Who Was Being Measured — Who "People" Meant When Psychology Said It

The Hidden Subject of Textbook Sentences

Conclusions in psychology papers are usually written like this: "People feel losses more deeply than gains." "People explain others' actions by personality rather than situation." The subject is "people." But the people actually sitting behind those sentences were almost all college students. The people who actually participated in an experiment and were measured are called the sample, and the process of claiming that the results obtained from the sample also apply to a broader group of people outside the sample is called generalization. Sentences like "People do ~" are sentences generalized from the sample of college students to all of humanity.

In 2010, a review gave a name to this practice: WEIRD, which is an acronym for Western, Educated, Industrialized, Rich, and Democratic societies. The critique had two layers. First, behavioral science experimental results vary considerably across population groups. Second, in the distribution by group, WEIRD samples were frequently outliers, meaning values that are far apart from the rest. The areas examined were not peripheral — visual perception, fairness, cooperation, spatial reasoning, categorization and inductive reasoning, moral reasoning, reasoning styles, self-concept, and the heritability of IQ. The conclusion was a restrained sentence. There is no a priori ground—meaning no clear basis to put forward in advance—to claim that a certain behavioral phenomenon is universal based solely on data drawn from a single subgroup (PMID 20550733, DOI: 10.1017/s0140525x0999152x). A commentary in the same issue added that brain imaging research shares the same bias (PMID 20546651).

One thing must be nailed down in advance. The point of this critique is not that any society is better or worse. It is that the standard was set incorrectly. It is a matter of sampling (who you bring into a study) and measurement (what and how you measure from those people), meaning that only a thin slice was measured and described in the name of the whole.

The Basis for the Critique

The most frequently cited evidence was economic games. There is a study that ran the ultimatum, public goods, and dictator games—which observe whether people pursue only their own self-interest through ways of dividing money—across 15 small-scale societies. It was a study trying to answer questions that could not be addressed by existing experiments targeting only college students (PMID 16372952). The standard model based on self-interest failed in every society studied, and the behavioral variation between societies was much larger than in existing research. The greater the degree of market integration and the benefit that cooperation provides in daily life, the higher the pro-sociality in the experiments.

Basic perception did not seem to be an exception either. In a change blindness task of noticing when one part of a screen changed, American participants were sensitive to changes in the focal object, while East Asian participants were relatively sensitive to changes in the context (PMID 21702819). The Tsimane' of the Amazon, who have almost no exposure to Western music, evaluated consonant chords, which sound harmonious, and dissonant chords, which sound clashing, as equally pleasant. Urban residents in the same country preferred consonant chords, and American respondents preferred them more strongly (PMID 27409816).

The self and morality domains were also shaken. When re-examining whether the desire to view oneself positively exists in everyone, this desire could be rooted in North American culture (PMID 10560328). Even when a behavior was done due to social pressure and thus told little about the person's attitude, the correspondence bias—assuming one held an attitude matching the behavior—was maintained only in Americans, unlike the Japanese (PMID 12416925). Rural Maya participants saw harm as a means to a greater benefit as more prohibited than harm as a side effect, yet they did not morally distinguish between causing harm through action and harm occurring through inaction (commission and omission) (PMID 20092817).

Counterattack — A Re-examination of a Representative Case

However, ending the story here would not be honest. Regarding the Müller-Lyer illusion, cited as the most famous example of cultural influence in the perceptual domain, a recent comprehensive review directly argued against it. This illusion is an image where two line segments of the same length look different in length when arrowheads are attached pointing inward and outward. A research lineage spanning over a century claimed that depending on what society one grew up in and how much one was exposed to a carpentry environment built with right angles, this illusion might not be seen at all. From this came the proposition that "even fundamentally appearing visual processing is a culturally evolved byproduct."

This review argued that this hypothesis was almost certainly wrong in 2025. The illusion also appears in non-human animals living in different ecologies, and is replicated by the statistics of natural scenes alone. It arises even without the straight lines specific to a carpentry environment, and even without a single line. It also appears in senses other than vision, and in people with congenital blindness. Furthermore, the authors reopened the historical data and ethnographic descriptions (field observation records) of the original case study, reporting that the evidence for cultural differences and the correlation those differences had with core cultural variables were actually very inconsistent. That evidence suffered from questionable research practices and was miscommunicated in later discussions (PMID 40388159). However, this rebuttal is an argument made by a single review, and its target is also an illusion case in visual perception, not evidence from other domains like economic games.

