Who We Actually Studied

The two figures at the top of this page were printed in a German psychology journal in 1892 by Franz Brentano, in a paper he called Über ein optisches Paradoxon. The vertical line in Fig. 1 and the vertical line in Fig. 2 are the same length. Almost everybody reading this will see one of them as longer.

The famous claim about this figure is not that it fools you. It is that it does not fool everybody, and that whether it fools you depends on where you grew up. In 1966 Marshall Segall, Donald Campbell and Melville Herskovits published a large study that ran this figure and others past people in three European and fourteen non-European cultures, from the United States to Senegal, the Congo basin, Central Australia and the Kalahari. The result that got repeated for the next sixty years was that the European-derived samples were the most strongly fooled, and that foragers in the Kalahari were barely fooled at all.

The interpretation that grew up around this was that people raised among right angles, in rooms and streets and rectangular buildings, learn to read converging lines as depth, and the arrowheads hijack that habit. People raised without carpentry do not acquire the habit and so do not see the illusion. Seeing, on this account, is something a culture installs.

It is the single most quoted piece of evidence that human minds are not built the same way everywhere. In May 2025 two researchers published a paper in Psychological Review arguing that it is almost certainly wrong.

Both halves of that are true, and they point in opposite directions.

Who psychology has actually been asking

In 2008 Jeffrey Arnett went through the leading journals in six subfields of psychology and counted where the people in them came from. Over 70 per cent of both the samples and the authors were American. The United States holds about 5 per cent of the world's population.

Two years later Joseph Henrich, Steven Heine and Ara Norenzayan wrote the paper that named the problem. Their review found that 96 per cent of samples in the behavioural sciences came from Western industrialised countries, which between them hold about 12 per cent of humanity, and they gave those countries an acronym that stuck: Western, Educated, Industrialised, Rich and Democratic.

Their argument was sharper than the label suggests, and it is regularly misquoted. They were not simply saying that a sample of American undergraduates is a bit narrow. They were saying that across visual perception, fairness, cooperation, spatial reasoning, categorisation, moral reasoning, reasoning style and the way people describe themselves, WEIRD subjects come out as frequent outliers. Not the middle of the human range. The edge of it.

Which means the problem is not that the findings might not generalise. It is that the population being generalised from is, on the available comparisons, one of the least representative you could have picked.

What happens when you ask somebody else

The clearest demonstration is not an illusion at all. It is money.

In the ultimatum game two people split a sum. One proposes a division, the other can accept, in which case both are paid, or refuse, in which case neither is. Pure self-interest says the second player should take anything above zero and the first should therefore offer almost nothing. Run it on university students and that is not what happens: offers cluster near half, and mean offers get refused.

For years that was read as a fact about human beings. In 2005 a large team led by Henrich took the ultimatum game, the dictator game and public goods games to fifteen small-scale societies across the world, from foragers to horticulturalists to nomadic herders.

The self-interest model failed in every single society, which supported the original finding. But the variation between groups was far wider than anything the student experiments had shown, and it was not random. The more a society was integrated into markets, and the more everyday life rewarded cooperation with people outside your family, the more generous the offers were. In some places the play in the laboratory looked like a direct import of how people already dealt with each other outside it.

Individual characteristics, age, sex, wealth, explained very little. Which society you belonged to explained a great deal.

The explanation on offer

If Western populations really are outliers, something made them that way, and the most ambitious answer came in Science in 2019.

Jonathan Schulz, Duman Bahrami-Rad, Jonathan Beauchamp and Henrich proposed that a large part of it traces to the medieval Western Church. From late antiquity onwards it banned marriage between cousins, discouraged polygamy, opposed adoption of heirs and pushed people towards small, independent households. Over centuries that dismantled the dense kin networks, clans and cousin marriage that organised most human societies then and organise many now.

The authors built a kinship intensity index out of historical anthropological data, cousin marriage preferences, polygamy, extended families sharing a roof, clan organisation, marrying within your own community, and set it against how long a region had been exposed to the Western Church and against 24 measured psychological outcomes. Longer Church exposure went with weaker kinship, and weaker kinship went with more individualism, less conformity and more willingness to deal fairly with strangers.

It is an elegant argument and it should be held loosely. It is historical and correlational, it runs across enormous spans of time, and it has attracted serious pushback from anthropologists who dispute both the kinship measures and the causal direction. What it does establish is that the psychology of the people in the journals has a history, which is a different thing from being the human default.

What the field did about it

Here the record is better than cynics would predict and worse than it should be.

In 2021 Amy Thalmayer, Cecilia Toscanelli and Arnett went back to the same six journals and counted again, this time for 2014 to 2018. Arnett's paper had been cited over 1,300 times. American researchers had gone looking for collaborators abroad. The share of American authors and samples had come down from over 70 per cent to a little over 60.

The improvement came almost entirely from other English-speaking countries and Western Europe. Their summary of the position is that 11 per cent of the world's population is now represented in the top psychology journals, and 89 per cent continues to be neglected. Authors and samples from the majority of the world still make up 4 to 5 per cent.

