The research behind every video
Every video stands on peer-reviewed research. Here is every paper we used, what we left out, and why. No sponsors, no affiliate links.
- 43studies
- 95%in top or leading journals
- 206quotes checked word for word
- 12videos, each approved by an editor
Choosing sources means leaving some out. These studies were considered and not used. Here is why.
Jebb et al. 2018 (Gallup World Poll, over 1.7 million people worldwide) · Nature Human Behaviour
Answers the same question worldwide and found a levelling-off for emotional well-being at $60,000 to $75,000. It uses the same Gallup yes/no questions about yesterday as the 2010 study; the 2023 paper showed that such questions cannot tell happy people apart. We only had the abstract.
Arslan 2024 (letter: The miseries that remain may yet be alleviated) · PNAS (letter)
A letter with a re-analysis of the published summary data, not a study of its own. It argues that the evidence for a levelling-off among the least happy is not strong and that a straight line fits about as well. We name it in the description because it questions a finding the video reports.
Rohrer & Wenz 2024 (letter: Inappropriate causal assumptions underlie Killingsworth, Kahneman, and Mellers' conclusions) · PNAS (letter)
A letter about method, not a study. It argues that these data cannot show that income causes happiness and that 'the unhappy group' is not a fixed group of people. The video therefore speaks of a link, not a cause.
Dwyer & Dunn 2022 (Wealth redistribution promotes happiness) · PNAS
An experiment: 200 people received $10,000 and reported more happiness afterwards. That is a one-off gift, not income, so it answers a neighbouring question. We only read the abstract. Worth knowing next to our caveat that the three studies show a link, not proof.
Newman et al. 2025 (Income and education show distinct links to health and happiness in daily life) · Nature Human Behaviour
Also uses daily reports and found that higher income went with higher well-being, but it compares income with education and does not test whether happiness levels off at higher incomes. We only read the abstract.
Easterlin et al. 2010 (The happiness-income paradox revisited) · PNAS
A different question: whether happiness in a country increases over the long term as that country's income rises.
McGuire et al. 2022 (meta-analysis of cash transfers in low- and middle-income countries) · Nature Human Behaviour
A different question and population: the effect of cash transfers on well-being and mental health in low- and middle-income countries, not whether day-to-day happiness levels off at higher incomes.
Loh et al. 2020 (activity breaks during sitting, meta-analysis) · Sports Medicine
Also includes moderate and vigorous activity breaks, not only light walking, and overlaps with the 2022 meta-analysis we used.
Henson et al. 2016 (standing or walking breaks) · Diabetes Care
A single small trial (22 women); the 2022 meta-analysis pools trials like this one.
Gallardo-Gómez et al. 2024 (dose of physical activity and HbA1c) · Diabetes Care
About long-term average blood sugar (HbA1c) with exercise in general, not the rise in blood sugar after a meal.
Ball et al. 2002 (ACTIVE trial) and follow-ups (JAMA 2006, J Am Geriatr Soc 2014) · JAMA
A strong trial in 2,832 older adults, but it used classroom training with an instructor, not brain-training games. Trained abilities improved; no effect on daily life was seen at 2 years, while the 10-year follow-up reported less self-reported difficulty with daily tasks. A different question, and worth its own video.
Simons et al. 2016 (Do "brain-training" programs work?) · Psychological Science in the Public Interest
A thorough review that reaches the same conclusion. We used original experiments; only the abstract was available.
Bediou et al. 2018 (action video games, meta-analysis) · Psychological Bulletin
About action video games and perception and attention, not brain-training games.
Jamshed et al. 2022 (early time-restricted eating) · JAMA Internal Medicine
Points the other way: in 90 adults, eating early in the day plus fewer calories gave 2.3 kg more weight loss than fewer calories alone. It lasted 14 weeks, against a year for the trials we show, and meta-analyses like the BMJ one include trials like it.
Lin et al. 2023 (time-restricted eating vs calorie counting) · Annals of Internal Medicine
Fits the question and points the same way (12 months, no significant difference). Left out of the video to keep it around a minute.
Lowe et al. 2020 (TREAT trial, 16:8) · JAMA Internal Medicine
The comparison group ate three structured meals a day, not fewer calories. It answers a neighbouring question.
Patikorn et al. 2021 (umbrella review) · JAMA Network Open
A review of earlier meta-analyses; the BMJ network meta-analysis from 2025 is newer and compares the diets directly.