Evidence for universality has also accumulated in the meantime. Pride measured in 567 people across 10 small-scale societies followed the value evaluations of local audiences despite large differences in language and subsistence modes. The correlation coefficient $r$, which measures the degree to which two values rise and fall together, was an average of $r=+0.66$ within the same region and $+0.29$ even across regions (PMID 30068602). Conversational turn-taking showed the same pattern of avoiding overlapping speech and minimizing silence across 10 languages, with differences by language falling within a range of 250 milliseconds on average (PMID 19553212). The prediction that shame is a defense against reputation damage was supported in 6 countries (PMID 41871240). The prediction that retributive sentiments to punish wrongdoers follow victim compensation and recurrence deterrence together was confirmed similarly in 74 Mentawai horticulturists in Indonesia and 600 Americans (PMID 42113676).

What is especially noteworthy is that even within a single study, some layers are the same across societies while other layers differ. The property of valuing rewards differently depending on surrounding options was similar across all 11 countries, but risk aversion measured by a separate task differed significantly by country (PMID 36909645). Although Seoul and Los Angeles samples diverged greatly on whether the causes of wrongdoing were attributed to situations or dispositions, the pattern of judging violations that are distant in time and space or backed by authority approval as less bad was the same in both (PMID 35987768).

Conversely, emotional perception, meaning recognizing emotions from facial expressions, remains divided. A study of Hadza hunter-gatherers found almost no evidence of universal emotional perception (PMID 32123191). There is an argument that evidence for universality depends on experimental methods (PMID 28950961), and a counter-rebuttal went back and forth that the manipulation check procedure of the experiment favoring universality may have instead taught Western emotion categories to the participants (PMID 25608863).

So the Takeaway Is "You Can't Predict It"

The conclusion to draw from this balance sheet is neither that "the West is bizarre" nor that "it's all the same in the end." The observation that WEIRD samples were frequently outliers has not been retracted, but rather that while some phenomena cross cultures and others do not, you cannot know which is which before measuring it. A recent one-line summary from developmental psychology captures this attitude. Looking at replication attempts involving multiple labs, cultural differences are certainly important in some cases, but they are often smaller and less systematic than assumed, and therefore cultural generalizability is a matter to be empirically tested rather than presupposed (PMID 42396701). The problem is not that any particular sample is inferior, but that you cannot know the answer in advance from any single sample.

15 years later, what has actually changed?

Interest in the critique has grown significantly. The results of recounting by field are as follows. Among papers in representative cognitive psychology journals from 2016 to 2020, about 7% considered culture—even when defined very broadly—and of those, 83% were language and bilingualism studies (PMID 36510095). Across 2,223 articles in 5 sex research journals, the proportion of WEIRD samples was 68 to 88% (PMID 33939579). In an analysis coding 1,682 papers and about 1 million participants from 4 infant development journals between 2011 and 2022, sociodemographic information was consistently not reported, and when it was reported, it was unwaveringly skewed toward white infants in North America and Western Europe (PMID 37211974). Comparing two time points 20 years apart across 1,762 studies in 8 relationship science journals dealing with bonds like romantic partners and spouses, some items like socioeconomic status did not improve, and even the items that did improve (sexual orientation) had limited reporting (PMID 36757951). The reporting rate for race and ethnicity in psychophysiology journals from 2010 to 2020 was 17% (PMID 37332087).

The most stinging figures come from an analysis of alcohol research journals from 2010 to 2022. Reporting rates for demographic items rose overall, and rose significantly in race, sexual orientation, gender identity, and socioeconomic status. Yet over the same period, the proportion of studies conducted outside the United States decreased significantly (PMID 38653579). Improving reporting and broadening samples are different matters. The results can also look different depending on the sample region. In one review, out of 34 papers from 2018 to 2020 dealing with adolescents' social media use and depression, over 70% used Global North samples, and less than 30% used Global South samples. The association between the two was significantly positive in the Global North, while being close to zero and nonsignificant in the Global South (PMID 37694229).

