the speed-up to rising production and consumption of content, which exhausts each topic's audience sooner. The authors keep this population-level attention separate from the individual attention that psychologists study, and so should anyone quoting them.
Psychologists who study sustained attention, also called vigilance, don't ask for a number of seconds. They set a dull watch task and track how detection changes over time. Norman Mackworth began the experimental study of it in 1948 with a clock test: watch a moving hand and report the rare moment it jumps two steps instead of one. Detection fell, and the drop was robust after the first half hour on the task.
I couldn't open Mackworth's 1948 paper, which sits behind a publisher paywall, so this account comes from Klein and Feltmate's 2025 review of 75 years of vigilance research. The review reports a detail from Mackworth's later monograph that matters for your workday. Urging people to pay closer attention did nothing. A phone call or a half-hour break reduced or removed the decline, and feedback on performance had the same effect.
Dr Gemma Briggs, a psychologist at the Open University, gave Maybin the plainest summary: "It's very much task-dependent." She told the BBC that an average attention span means little, since the attention people apply depends on what the task demands and what they bring to it.
Claim | Where it came from | What the evidence says |
|---|---|---|
The average attention span is 8 seconds | Page 6 of Microsoft Canada's 2015 report, credited to Statistic Brain | No traceable study. The listed sources found no record of it (BBC). |
The attention span was 12 seconds in 2000 | The same Statistic Brain table, online by February 2012 | No traceable study. |
A goldfish's attention span is 9 seconds | The same Statistic Brain table | No study found. Goldfish learned a once-a-day feeding window in a 1994 experiment. |
People spend 47 seconds on a screen before switching | Measured: mean time per screen for 40 office workers. It measures switching. | |
Attention spans are shrinking | Press coverage of the 8-second claim | No long-term data on capacity (King's College London). Screen switching and topic turnover have sped up. |
Full disclosure: I build Hugo, so weigh this section with that in mind.
Hugo doesn't try to raise your capacity. It works at the moment of the switch. Hugo is an AI focus app for Mac: you start a session by giving it a goal, and in Hugo AI mode it judges each new tab, chat, video and app against that goal. If it's unsure about a page, it asks how the page helps your goal, and you type a real reason. "Open anyway" stays available at the last step.
Set that next to the two findings above. Mark's team changed the setup, and switching changed. Mackworth told people to try harder, and nothing changed. In Deep Work and Hugo AI sessions, the switch you'd make on reflex can meet that question first: the inbox tab mid-draft, or the other client's deck while you're on this one. If the break findings persuade you, Pomodoro mode runs work and break cycles and pauses enforcement during the breaks.
Download Hugo and run one session on the task you keep leaving. The 14-day trial covers everything, and you won't need a card.
Ready to stop losing time to distractions?
Hugo uses AI to keep you on task. No blocklist required. Set your goal, and Hugo evaluates every tab in real time.
Set it once. Hugo keeps you on it. 14-day free trial, no card.