Infinite Scroll: The Stopping Cue That Was Removed

Before 2006, a list of things on the internet ended. You reached the bottom of the page and there was a row of numbers - 1, 2, 3, Next - and to continue you had to decide to continue.

That row of numbers was doing something nobody thought of as important. It was a stopping cue: a point at which the activity paused and asked whether you wanted more. Infinite scroll removed it. Now the page has no bottom, and the decision to keep going is never presented, because continuing requires no decision at all.

This is a good story and it is mostly true. It is also surrounded by claims that do not survive checking, including the one about who invented it. What follows separates the documented history from the folklore, and then, more importantly, marks the exact point where the evidence about what infinite scroll does to you runs out.

That point arrives earlier than you would expect.

Who invented it, honestly

The standard account is that Aza Raskin invented infinite scroll in 2006 at his company Humanized, and later came to regret it. Raskin says this himself. His own website states that he “worked on natural language interfaces and invented infinite scroll (2006),” and in a 2024 interview he put it plainly: “I did invent infinite scroll.”

The claim is contested, and by evidence that predates the controversy.

Hugh Williams, formerly of Microsoft, wrote on his own blog in March 2012, six years before the Raskin attribution became widely known, that “with two remarkable people, Nick Craswell and Julie Farago, I invented infinite scroll in MSN Search’s image search in 2005.”

There is a supporting document. Microsoft patent US7664739B2, filed 14 February 2006, with Farago, Williams and Whyte among its named inventors, uses the term explicitly: “To view off-screen objects, an infinite scroll component can be employed.” Windows Live Image Search shipped on 8 March 2006.

There was also a broader current. In June 2005 the designer Bill Scott published a post titled “Death to Paging!” announcing a component that replaced “the standard Previous/Next page controller bar” with a scrollbar, though in fairness that was buffered scrolling over a bounded dataset, not an endless feed. The idea of killing pagination was in the air.

We could not find any contemporaneous 2006 document from Raskin describing the invention; his old blog no longer hosts it, and the Humanized weblog archive was not reachable. We also found no statement in which he acknowledges the Microsoft prior art.

The honest sentence is therefore:

Raskin says he invented infinite scroll at Humanized in 2006; a Microsoft team says they built it for MSN image search in 2005 and filed a patent using the term in February 2006. Anything more confident than that is choosing a side of a live dispute.

Two further corrections, because both circulate widely. Raskin’s first-person expressions of regret that we could verify are from 2024 and 2026, not from the 2018 BBC Panorama segment; the 2018 regret is the journalist’s paraphrase, not a quotation. And he has not testified before the US Congress on this; the testimony people are thinking of was given in a New Mexico state trial against Meta in February 2026. His striking figures about “human lifetimes” wasted are unsourced and mutually inconsistent across two interviews by a factor of roughly six. We are not repeating them as facts.

What a stopping cue is

The concept is borrowed from food research, and the borrowing is worth being explicit about.

The classic demonstration is the bottomless bowl: diners fed from a bowl that secretly refilled ate substantially more than diners fed from a normal bowl, because the empty bowl, not fullness, was what told them to stop. The idea generalises as unit bias: Geier, Rozin and Doros (2006) showed in Psychological Science that people treat “a unit” of something as the appropriate amount, and that increasing the unit size increases consumption even when quantity is unlimited.

A necessary disclosure about the bottomless bowl. The 2005 paper is by Brian Wansink, whom Cornell found in 2018 to have “committed academic misconduct in his research and scholarship, including misreporting of research data, problematic statistical techniques, failure to properly document and preserve research results, and inappropriate authorship.” Around eighteen of his papers were retracted.

This particular paper does not appear to be among them. It is not on Wansink’s own list of retracted articles and carries no retraction flag in its Crossref record, but an independent analyst has argued its numbers do not add up, and the statistician Andrew Gelman, reviewing that analysis, concluded only: “Did the experiment ever happen? I just don’t know!”

We are citing it anyway, for one reason: a preregistered replication was published in 2024 in the Journal of Experimental Psychology: General by Lopez and colleagues, and the main finding held, with a smaller effect than originally reported. Cite the replication. Do not cite the original as a standalone authority.

And now the caveat that governs this whole section. All of this is food research. Nothing in it tests whether removing a stopping cue from a screen interface produces the analogous effect. The move from soup bowls to feeds is an analogy, a good one, and a useful way to see what the pagination bar was doing, but it is not evidence about scrolling.