There have also been attempts to recount the whole picture from a more recent viewpoint. This is the result of counting 45 papers and 133 studies from the June–December 2023 issues of Psychological Science, as a preprint, which is a manuscript before review by researchers in the same field (peer review). Samples were still 85% WEIRD, and culturally relevant characteristics were reported in only 54% of abstracts. Overgeneralization beyond the sample, meaning stating conclusions about groups wider than the people measured, occurred in 97% of abstracts (preprint before peer review) (DOI: 10.31234/osf.io/mkpce_v1). The same preprint notes that WEIRD countries account for about 12% of the world's population.

Why doesn't it change? One commentary answers coldly. The reason it has not moved even after being argued for over 50 years is not a lack of knowledge, arguments, or clever acronyms, but rather lies in the discipline's self-image and priorities of being a universal, decontextualized, and generalizable science (DOI: 10.1017/s0140525x25104172).

Just Adding Diversity First — Who Do We Add

There was optimism in 2010 as well. A commentary analyzing 564,502 online personality test respondents suggested the potential of the internet by noting that 19% were from outside advanced economies and 20% were from non-Western societies (DOI: 10.1017/s0140525x10000300). Yet subsequent research points out that as researchers gather many highly educated individuals in samples like college students or online workers, high education is not a neutral axis. A preprint measuring the cultural distance between high- and low-education groups using World Values Survey (2005–2022) data from 268,992 respondents across 95 countries reported that the high-education group is significantly more similar to WEIRD countries in beliefs, values, and behavior. This pattern did not appear as similarity with other large and influential countries like China, Russia, or India, and neither income nor subjective status produced the same pattern. The authors suggest that "international" samples made of college students and online workers in each country may overrepresent values typical of WEIRD countries (preprint, not peer-reviewed) (DOI: 10.31234/osf.io/8wr5d_v2). A 2010 commentary wrote that sample diversification alone cannot reach generalizable principles, and we must simultaneously broaden the context of data collection, meaning the range of situations in which experiments take place (DOI: 10.1017/s0140525x10000063).

Some shortcuts are also proposed. These are synthetic respondents made by having large language models answer surveys like humans. In a study comparing the United States, Saudi Arabia, and the United Arab Emirates, GPT-4's synthetic responses were plausibly similar to humans on average, but correlations tended to be weaker in Saudi Arabia and the United Arab Emirates (DOI: 10.1177/23794607241311793). There are also more fundamental criticisms. Because the training corpus is skewed toward the West, English, and third-person narration, synthetic responses may reproduce the way others described them rather than how people express their own experiences. It is also argued that replacing people with synthetic responses could weaken incentives to invest in infrastructure like translation, local expertise, and community collaboration (preprint, not peer-reviewed) (DOI: 10.31234/osf.io/8x37h_v1).

Does Just Increasing Diversity Suffice ② — The Tools and Theories Are Already Aligned

Once you widen the sample, the next wall appears immediately. The tools themselves have been polished to fit specific participant groups. Records from a research team attempting to run a social discounting task (a task measuring how the share I give away shrinks according to social distance from the other person) in rural Bangladesh demonstrate this concretely. Through multiple pilot tests and revisions, the implicit premises that came to light were familiarity with numbers and letters, the ability to read abstract vertical lines and two-dimensional shapes, and an attitude of treating a series of choices as mutually isolated events. Through field observation, the team built a new protocol that functions for both Bangladeshi and American college students (PMID 30397138).

Theories are the same. An audit revealed that two representative theories of developmental psychology have the same bias. Shared intentionality theory, which deals with humans' ability to share intentions with others, has a biased foundational database, weakening the claim that it explains human social cognition in general. Attachment theory, which deals with the emotional bond between a child and a caregiver, excludes community-oriented constructs observed in various societies, such as interdependence or attachment networks, from its core assumptions (PMID 40407757). There is also the criticism that the category WEIRD itself is a lump sum — because each item is not properly operationalized, it erases people of diverse identities living within the WEIRD context (DOI: 10.1146/annurev-anthro-102218-011133).

For balance, claims in the opposite direction are also introduced. One paper points out that sample diversity recommendations are sometimes applied even to research that does not aim to generalize across people in the first place. In addition, it states that WEIRD samples and convenience samples are not the same thing, and that convenience samples—composed of people researchers can easily gather—are rather an important means for globalizing psychology (PMID 39509224).