Thirteen years of effort moved the neglected 95 per cent to the neglected 89 per cent.

And then the arrows

Which brings us back to Brentano's figure, and to Dorsa Amir and Chaz Firestone, who went and checked.

Their case has two halves. The first is that the illusion behaves nothing like a cultural habit. It shows up in non-human animals from a range of environments, none of which have been exposed to carpentry. The statistics of ordinary natural scenes are enough to produce it, no rectangular rooms required. It works without straight lines, and in fact without any lines at all. It appears in senses other than vision. And it appears in people who have been blind since birth.

The second half is more uncomfortable. They went back to the original cross-cultural data and the ethnographic descriptions behind it and found the evidence for cultural variation, and for its correlation with the cultural variables it was supposed to track, to be highly inconsistent, affected by questionable research practices, and misreported by the people who cited it afterwards. They extend the same argument to the Ebbinghaus, Ponzo, Poggendorff and horizontal-vertical illusions.

So the flagship exhibit for the claim that culture builds perception was, on this reading, never solid, and got firmer every time it was retold by somebody who had not read it.

What survives

Nearly all of it, in fact, which is the part people get wrong when they hear about this.

The sampling numbers are not an interpretation of an experiment. They are a count of where the participants came from, and they are what they are. The bargaining game results held up across fifteen societies and were not about perception. The claim that took the damage was one specific claim about one specific figure, made in the 1960s, on data that could not carry it.

What that leaves is a narrower and more useful position. Human beings almost certainly do share a great deal of basic machinery, including, it now seems, the machinery that gets fooled by two arrowheads. Where they differ, and differ enormously, is in the things that machinery gets used for: what counts as a fair split, who you owe things to, whether a stranger is a risk or an opportunity, how much you conform. Those are the parts the journals have been measuring on a tenth of humanity.

The line in Fig. 1 and the line in Fig. 2 are the same length. That is true in Baltimore and it is true in the Kalahari, and it took about sixty years and a lot of arguing to establish it properly. Almost every sentence that begins "humans are" is doing more work than the evidence behind it can support, and the way you find that out is that somebody eventually goes back and reads the original paper.

Sources

• Behavioral and Brain Sciences, Henrich, Heine and Norenzayan, "The weirdest people in the world?" (the WEIRD acronym; the finding that WEIRD subjects are frequent outliers rather than merely narrow, across visual perception, fairness, cooperation, spatial reasoning, categorisation, moral reasoning, reasoning styles and self-concept; and the argument that there are no grounds for assuming a behavioural finding is universal on the basis of one subpopulation. The figures of 96 per cent of samples and 12 per cent of world population are from the body of the same review).
• American Psychologist, Thalmayer, Toscanelli and Arnett, "The neglected 95% revisited: Is American psychology becoming less American?" (Arnett's 2008 finding of over 70 per cent of samples and authors from the United States, a country holding 5 per cent of the world's population; the fall to a little over 60 per cent for 2014 to 2018; the improvement coming mainly from other English-speaking and Western European countries; 11 per cent of the world now represented and 89 per cent still neglected; and majority world authors and samples still at 4 to 5 per cent).
• Behavioral and Brain Sciences, Henrich and others, "Economic man in cross-cultural perspective: behavioral experiments in 15 small-scale societies" (the canonical self-interest model failing in all fifteen societies; substantially more behavioural variation across groups than previous work had found; the association with market integration and with the payoffs to cooperation in everyday life; and individual-level economic and demographic variables failing to explain game behaviour).
• Science, Schulz, Bahrami-Rad, Beauchamp and Henrich, "The Church, intensive kinship, and global psychological variation" (the proposal that the Western Church's transformation of European kinship towards small nuclear households, weak family ties and residential mobility fostered individualism, less conformity and more impersonal prosociality; the kinship intensity index built from cousin marriage, polygamy, extended family co-residence, clan organisation and community endogamy; and the combination of 24 psychological outcomes with historical measures of Church exposure).
• Segall, Campbell and Herskovits, The Influence of Culture on Visual Perception (1966), as described in "Cross-Cultural Studies of Perception" (the study covering subjects from three European and fourteen non-European cultures, and the comprehension checks the authors ran before testing). The precise misperception percentages usually quoted from this work are deliberately not reproduced here, for the reason the Amir and Firestone paper gives.
• Psychological Review, Amir and Firestone, "Is visual perception WEIRD? The Müller-Lyer illusion and the cultural byproduct hypothesis" (the argument that the cultural byproduct hypothesis is almost certainly false; the illusion appearing in non-human animals, being captured by natural scene statistics, arising without straight lines or any lines, arising in other sense modalities and in congenitally blind subjects; the re-examination finding the original evidence inconsistent, beset by questionable research practices and misreported by later discussions; and the extension to the Ebbinghaus, Ponzo, Poggendorff and horizontal-vertical illusions).
• Wikimedia Commons, "Müller-Lyer illusion, Franz Brentano, 1892" (the plate reproduced at the top of this piece, from Brentano's paper Über ein optisches Paradoxon in the Zeitschrift für Psychologie und Physiologie der Sinnesorgane, volume 3, 1892).