Who the Studies Were Run On
The most-quoted statistic in psychology is two separate counts nested into one sentence, and both of them belong to somebody the retelling leaves out. A decade of alarm moved the American share down by roughly ten points and handed the difference to Britain, Canada, Australia and Western Europe. Consumer research, meanwhile, has audited its platforms obsessively and has never audited its countries.
Answer to the question
Are WEIRD samples still a problem in psychology?
As published in 2010, 96 per cent of subjects in the top journals of six psychology subdisciplines came from Western industrialised countries and 68 per cent from the United States. A decade on, American authors and samples had fallen only to a little over 60 per cent, with the slack taken up by other Western countries.
Key takeaways
- As published in 2010, 96 per cent of subjects came from Western industrialised countries and 68 per cent from the United States, across the top journals of six subdisciplines for 2003 to 2007.
- The famous undergraduate figure is a separate count from a single journal, and both numbers belong to Jeffrey Jensen Arnett rather than to the paper that made them famous.
- A decade later the American share had fallen to a little over 60 per cent, and the change came almost entirely from other English-speaking and Western European countries.
- No audit of participant country or culture in the leading consumer and marketing journals was found, and the field's own recent audit codes samples as students, Mechanical Turk or other.
You are reading a consumer finding and deciding whether to spend money on it. Ask first who it was measured on. For most of modern psychology the answer is a narrow slice of humanity, and the shorthand for that slice is WEIRD samples: Western, educated, industrialised, rich and democratic.
Where did the famous number come from?
Joseph Henrich, with Steven Heine and Ara Norenzayan, printed the count in Behavioral and Brain Sciences in 2010. Across the top journals of six subdisciplines of psychology between 2003 and 2007, 96 per cent of subjects came from Western industrialised countries and 68 per cent came from the United States. The same paper gives the comparison that bites: those countries hold 12 per cent of the world's population, and a randomly selected American undergraduate is more than four thousand times more likely to be a research participant than someone outside the West.
The undergraduate figure everybody quotes is a separate count. In the Journal of Personality and Social Psychology, 67 per cent of the American samples were only undergraduates in psychology courses.
Two counts, two denominators, one sentence.
The most-quoted statistic in psychology is two separate counts of two separate things, nested into one sentence.
The circulated version nests them, so that 96 per cent of participants were Western and about seventy per cent of those were American undergraduates. That is the form Henrich uses in his own preface to The WEIRDest People in the World, and it is defensible as a gloss, because 68 per cent of subjects were American and 67 per cent of American social-psychology samples were students. What it loses is the journals, the years and the two different denominators. Both underlying figures belong to Jeffrey Jensen Arnett, from a 2008 audit that Henrich and colleagues cite rather than conduct and that we have not read in the original.
This publication repeated the nested version for a year before reading either source. The evidence for WEIRD samples travelled further than its own frame.
Have WEIRD samples got any wider?
Amber Gayle Thalmayer, with Caterina Toscanelli and Arnett, re-counted the same six journals for 2014 to 2018. Authors and samples fell from over 70 per cent American to a little over 60 per cent. The change came mainly from other English-speaking and Western European countries. Eleven per cent of the world's population is now represented in those journals and 89 per cent still is not.
Leah Petrutiu and colleagues ran the newest audit, across five British Psychological Society journals from 2021 to 2023, coding 1,293 samples. Western countries supplied 67.6 per cent, non-Western countries 18.6 per cent, and 13.1 per cent named no country at all. Their own reading of that silence:
> "Given the tendency to leave groups considered normative unexplained ... it is possible that many of these were also recruited in the West."
Add that unstated share to the Western one and eighty per cent of what they coded could be Western. That is our arithmetic on their two figures. They close by quoting Quinn McNemar, writing in 1946, that the existing science of human behaviour is largely the science of the behaviour of sophomores. Eighty years on, WEIRD samples are still most of what the literature holds. One limit: it is a British audit of British journals, and its authors say it is not a clean comparison with Thalmayer's window. Who gets left out of a sample carries a cost, as [Why Women's Health Was Left Out](https://thewhyweproject.com/articles/why-womens-health-was-left-out) sets out.
Who does consumer research actually test?