A Harder Problem — Does Translation Measure the Same Thing?

A scale is a questionnaire that turns invisible mental traits, like mood or depression, into scores across multiple items. When you translate the same scale and run it in different countries, you naturally want to place the scores side by side and compare them. However, in a study that ran a scale asking about positive and negative experiences across 12,635 people in 13 countries, three items—"sad," "fearful," and "angry"—did not operate by the same yardstick across countries. Excluding those three items, a state where the baseline of scores between groups is the same—in other words, partial scalar invariance—was supported. This means caution is necessary when comparing countries using these three items. The abstract does not clarify whether the source of this difference is translation or a difference in cultural meaning (PMID 34105378).

To say "this country is higher," a scale must be measuring the same thing in the same way in both groups. This condition is called measurement invariance (also known as measurement equivalence).

The verification procedure is well-organized. You confirm through qualitative research whether that concept has meaning in that culture, translate and adapt it, and then go through a pilot test and a cognitive interview (an interview where participants are asked directly how they understand the items). After that, you explicitly test the invariance hypothesis using statistical methods like confirmatory factor analysis and the Rasch model. Yet this review leaves an uncomfortable phrase. Even if invariance holds, threshold effects arising from cultural differences in norms, expectations, and expressions of distress—in other words, the effect where the baseline for answering "yes" to the same degree of pain varies by culture—can remain. This 2008 narrative review evaluated that cases actually proving measurement-fair comparisons in the mental health literature are rare (PMID 18924560).

Actual results match that warning. A study that validated the short form of the Moral Foundations Questionnaire, which measures five foundations of moral judgment across 27 countries on 5 continents, reported that it is difficult simply to craft items so that the five-factor model—the structure where the five foundations divide into five clusters—fits without problems across a wide range of populations (PMID 30596252).

The clearest case is the psychotic experiences scale. A study analyzing 7,141 people across 13 countries using this scale, which self-reports symptoms like hallucinations, lack of motivation, and depressive symptoms, examined whether previous results showing psychotic experiences are more common in low- and middle-income countries were due to measurement bias. Partial scalar invariance held, making it possible to compare latent scores, which estimate the common trait behind the items using a statistical model. However, residual invariance—where even the size of the item errors is the same between groups—did not hold, leading to the report that comparing sum scores simply adding up item scores is biased. Comparing before and after correction, one factor was overestimated and another was underestimated. The difference between low- and middle-income countries compared with high-income countries (effect size $d$, meaning the difference between the two groups' averages divided by the degree of scatter) changed from $d=0.03$ to $d=−0.42$ for amotivation, and from $d=−0.03$ to $d=0.33$ for magical thinking (PMID 33023691). This means previous studies comparing countries using sum scores are likely also biased. At the same time, the same study states that even after correction, low- and middle-income countries were significantly higher in most factors (only amotivation was significantly lower), and the basic finding that psychotic experiences are more common in low- and middle-income countries was confirmed—bias shook the size and direction of individual factors, but did not overturn the big picture.

Not all validations end in failure. A study examining 15 items of a depression scale in 4,732 people across 8 low- and middle-income countries using item response theory—an analysis that looks at how well each item screens out that hidden trait—reported that 13 out of 15 items function well in various environments. The two items with low discriminative power were suicidal ideation and "loss of sexual interest or pleasure." While differential item functioning, where item scores vary by country despite the same depression level, existed to some degree in most items, the only place it affected the total score level was one country (Indonesia) (PMID 27083900). The same problem occurs when the measuring tool is an algorithm instead of a person. Two automated facial expression recognition systems showed high accuracy in Western datasets but dropped in performance in Eastern datasets, particularly for negative emotions, thereby violating measurement invariance. Happiness measurement was accurate and invariant (PMID 38130968).

Problems remain even though standards for translation practice already exist. One paper suggests adding one more cause to the list. This issue has long been blamed on a lack of knowledge about translation procedures, but even after solutions based on that diagnosis were widely adopted, the problem was not solved. The authors consider that a lack of information about the scale being translated itself may be an additional cause. If so, since this is not a problem translators can solve on their own, they propose exploring a plan where original scale developers create translation guides for each scale (PMID 36355577).