Chieng et al. 2022 (coffee subtypes, UK Biobank) · European Journal of Preventive Cardiology
Uses the same UK Biobank data as the JAMA Internal Medicine study we used, split by coffee type. Specialist journal.
Klatsky et al. 2011 (hospital admissions for arrhythmia) · The Permanente Journal
An older observational study in a health-plan population, in a specialist journal. The larger and newer UK Biobank study answers the same question.
Review 2023 (coffee and cardiac arrhythmias) · Cardiology Journal
A narrative review in a specialist journal; we used the original studies.
Hoover & Milich 1994 · Journal of Abnormal Child Psychology
All boys got a placebo, but some mothers were told it was sugar; those mothers rated their sons as more hyperactive. A striking result, but it answers a neighbouring question (how parents perceive behaviour), not what sugar does. Specialist journal; only the abstract was available.
Wender & Solanto 1991 · Pediatrics
A small trial (17 children with ADHD, 9 without) with a single dose of sugar. The 1995 meta-analysis pools blinded studies like this one.
Studies on sugar intake and ADHD symptoms (e.g. Front Nutr 2026, Nutr Res Pract 2011) · various
Observational: they compare children who eat more or less sugar, so they cannot show what sugar itself does. Our question is about blinded trials.
Saint-Maurice et al. 2020 · JAMA
Fits the question and points the same way (8,000 vs 4,000 steps: lower mortality in 4,840 US adults). Left out of the video to keep it around a minute; the two meta-analyses pool many cohorts like this one.
Rodríguez-Gutiérrez et al. 2024 (umbrella review) · Preventive Medicine
A review of existing reviews in a specialist journal. The Lancet Public Health meta-analyses analyse the cohort data more directly and are in a leading journal.
Xu et al. 2024 (umbrella review) · BMJ Open
A review of reviews covering many outcomes; our question is only about dying early.
Paluch et al. 2023 (harmonized meta-analysis) · Circulation
About heart and vascular disease, not death from any cause.
Macpherson et al. 2013 · American Journal of Clinical Nutrition
A meta-analysis of trials that also included people who were already ill (secondary prevention). The 2022 review for the US task force covers only adults without heart disease or cancer, and is more recent. Only the abstract was available.
Sesso et al. 2012 (Physicians' Health Study II, heart disease) · JAMA
Same trial as the cancer paper we used; the deaths from all causes are reported there. This paper is about heart disease.
Fortmann et al. 2013 · Annals of Internal Medicine
The earlier evidence review for the US task force. It was replaced by the 2022 update, which we used.
Linxian trials (Blot et al. 1993) · Journal of the National Cancer Institute
Tested vitamin and mineral combinations in a region of China where many people lacked nutrients. Our question is about healthy adults without a deficiency.
Fawzi et al. 2004 · New England Journal of Medicine
Multivitamins for people with HIV in Tanzania: a different population and question.
ATBC trial 1994 (vitamin E and beta-carotene) · New England Journal of Medicine
Tested single vitamins in male smokers, not a multivitamin.
Raman et al. 2004 · American Journal of Physiology – Renal Physiology
Shows a wide range in daily water intake (1.2 to 7.7 litres), but makes the same point as the much larger Science study.
Belenky et al. 2003 · Journal of Sleep Research
Supports the finding that attention does not fully recover after three recovery nights, but only the abstract was available to check.
Prokopidis et al. 2023 (meta-analysis) · Nutrition Reviews
After a published letter pointed out double counting, the authors reanalysed their data: the overall effect on memory was no longer significant. The corrected figures sit behind a paywall we could not check automatically.
Xu et al. 2024 (meta-analysis) · Frontiers in Nutrition
Not on our list of top and leading journals, and it mixes patients with healthy people.
Van Cutsem et al. 2020 · Medicine & Science in Sports & Exercise
A different question: mental fatigue, not getting smarter.
McMorris et al. 2006 · Psychopharmacology
A different question: performance during sleep deprivation. Saved for a separate video.
Schöllhorn et al. 2023 · Communications Biology
Strong direct evidence that less blue screen light shortened the time to fall asleep, but the journal is not on our list of top and leading journals.
Silvani 2022 · Shechter 2020 · Brosnan 2024
Considered for an earlier draft. Dropped for a weaker method, a journal outside our list, or answering a different question (screens rather than blue light).