Consumer research has counted its platforms with great care and has never counted its countries. Joseph Goodman and Gabriele Paolacci, in a 2017 tutorial for the Journal of Consumer Research whose note records equal and alphabetical authorship, analysed 1,350 surveys and experiments across four volumes. Mechanical Turk carried 27 per cent of them, rising from 9 per cent in the first volume to 43 per cent in the last. Then the sentence that explains the situation:
> "The majority of workers reside in the United States, and most researchers restrict participation to US residents in order to increase homogeneity."
Read as a mechanism rather than a lapse, it is obvious. Homogeneity lowers variance, lower variance makes an effect easier to detect, and a detectable effect is what gets published. WEIRD samples are not a correction the field keeps forgetting to make. They are the output of a design preference working as intended.
Homogeneity lowers variance, lower variance makes an effect easier to detect, and a detectable effect is what gets published.
The field's own recent audit records that preference by what it omits. Evan Weingarten, On Amir and Andrea Morales examined methods, topics and authorship across the Journal of Consumer Research, the Journal of Marketing Research and the Journal of Consumer Psychology, coding each sample as students, Mechanical Turk or other. There is no country in it and no culture.
We searched those three journals and the Journal of Marketing for an audit coding participant country or culture, in several phrasings, and found none. The nearest is in food and sensory science, where Otterbring and Folwarczny coded author affiliations rather than samples, and opened by saying nobody had measured the WEIRD bias there before.
Crowdsourced workers match or beat undergraduate WEIRD samples on attention and conscientiousness, produce the same biases at similar effect sizes, and let researchers without a laboratory run properly powered work. Goodman and Paolacci's warning stands: more representative than students does not mean representative.
Does a narrow WEIRD sample make findings wrong?
Richard Klein and 185 colleagues ran Many Labs 2: 28 published findings, 125 samples, 15,305 participants, 36 countries. Exploratory comparisons revealed little heterogeneity between WEIRD and less WEIRD cultures, and variability in the effect sizes tracked which effect was studied more than where. [What Forty-One Participants Buys](https://thewhyweproject.com/articles/what-forty-one-participants-buys) already put that project against the hidden-moderator defence of social priming.
Which WEIRD samples findings travel
The reply came in the same journal in 2024, in a commentary whose co-authors include Norenzayan, Heine and Henrich. Robin Schimmelpfennig and Rachel Spicer, sharing first authorship, report that 39 per cent of the Many Labs 2 pool was United States-based and that participants came mostly from university subject pools and Mechanical Turk, which skew toward education, digital literacy and socio-economic status. They twice call the project a milestone: this is an argument about design.
Two of their points need nothing further. Many Labs 2 had not planned to examine cultural moderation and did so in an after-the-fact exploratory analysis, which its own abstract confirms. And it tested moderation by decomposing the letters of the backronym, which these authors say was only ever a rhetorical device, country-level measures being, on their reading, a poor test. A preprint they also cite, which we have not read, reports that more formal education predicts cultural alignment with the United States, and we put it no higher than that.
The 2010 paper's own review of the universals sorts the candidates. Industrialised populations sit at the edge of the measured distribution on social motivation, fairness, folkbiological and spatial cognition, and some visual illusions. They do not on colour perception, the false-belief task, analog numeracy or facial emotion recognition, where the main effect across populations accounted for 58 per cent of variance against 9 per cent for cultural specificity.
The two claims therefore have different standing. Composition is a count, three audits over twenty years agree, and it has barely moved. Invalidity is a hypothesis, tested a few times, with results that depend on what counts as a different culture. So discount by subject rather than reject wholesale.
A perceptual or developmental result probably travels. A result about fairness, generosity, self-regard or the response to a nudge was measured where that response is most extreme and most movable: non-student American adults offered 47 per cent of the stake in dictator games against freshmen's 32 per cent, and a religious prime that shifted student offers in Canada did nothing among non-religious non-student adults.
Such a finding is untested outside that pool rather than false, and the audits say where the untested part is: majority-world authors and WEIRD samples at 4 to 5 per cent of the top psychology journals, and Latin America, Eastern Europe, the Middle East and Africa together at 8.7 per cent of the newest British count.