What the Scale Simply Does Not Ask

Even if the translation is perfect, a layer remains. These are the things that the scale does not ask about in the first place.

When low-income adolescents in Kenya were asked directly about their "biggest problem" without fixed choices, 61% cited social problems, 38% academic problems, and 35% economic problems. However, only 17% of participants cited emotional and behavioral problems—measured by standardized mental health scales—as their top problem. This means there are problems that standard scales can miss (PMID 33528725). An exploratory factor analysis in 116 people suggested that a locally co-developed depression questionnaire for the Luo people of western Kenya had a three-factor structure of cognitive, emotional, and physical—meaning a structure where the items split into three bundles—whereas the standard tool was a one-factor structure. The local tool captured features that the standard tool missed (PMID 33818199).

"Thinking too much" is an idiom for expressing distress in various societies. A systematic review synthesizing 138 papers from 1979 to 2014 reported that this expression variably overlaps with constructs (concepts that cannot be seen directly and are measured with scales) like depression, anxiety, and PTSD, while capturing aspects that diagnoses fail to catch, and showing large variation not only between cultures but also within cultures (PMID 26584235). A meta-analysis (an analysis that combines and recalculates the results of multiple studies) of 45 papers and 18,782 people comparing cultural distress concepts and psychiatric diagnoses yields a more uncomfortable result. Only 27% of the studies met moderate-or-higher quality criteria, and the higher the quality of the study, the weaker the association between the two concepts. The authors conclude that cultural distress concepts are not inherently unsuited for epidemiological research, but rather that research quality is the stumbling block (PMID 24366490).

Proposals and Attempts

Leaving only pessimism is not accurate. However, the following are mostly designs and proposals, so their effectiveness is not confirmed by this material. Permanent and distributed multi-lab networks like the Psychological Science Accelerator (PSA) are designed to conduct a single study together across multiple countries (PMID 31886452), and scales have also emerged such as a survey where 404 researchers gathered romance and mate preference data from 117,293 participants across 47 languages and 175 countries (PMID 40593913). There is also a proposal to let theory drive sampling instead of expanding blindly — arguing that socioecological hypotheses, such as how environmental harshness shapes time preference or how market integration shapes personality structure, specify which groups need to be measured (PMID 30397108). A framework that breaks down external validity — the degree to which results hold up in different situations — into four domains of populations, situations, outcomes, and times to inspect them is in the same direction (PMID 31743074), and a paper organizing the difficulties of multi-region research recommends an approach that puts community wishes first, noting that such considerations are part of scientific rigor (PMID 32962541). And there is the assessment that reproducibility, the property where getting the same results when trying again, and generalizability are not problems to be solved separately but entangled problems that must be solved together (PMID 36104999). This is where the replication crisis covered in "Psychology's Replication Crisis — Why the Findings We Believed In Wobbled" and the sampling problem of this primer meet.

Honest Unsolved

To sum up, fewer things are settled in this debate than you might think.

What is universal can still only be answered on a case-by-case basis. The structure of pride and the alternation of turns in conversation crossed over, the preference for consonant chords and the focus of change detection did not cross over, and emotion perception is still being fought over. No theory guessed this list in advance in the data for this primer, and the proposal to let theories drive sampling (PMID 30397108) is still at the proposal stage. However, universality and variation can also be split into layers within a single study. Children's peer cooperation preference differed greatly across three groups, but the relationship between positive emotion and social-cognitive ability during cooperation was consistent across groups (PMID 33211711). The three-factor structure of advanced theory of mind—the ability to read others' minds as measured by social reasoning and norm recognition—was common in Japanese and German children, but the developmental pathway was different (PMID 41983731). The real question might not be "what," but "which layer" is universal.

When invariance testing fails, the choices diverge. Whether to rely on partial invariance to compare group averages only at the level of hidden traits, drop problematic items and change the scale, or give up on comparison altogether—which side is the standard cannot be known from the data in this primer.

There is no guarantee that diversification leads to representativeness. This paper also points out that it is difficult to distinguish what counts as "sufficient" generalizability (PMID 39509224). The boundary between genetics and culture is also not fixed—when cumulative culture functionally overlaps with genes, genetic effects can be masked, revealed, or reversed, and that pattern is unique to a specific time and society, with some analyses suggesting that the WEIRD sampling problem is precisely what hides this condition dependence (PMID 34016199).

