Doomscrolling and what it actually does to your nervous system.

The word doomscrolling entered the public vocabulary during the first wave of the pandemic, and it has stayed because it names something most adults now recognise in themselves: the involuntary thirty- or sixty- or one hundred and twenty-minute drift through bad news, threaded with outrage, threaded with celebrity, threaded with one good joke, that ends with a vague sense of being slightly worse off than when it began. The behaviour predates the word. The architecture that produces it does not.

What follows is a calm look at what the available evidence says about what happens to your physiology, your sleep, and your mood when you do this every day for an hour, and what the trial data shows actually helps.

§ The architecture: intermittent reinforcement on tap

Every infinite feed — Twitter/X, Instagram, TikTok, the comments under a news article, Reddit — runs on the same psychological substrate that makes a slot machine effective. You pull, sometimes you get nothing, sometimes you get a small reward, very rarely you get something extraordinary. This is intermittent variable-ratio reinforcement, and the behavioural literature is unambiguous: it is the schedule that produces the most persistent, hardest-to-extinguish behaviour of any known reinforcement structure.

The product designers know this. They did not arrive at infinite scroll, autoplay, and pull-to-refresh by accident. These mechanisms exist because they keep the user engaged for marginally longer per session, and the business model — advertising revenue priced per session — depends on those marginal minutes. The feed is not a neutral information surface. It is a behavioural delivery system, and you are the behaviour.

"The feed is not free. The price is that your attention has been turned into a commodity and your nervous system has been turned into the factory."

§ What this does to the stress system

The autonomic stress response is exquisitely calibrated for short, intense events: a near-miss with a car, a difficult conversation, a sprint up a flight of stairs. Cortisol rises, the sympathetic branch fires, the parasympathetic damps. After the event the system returns to baseline within minutes or hours. This is healthy. The system is meant to do this.

What it is not meant to do is maintain a low-grade activation for hours at a time, every day, for years. Robert Sapolsky's body of work on chronic stress in baboons and humans showed three decades ago that the costs of chronic low-grade sympathetic activation are not in the moment but in the slow erosion of systems that depend on parasympathetic dominance to do their work: digestion, immune function, sleep architecture, emotional regulation.

Doomscrolling sits squarely in this regime. Each piece of bad news, each piece of provocative content, produces a small spike. The spikes do not feel like much in the moment because they are small. They add up to a day in which the system never returns to a true rested baseline. The body learns to operate at a slightly elevated set point. That set point is the substrate for the sleep complaint, the gut complaint, the irritability, the difficulty concentrating on long-form work that most adults now report as a vague background condition of modern life.

§ What the longitudinal data shows

The peer-reviewed evidence on social media use and mental health is genuinely contested, in part because effect sizes for general population averages are small. The evidence on heavy use, and on adolescent and young-adult use specifically, is less contested. Twenge and Haidt's work on adolescent mental health across thirty-two countries showed substantial associations between heavy smartphone-based social media use and depressive symptoms, particularly in adolescent girls, with the relationships strongest where use exceeded three hours a day.

Two important randomised trials inform the picture. Hunt's 2018 study at Penn randomly assigned undergraduates to limit social media to ten minutes per platform per day and found significant decreases in loneliness and depressive symptoms over three weeks compared to a control group that used social media normally. Allcott's 2020 paper in the American Economic Review randomised paid Facebook deactivation for four weeks and found measurable improvements in subjective wellbeing and a reallocation of time to social activities and television (which is, on most measures, less harmful).

These are not earth-shaking effect sizes. They are real, they replicate, and they point the same direction: people who reduce their feed time feel better, sleep better, and report better attentional control within weeks.

§ The sleep mechanism, specifically

The most consistent finding across the literature is the relationship between evening feed use and degraded sleep. Three mechanisms operate at once:

• →The light from the device suppresses melatonin and delays the circadian phase, particularly within the two hours before intended sleep.
• →The content itself activates the sympathetic system — outrage, novelty, social comparison — at the exact moment the system needs to wind down.
• →The intermittent reinforcement schedule produces a behavioural loop that delays sleep onset; one more swipe, one more video, one more thread.

The result is later sleep onset, more fragmented sleep, less REM, and a wake the next morning that is measurably worse on every cognitive metric the lab can measure. The compounding effect across years is the most worrying part of the data.

§ What actually helps — by effect size

The interventions with the largest effect sizes in trials are, dispiritingly, the simple ones:

• →Remove the device from the bedroom. Charge it in another room. This single change has the largest documented impact on time-to-sleep and morning subjective alertness.
• →Delay the first feed-check of the morning by at least an hour. The first hour after waking sets the autonomic tone for the day. Letting that tone be set by a stranger's outrage is a self-imposed handicap.
• →Replace, do not restrict. The mind, denied one source of intermittent reinforcement, will route to the next. Have a deliberate replacement ready — a book in the bedroom, a walk after coffee, a habit that occupies the hand and the eye.
• →Use the platform's own tools. App timers, grayscale mode, and disabling autoplay all reduce time-on-feed by margins large enough to show up in the data.
• →Reduce frequency before reducing total time. Five short check-ins a day is less bad for autonomic tone than one ninety-minute session, even if total minutes are similar — but both are worse than three deliberate check-ins.

§ A closing point

None of this is moralistic. The feed is engineered by some of the best behavioural scientists in industry to do exactly what it does, and the failure to resist it is not a failure of character. It is the predictable outcome of a system designed to be irresistible meeting a brain that evolved before the system existed.

The useful framing is mechanical, not moral. The device is a tool. The feed inside the device is a behavioural delivery system. The cost of letting it operate freely is paid in your sleep, your attention, and your baseline stress physiology. The interventions that work are the ones that put friction between you and the loop, and the data on those interventions is, at this point, clear enough to act on.