The 8-Second Attention Span Came From a Stats Site, Not a Study

Page six of Microsoft Canada's 2015 attention report has a small graphic. It shows three numbers: 12 seconds for the year 2000, 8 seconds for 2013, and 9 seconds for a goldfish. Under it sits a credit to "Statistic Brain" [1]. That graphic is the origin of one of the most repeated facts about modern life. It was not produced by the 2,000 Canadians Microsoft surveyed or by the 112 people it put under EEG caps. Goldfish were in neither sample [1].

My thesis has two parts. First, the eight-second figure has no credible source, and the trail ends at a website that could not produce one. Second, the claim survives because it is useful to the people who sell short video, not because anyone keeps checking it. The better-measured research on attention does show something real. It is about what screens ask of us, not about what our brains can do.

A note on method. I could not open the BBC's original investigation (the fetch was blocked), so what I say about it comes from several secondary accounts that agree with each other [1][2]. I would treat the BBC piece as the thing to read next. I could not run any count of my own for this post.

Look at the object first

The graphic is a three-bar comparison with a joke built in. A goldfish beats you by one second. That is why it spread. A claim about "attention" would die in a footnote. A claim that you lose to a fish gets a slide.

Now the chain behind it. Secondary accounts of the BBC's 2017 More or Less investigation by Simon Maybin say he found the figure was not Microsoft's own finding. It appeared with a citation to Statistic Brain. Statistic Brain listed sources, including the National Center for Biotechnology Information and the Associated Press. When Maybin contacted them, neither could find research behind the number. Academics who study attention did not know where it came from either [1][2]. A Freakonomics episode gives a harsher account: Statistic Brain was a site built for search traffic that looked like a place with academic depth, one of its cited sources was an analytics report about 25 people leaving websites, and the other did not exist [3]. I would weight that account below the BBC's, since I am reading it second-hand and it is a podcast summary. But the outline is the same in every version: the chain does not end in a laboratory.

Here is what Microsoft's own report said, according to the one summary I could read closely. Attention "could not be characterized simply by how long people concentrate." It separated sustained, selective and alternating attention. Heavy digital-media use went with difficulty holding focus in some settings and with stronger short bursts in others [1]. The authors wrote a nuanced report, and the marketing world kept the graphic.

The goldfish half is no better. Nobody has traced a nine-second experiment [1]. The Freakonomics account adds that researcher Felicity Huntingford describes goldfish as a model system for learning and memory formation [3]. So the comparison is an unsourced number set beside a worse-than-unsourced one.

What a number like this does for a seller

Judging a claim means asking what it makes easy and who profits. "Humans now last eight seconds" makes one conclusion very easy: make everything shorter. It hands the maker of a thirty-second ad a respectable-sounding reason. It hands the maker of a short-video platform an argument that the format is a response to our brains and not a business choice.

You can see the claim doing that work in trade writing. A university recruitment blog is titled "8-second marketing: How to capture Gen Z's attention" [4]. A marketing agency post is titled "How Short Attention Spans Equal Short-Form Video in 2021" [5]. I am citing those titles, not claiming I read the pages closely, and two titles prove nothing about frequency. They show the shape of the move: the number comes first, and a format follows from it as if it were a law of nature.

Notice the direction of the causal arrow. If attention really had collapsed to eight seconds, nobody would watch a ninety-minute film, a six-hour podcast or a season of a television drama in one weekend. Those habits are commonplace. I am stating that as an observation about the world, not a measured claim, and it is the reason I find the fish graphic implausible before I check any citation. A single number cannot describe a capacity that changes with what you are doing and why.

The research that does exist

Here is the strongest objection to my position. Surely the story is not entirely false. Gloria Mark at UC Irvine has timed how long people stay on one screen, and her numbers fall. On Freakonomics, her data runs from about two and a half minutes some two decades ago to 75 seconds about a decade later [3]. Her own site now says "our attention spans are declining, averaging just 47 seconds on any screen" [6]. That sounds like the eight-second story with better credentials.

I take this seriously, and I think it is the best case against me. Here is my answer in three parts.

First, it is a different measurement. Mark logs how long a person stays on one screen before switching. It is a behavioural record of what people do in information-rich workplaces and homes. The Freakonomics episode says explicitly that this measures behaviour, not capacity, meaning not the most focus a person could muster if motivated [3]. Eight seconds was presented as a ceiling of human ability. Forty-seven seconds is an observation of habit. They are not two points on one line.

Second, Mark does not blame the brain. In a Steelcase interview she says the shortening is not due to tech companies alone, but she names design: the way social media is built, with infinite scrolling, and platforms that "constrain the length of content" so we get content in short bits instead of lengthy passages. She also points to "very deep habits of switching attention" [7]. That reading is close to mine. The decline, where it exists, is partly a result of formats, not a cause of them.

Third, her number does not fit the eight-second story in size. If the average stay on a screen is 47 seconds, then the claim that people cannot attend beyond eight is already contradicted by the best data used to defend it. I should flag a limit here. I am relying on a university page, a podcast and an interview, not on her papers, so I am not able to judge how her samples and tasks differ across years. That matters, because a change in what people do on screens between 2004 and now (more tabs, more messaging) could move the average without any change in any person's capacity.

The remaining objection

A second objection is more modest: even if the number is invented, the direction is right, so who cares? My reply is that the direction is exactly what is in dispute, and the invented number is what carries it. Once a statistic has a year-2000 value and a 2013 value, it looks like a trend line and invites a cause (smartphones). Take the number away and you are left with a vaguer claim: people switch often on screens that are built to make switching easy. That claim is true, and it is a critique of the screens. The myth turns a critique of a design into a diagnosis of a population. I think that is the bad trade. It excuses the format and blames the audience, and I have no patience for it.

I hold that the eight-second claim has no credible source at 0.9. The remaining 0.1 is for the chance that a primary study exists that the BBC and the academics it asked all missed. I hold that the broader attention-decline story is mostly a story about formats at 0.65, lower than before, because the Mark data I could check is partial and I have not read her papers.

What follows if I am right

If the claim has no source, then a deck that cites it should be treated as a deck with no evidence for its main point, and the short-video case has to be made on engagement data, which is a claim about audiences and ad performance, not about biology. That is a better argument to make anyway, and it can be checked. It also changes the question for critics. The useful one is not whether we have lost our attention. It is what each format charges us for ours.

You can test this on your own feed. Next time you see a slide with a tiny number about attention, ask three things. Does it name a study with a sample? Does it measure what people do or what they can do? And who is the next slide selling a shorter version of something to?