Responding to 28 commentaries, the authors who raised the original point also wrote that a broad-based behavioral science must apply diverse methods, including experiments and ethnography, to diverse populations (DOI: 10.1017/s0140525x10000725). Fifteen years later, the remaining question is not very different from the beginning. When we write "People do ~," who were we measuring? And when we say we are measuring more people, are we really measuring the same thing?

Source Papers

• "The weirdest people in the world?" (2010) — PMID 20550733 / DOI: 10.1017/s0140525x0999152x — The original paper on the WEIRD concept. Reviews high variation across populations, with WEIRD samples frequently acting as outliers, examining visual perception, fairness, cooperation, spatial reasoning, categorization, moral reasoning, reasoning styles, self-concept, and IQ heritability.
• "Beyond WEIRD: Towards a broad-based behavioral science" (2010) — DOI: 10.1017/s0140525x10000725 — The authors' response to 28 commentaries. Emphasizes the integration of experiments and ethnography.
• "The weirdest brains in the world" (2010) — PMID 20546651 — A commentary arguing that brain science shares the same sampling bias.
• "'Economic man' in cross-cultural perspective: behavioral experiments in 15 small-scale societies" (2005) — PMID 16372952 — The self-interest model fails in all 15 societies, with prosociality increasing as market integration and the gains from cooperation grow.
• "Culture and change blindness" (2006) — PMID 21702819 — Americans are relatively sensitive to focal objects, while East Asians are relatively sensitive to contextual changes.
• "Indifference to dissonance in native Amazonians reveals cultural variation in music perception" (2016) — PMID 27409816 — The Tsimane' evaluate consonance and dissonance equally. Preference for consonance increases in the order of urban Bolivia and the United States.
• "Mayan morality: an exploration of permissible harms" (2010) — PMID 20092817 — The distinction between means and side effects is shared, but the distinction between acts and omissions is not observed among rural Mayan participants.
• "Is there a universal need for positive self-regard?" (1999) — PMID 10560328 — A re-examination suggesting that the need for positive self-regard may be rooted in North American culture.
• "Cultural variation in correspondence bias" (2002) — PMID 12416925 — Only Americans maintain correspondence bias when the attitude diagnosticity of behavior is low.
• "Is visual perception WEIRD? The Müller-Lyer illusion and the cultural byproduct hypothesis" (2025) — PMID 40388159 — A comprehensive rebuttal to the cultural byproduct hypothesis. Evidence from non-human animals, natural scene statistics, line-free stimuli, other senses, and congenital blindness, plus a re-examination of raw data.
• "Invariances in the architecture of pride across small-scale societies" (2018) — PMID 30068602 — 567 participants across 10 small-scale societies. Pride tracks local audience valuation at r=+0.66 (within-region) and +0.29 (between-region).
• "Universals and cultural variation in turn-taking in conversation" (2009) — PMID 19553212 — Avoidance of overlap and minimization of silence are common across 10 languages, with differences being quantitative variations within a 250ms range.
• "Outcome context-dependence is not WEIRD" (2023) — PMID 36909645 — Context-dependent reward encoding is identical across all 11 countries. In contrast, risk aversion differs significantly by country.
• "Cross-cultural evidence that shame is a defense against reputational damage" (2026) — PMID 41871240 — Both predictions are supported in 6 countries and two regions of the United States.
• "Moral parochialism and causal appraisal of transgressive harm in Seoul and Los Angeles" (2022) — PMID 35987768 — Situational and dispositional attributions differed greatly, but patterns of moral parochialism were identical in both places.
• "Retributive Sentiments Track Both Deterrent and Compensatory Concerns in a Small-Scale Society and a WEIRD Sample" (2026) — PMID 42113676 — 74 Mentawai participants and 600 Americans, four experiments.
• "Emotion Perception in Hadza Hunter-Gatherers" (2020) — PMID 32123191 — Found almost no evidence for universal emotion perception.
• "Revisiting diversity: cultural variation reveals the constructed nature of emotion perception" (2017) — PMID 28950961 — The method-dependence of universality evidence.
• "Cultural variation in emotion perception is real" (2015) — PMID 25608863 — Raises the possibility that manipulation checks taught Western emotion categories.
