Why you can't stop scrolling
You didn't decide to be here for forty minutes. You decided to check one thing. That gap isn't a character flaw. It's a design that works exactly as intended.
The usual explanation is that your phone gives you a dopamine hit, and you keep chasing it. It's a tidy story. It's also the part of this that the research supports least.
What the evidence does support is stranger, and more useful: the thing holding you there is probably not the reward. It's the timing of the reward.
The mechanism is a schedule, not a substance
Most of what you scroll past is forgettable. Occasionally something genuinely lands — funny, useful, or exactly what you needed to see. You can't predict which one, and you can't predict when.
That pattern has a name. In behavioural psychology it's a variable-ratio schedule: a reward that arrives after an unpredictable number of attempts. It was described in detail by B. F. Skinner in 1953 and mapped out exhaustively in Schedules of Reinforcement a few years later. Of every reinforcement pattern studied, this one produces the most persistent behaviour and the hardest to extinguish.
It's how a slot machine works, and not as a metaphor. It runs the same underlying schedule. A feed that reliably gave you something good every tenth post would be easier to put down, because you'd know where the finish line was. The not-knowing is the active ingredient.
A predictable reward is easy to leave. An unpredictable one is what keeps the hand moving.
The dopamine story is weaker than you've been told
Here's where the popular version gets ahead of the evidence. A 2021 study published in iScience used PET imaging to measure dopamine synthesis capacity in 22 adults, alongside logs of their actual, unconstrained phone behaviour rather than self-reports.
The people whose phone time skewed most heavily toward social apps showed lower dopamine synthesis capacity in the putamen — not higher. The opposite direction to the story usually told online.
Two caveats matter here. Twenty-two people is a small sample. And the study is correlational: it can show that two things move together, but not that one causes the other. The relationship could run the other way, or something else could drive both.
The accurate takeaway is narrower than "scrolling drains your dopamine." The confident version you've heard isn't settled science, and the one real imaging study points the other way.
And twenty minutes probably isn't the problem
In 2024, researchers put 59 people aged 13 to 55 through twenty-minute sessions of social media and of YouTube, measuring heart rate and cortisol throughout.
Neither spiked. Both drifted downward across the session, regardless of what people were doing.
Which is worth sitting with if you've been treating every scroll as self-harm. A short scroll is not, by itself, a measurable physiological stressor. What changes the picture is when it happens, how long it actually runs once the schedule takes over, and what it quietly replaces — which, at 11pm, is usually sleep.
What this changes about stopping
If the pull came from weak discipline, more discipline would fix it. It doesn't, reliably, and now you know why: you're up against a reinforcement schedule that was characterised seventy years ago precisely because it is so resistant to being extinguished.
Which points the fix somewhere other than willpower in the moment.
None of this makes the feed less engineered. It just moves the decision to a point in the evening where you're still the one making it.
- Skinner, B. F. (1953). Science and Human Behavior. Macmillan. — original description of variable-ratio reinforcement. Expanded in Ferster, C. B. & Skinner, B. F. (1957), Schedules of Reinforcement.
- Westbrook, A. et al. (2021). Striatal dopamine synthesis capacity reflects smartphone social activity. iScience. PubMed 34113831. — n=22, [18F]-DOPA PET imaging alongside logs of natural, unconstrained smartphone behaviour. Correlational.
- Boise State University (2024). Acute physiological responses to short-duration social media and video use. PLOS ONE. — n=59, ages 13–55, heart rate and cortisol measured across 20-minute sessions.
Common questions
The popular version of this is stronger than the evidence. A 2021 study using PET imaging found that a higher proportion of social app use tracked with lower dopamine synthesis capacity in the putamen, not higher — the opposite direction to the usual claim. It was a small, correlational study of 22 adults, so it cannot show cause either way. The honest summary is that the dopamine story is far less settled than it sounds online.
The strongest explanation is the reward schedule rather than the reward itself. Most posts are unremarkable and an occasional one is genuinely good, so the payoff arrives unpredictably. Unpredictable reward timing — a variable-ratio schedule — produces the most persistent, hardest-to-extinguish behaviour of any reinforcement pattern studied. It's the same schedule a slot machine runs on.
Not measurably, on its own. A 2024 study of 59 people found that a twenty-minute session of social media or YouTube did not spike heart rate or cortisol — both actually drifted downward across the session. That suggests the problem is less about a short scroll being physiologically stressful, and more about when it happens, how long it runs, and what it displaces.
Changing the schedule tends to work better than relying on willpower in the moment, because the pull comes from unpredictable rewards rather than weak discipline. Practically: move the scroll earlier in the evening rather than trying to cut it entirely, put a real physical gap between the phone and the bed, and give the last stretch of the night a different, duller default.
We are building the things that make the end of the day quieter.
dusk embrace is building the small things that end a night well, starting with the one that covers your eyes. Come be first.
Get early access