Which leaves a buyer of research with two possible worlds. In one, the effect in front of you is a fact about people, and it survives contact with a market in Lagos or Jakarta. In the other it is a fact about educated Anglophone adults who answer questionnaires for money, and a finding can [replicate perfectly inside a population and still be local](https://thewhyweproject.com/articles/the-psychology-you-learned-didnt-replicate). Those are different products. Nobody is pricing them differently.
Common questions
Are WEIRD samples still a problem in psychology?
Yes, and the improvement so far is mostly substitution. American authors and samples in six leading journals fell from over 70 per cent to a little over 60 per cent across a decade, with the change coming from other English-speaking and Western European countries. A 2021 to 2023 audit of five British journals put Western samples at 67.6 per cent, with 13.1 per cent of samples naming no country. Add that unstated share to the Western one, as we do above, and the figure could reach eighty per cent.
What does WEIRD stand for in research?
Western, educated, industrialised, rich and democratic. Henrich, Heine and Norenzayan used it in 2010 to label the populations psychology had mostly tested, and WEIRD samples became the field's shorthand. A 2024 commentary co-authored by all three says the backronym was only ever a rhetorical device, and that decomposing its letters into country-level measures is a poor test of cultural variation.
Do marketing journals report where participants came from?
We looked, in several phrasings, across the leading consumer and marketing journals, and found no audit that codes participant country or culture. The field's own recent audit of three of them codes samples as students, Mechanical Turk or other, with no geography in the taxonomy.
Sources
Henrich, J., Heine, S. J., & Norenzayan, A. (2010). 'The weirdest people in the world?' Behavioral and Brain Sciences, 33(2-3), 61-83. doi:10.1017/S0140525X0999152X
Arnett, J. J. (2008). 'The neglected 95%: Why American psychology needs to become less American.' American Psychologist, 63(7), 602-614. doi:10.1037/0003-066X.63.7.602
Thalmayer, A. G., Toscanelli, C., & Arnett, J. J. (2021). 'The neglected 95% revisited: Is American psychology becoming less American?' American Psychologist, 76(1), 116-129. doi:10.1037/amp0000622
Petrutiu, L., Birney, M. E., Cooke, R., & Stewart, S. (2026). 'WEIRD but Also Inconsistent: An Analysis of the Reporting Practices of Participant Samples Across Five Areas of Psychology.' International Journal of Psychology, 61(2), e70168. doi:10.1002/ijop.70168
McNemar, Q. (1946). 'Opinion-attitude methodology.' Psychological Bulletin, 43(4), 289-374
Goodman, J. K., & Paolacci, G. (2017). 'Crowdsourcing Consumer Research.' Journal of Consumer Research, 44(1), 196-210. doi:10.1093/jcr/ucx047
Weingarten, E., Amir, O., & Morales, A. C. (2025). 'The evolving field of consumer research through the lens of its top journals.' Marketing Letters, 36(2), 227-242. doi:10.1007/s11002-024-09731-9
Otterbring, T., & Folwarczny, M. (2024). 'From WEIRD to worldwide: Auditing authors' affiliation countries over time across three leading journals publishing food-related research.' Food Quality and Preference, 117, 105175. doi:10.1016/j.foodqual.2024.105175
Stewart, N., Ungemach, C., Harris, A. J. L., Bartels, D. M., Newell, B. R., Paolacci, G., & Chandler, J. (2015). 'The average laboratory samples a population of 7,300 Amazon Mechanical Turk workers.' Judgment and Decision Making, 10(5), 479-491
Klein, R. A., and 185 colleagues (2018). 'Many Labs 2: Investigating Variation in Replicability Across Samples and Settings.' Advances in Methods and Practices in Psychological Science, 1(4), 443-490. doi:10.1177/2515245918810225
Schimmelpfennig, R., Spicer, R., White, C. J. M., Gervais, W. M., Norenzayan, A., Heine, S., Henrich, J., & Muthukrishna, M. (2024). 'The Moderating Role of Culture in the Generalizability of Psychological Phenomena.' Advances in Methods and Practices in Psychological Science, 7(1), 1-6. doi:10.1177/25152459231225163
White, C., & Muthukrishna, M. (2023). 'More educated, more WEIRD: Higher education predicts worldwide cultural similarity to the United States and other WEIRD countries.' PsyArXiv. doi:10.31234/osf.io/8wr5d