• "Cross-cultural generalizability is an empirical question" (2026) — PMID 42396701 — Cultural differences are often smaller and less systematic than assumed, making generalizability an object of testing rather than an a priori assumption.
• "Consideration of culture in cognition" (2023) — PMID 36510095 — Only about 7% of papers in cognitive psychology journals from 2016 to 2020 considered culture, and 83% of those focused on language and bilingualism.
• "How WEIRD and Androcentric Is Sex Research?" (2022) — PMID 33939579 — 2,223 papers across 5 journals (2015–2019), with 68% to 88% WEIRD samples.
• "Diversity and representation in infant research" (2023) — PMID 37211974 — 1,682 papers and roughly 1,000,000 participants across 4 journals (2011–2022). Persistent underreporting plus a heavy bias toward white infants in North America and Western Europe.
• "'Mostly White, heterosexual couples'" (2023) — PMID 36757951 — 1,762 studies across 8 journals, comparing 1996–2000 versus 2016–2020. Certain items showed no improvement.
• "Participant diversity in Psychophysiology" (2023) — PMID 37332087 — Race and ethnicity reported in 17% of papers from 2010 to 2020, with over 60% left unanalyzed after reporting.
• "Participant diversity in ACER: 2010-2022" (2024) — PMID 38653579 — Reporting rates generally increased, but the proportion of studies outside the United States decreased significantly.
• "Lack of Sample Diversity in Research on Adolescent Depression and Social Media Use" (2023) — PMID 37694229 — Over 70% of 34 papers were from the Global North. Associations were significant in the Global North and absent in the Global South.
• "Beyond WEIRD: The Persistence of Limited Samples and Overgeneralisation in Psychological Science" (2026) — DOI: 10.31234/osf.io/mkpce_v1 — WEIRD countries account for roughly 12% of the world's population. In 45 papers and 133 studies from Psychological Science in 2023, 85% of samples were WEIRD, and 97% of abstracts featured overgeneralization (preprint, not peer-reviewed).
• "Psychology wants to stay WEIRD, not go WILD" (2026) — DOI: 10.1017/s0140525x25104172 — A commentary arguing that 50 years of stagnation stems from the field's priorities and values rather than a lack of knowledge.
• "Wired but not WEIRD: The promise of the Internet in reaching more diverse samples" (2010) — DOI: 10.1017/s0140525x10000300 — Among 564,502 online participants, 19% were from outside advanced economies and 20% were non-Western.
• "More educated people are more WEIRD" (2025) — DOI: 10.31234/osf.io/8wr5d_v2 — 268,992 participants across 95 countries. Higher education predicts cultural similarity to WEIRD countries, whereas income and subjective status do not (preprint, not peer-reviewed).
• "The WEIRD are even weirder than you think: Diversifying contexts is as important as diversifying samples" (2010) — DOI: 10.1017/s0140525x10000063 — A commentary arguing that contexts must be diversified alongside samples.
• "Beyond WEIRD: Can synthetic survey participants substitute for humans in global policy research?" (2024) — DOI: 10.1177/23794607241311793 — GPT-4 synthetic responses were similar on average, but correlations weakened in non-WEIRD countries.
• "Synthetic survey participants cannot substitute for sample diversity in policy" (2026) — DOI: 10.31234/osf.io/8x37h_v1 — Criticizes demographic labeling as distinct from true representation, pointing to training corpus biases and weakened incentives for infrastructure investment (preprint, not peer-reviewed).
• "Learning from failures of protocol in cross-cultural research" (2018) — PMID 30397138 — Social discounting task in rural Bangladesh. Implicit assumptions of familiarity with numbers, letters, vertical lines, and 2D shapes, as well as isolated choices.
• "Let's go WILD: increasing inclusivity in theories of developmental psychology" (2025) — PMID 40407757 — Data biases in shared-intentionality theory and the exclusion of community-oriented constructs in attachment theory.
• "Soylent Is People, and WEIRD Is White" (2019) — DOI: 10.1146/annurev-anthro-102218-011133 — The failure to operationalize WEIRD categories erases diversity within WEIRD contexts, maintaining an unstated axis of whiteness.
• "There is nothing WEIRD about basic research: The critical role of convenience samples" (2025) — PMID 39509224 — The issue of misapplied diversity recommendations. WEIRD samples and convenience samples are distinct, and convenience samples are important for globalization.