Where the evidence actually stops

Here is the finding that should reshape how you read everything else written about infinite scroll:

We could not find a single experiment that randomly assigns people to an infinite-scroll feed versus a paginated feed and measures time spent, wellbeing, or attention.

We looked hard. What exists instead is:

Field studies of scrolling behaviour, which are observational. Rixen and colleagues (2023) ran a seven-day experience-sampling study with 46 participants across 6,605 sessions and 1,064 questionnaires. They found that people “reported higher levels of Valence on sessions that were not only composed of scrolling activity”: sessions that were purely scrolling felt worse. That is a correlation within people’s own reports, and the paper does not compare infinite scroll to pagination at all.

Intervention studies that do not include a pagination arm. Meinhardt and colleagues (CHI 2025) ran a seven-day field study with 72 participants testing an overlay that appeared after fifteen minutes of scrolling. Useful, and not a test of the design itself.

One small controlled comparison, whose result cuts the wrong way. Ruiz, Molina León and Heuer (2024) ran a between-subjects experiment with 30 university students, comparing infinite scroll against a version requiring a reaction to advance. The friction group showed better memory for what they had seen (d-prime = 2.89 versus 1.20, U = 31, p < 0.001). But they also spent more time, not less: 8.67 minutes versus 3.33 (p < 0.001), 53% reported frustration, and only 20% wanted the feature elsewhere.

Thirty students is not a basis for a general claim in either direction. But it is the closest thing to a controlled test of the central proposition that exists, and it did not find that infinite scroll increases time spent.

Survey studies, which cannot establish causation. A 2024 survey study in Bangladesh reports that 55.2% of respondents said triggering posts significantly affect their anxiety and fatigue. That is people’s account of themselves, and it belongs in a different category from the above.

So what can you actually say?

Supported: infinite scroll is widely deployed. People self-report difficulty stopping. Adding friction changes behaviour in small studies. The design demonstrably removes a decision point that used to exist.

The mechanism is documented far better than its effect is measured. We can show you the patent, the ship date, and the removal of the pagination control. We cannot show you an experiment demonstrating what that removal did to the people using it.

Not supported by anything we found: that infinite scroll causes more time spent than pagination would; that it shortens attention span; that it causes addiction in any clinical sense. These are the three claims the design is best known for and they are the three with no experimental basis we could locate. If you have read them stated as fact, and you have, they were stated beyond the evidence.

The honest position is that a design change of enormous scale was made without anyone measuring what it did, and that fifteen years later almost nobody has gone back and measured it. That is a more disquieting fact than the confident version, not a less one.

The regulation nobody actually passed

If you have the impression that infinite scroll is being banned, check what has actually been enacted. As far as we could verify: nothing, anywhere.

• The SMART Act (S.2314, Sen. Josh Hawley, introduced 30 July 2019) would have prohibited infinite scroll, autoplay and auto-refill. It died without a vote.
• The European Parliament adopted a report in October 2023, 38-0 with one abstention, naming “infinite scroll, default auto play, constant push and read receipt notifications” as features that should be prohibited. It is a non-binding own-initiative report urging the Commission to legislate. It is not law.
• Massachusetts passed S 3164 through its state Senate 38-2 in July 2026, which would disable “addictive feeds, ‘autoplay’ and ‘infinite scroll’ capabilities” by default for minors. One chamber. Not enacted.
• The UK has a proposal and a public consultation, with the Prime Minister quoted saying the mechanisms “need to go.” Not law.

And the one significant law that did pass, California’s SB 976, signed in September 2024, does not contain the word “scroll.” We checked the bill text. It regulates “addictive feeds,” defined by algorithmic recommendation based on user data. That is a different design feature. Conflating the two misstates what the law does.

What this leaves you with

A design decision, made twice by different people around 2005 and 2006, that removed the last structural point at which an interface asked whether you wanted to continue. The removal is documented. The dispute over credit is documented. The regret is documented, if later than usually claimed.

What is not documented is the part everyone is certain about.

The useful thing to take from this is not a conclusion about harm, because the evidence does not support one, but a piece of noticing. The reason you scroll for longer than you meant to is not that a page is compelling you. It is that the moment when you used to decide has been deleted, and nothing has replaced it. You can put one back. A timer, a tab close, the bottom of a saved article. It does not have to be much. It only has to be a place where the question gets asked.

Infinite scroll sits inside a wider set of design decisions covered in the persuasive design guide. For what happens when the scrolling is specifically of bad news, see the doomscrolling loop; for the same behaviour at night, why you can’t stop scrolling in bed. A related removal of a stopping point, the streak that never lets you take a day off, is in the Duolingo streak.