• "Measurement validity in cross-cultural comparative research" (2008) — PMID 18924560 — Outlines procedures for testing measurement invariance. Complete criterion validation may be a mirage, and threshold effects remain even after invariance is established.
• "Testing Measurement Invariance of the Moral Foundations Questionnaire Across 27 Countries" (2020) — PMID 30596252 — It is difficult to converge a five-factor model without problems across diverse populations.
• "Measurement Invariance of the Scale of Positive and Negative Experience Across 13 Countries" (2022) — PMID 34105378 — 12,635 participants. Items for sadness, fear, and anger are culturally non-invariant.
• "Comparing psychotic experiences in low-and-middle-income-countries and high-income-countries with a focus on measurement invariance" (2022) — PMID 33023691 — 7,141 participants. Comparing sum scores is biased. After correction, amotivation shifts from d=0.03 to −0.42, and magical thinking from d=−0.03 to 0.33.
• "Depression symptoms across cultures: an IRT analysis" (2016) — PMID 27083900 — 4,732 participants across 8 countries. 13 out of 15 items function well, while suicidal ideation and sexual interest items show low discriminative power.
• "The cross-race effect in automatic facial expression recognition violates measurement invariance" (2023) — PMID 38130968 — Two automated facial expression recognition systems show reduced recognition of negative emotions on East Asian datasets, while happiness measurements remain invariant.
• "Improving Scale Equivalence by Increasing Access to Scale-Specific Information" (2023) — PMID 36355577 — A proposal to change diagnoses. Argues that original creators should produce scale-specific translation guides.
• "In Their Own Words: Using Open-Ended Assessment to Identify Culturally Relevant Concerns among Kenyan Adolescents" (2022) — PMID 33528725 — Open-ended responses identify social concerns (61%), academic concerns (38%), and economic concerns (35%), with only 17% listing standard scale items as top problems.
• "Complementing standard western measures of depression with locally co-developed instruments" (2021) — PMID 33818199 — The Luo LDQ-17 has 3 factors, whereas the PHQ-9 has 1 factor. Local tools capture features missed by standard tools.
• "'Thinking too much': A systematic review of a common idiom of distress" (2015) — PMID 26584235 — 138 papers (1979–2014). Overlaps variably with psychiatric constructs while showing large within-culture variation.
• "Cultural concepts of distress and psychiatric disorders" (2014) — PMID 24366490 — 45 papers, 18,782 participants. 27% of moderate-or-higher quality, 29% representative samples. Higher quality yields weaker associations.
• "The Psychological Science Accelerator" (2018) — PMID 31886452 — A standing, distributed, inclusive network of multiple laboratories.
• "Cross-cultural data on romantic love and mate preferences from 117,293 participants across 175 countries" (2025) — PMID 40593913 — 404 researchers, 45 languages.
• "Broadening horizons: Sample diversity and socioecological theory" (2018) — PMID 30397108 — A proposal to let theory guide sampling. Environmental harshness and market integration explain the expression of psychological traits.
• "Weighing People Rather Than Food: A Framework for Examining External Validity" (2020) — PMID 31743074 — A framework for checking external validity across four domains: population, situation, outcome, and time.
• "Navigating cross-cultural research: methodological and ethical considerations" (2020) — PMID 32962541 — Argues that community-centered approaches are part of scientific rigor.
• "Integrating open science and multiculturalism to restore trust in psychology" (2022) — PMID 36104999 — Summarizes reproducibility and generalizability as intertwined issues.
• "Cultural variation in young children's social motivation for peer collaboration" (2020) — PMID 33211711 — 240 participants across three groups. Preferences for cooperation varied widely, but the relationship between positive affect and social cognition during cooperation remained consistent.
• "Reading between the lines: Universal structure and cultural variation in advanced theory of mind" (2026) — PMID 41983731 — Japanese and German children. A three-factor structure is shared, but developmental trajectories differ.
• "Cultural evolution of genetic heritability" (2021) — PMID 34016199 — Heritability estimates vary according to cultural dynamics, and WEIRD sampling conceals this condition-dependence.