In 2010, Joseph Henrich, Steven Heine, and Ara Norenzayan showed that 96 percent of the data that scientists had relied on to understand human psychology and behavior was from W.E.I.R.D. research participants—people living in Western, educated, industrialized, rich, and democratic countries. That biased sample, they argued, severely limited what we could say we really knew about ourselves. Their work was a wake-up call to many scientists to expand where and how they conduct their research.
But recognizing the need to expand and diversify research is one thing, and actually doing it is another. A decade later, Henrich, Norenzayan, and Coren Apicella conducted a follow-up study to see if participant samples had gotten more diverse. They hadn’t. The research team found that 94 percent of studies were still conducted with W.E.I.R.D. participants.
The W.E.I.R.D problem persists not because scholars fail to see it as a problem worth solving but because conducting research in new, unfamiliar places is difficult. There are language and cultural barriers, obstacles to finding participants, and complications with technology. But overcoming these difficulties is doable, and it’s essential.
The W.E.I.R.D problem persists not because scholars fail to see it as a problem worth solving but because conducting research in new, unfamiliar places is difficult.
In 2023, I founded Besample, a platform designed to help scientists reach participants all over the globe, in order to help solve these challenges. As a social psychologist trained in the United States, but born and raised in Russia, helping researchers address the W.E.I.R.D. problem feels both personal and professional. And in Besample’s first three years, we’ve learned that many scientists, when given the opportunity, are ready to address the W.E.I.R.D. problem too, by expanding where they conduct their research.
So far, we’ve worked with over 1,700 scholars on more than 880 studies, collecting nearly 330,000 data points from participants in 42 countries across Asia, Africa, Eastern Europe, and South America. Data collected through Besample has covered a range of phenomena and is now making its way into the academic literature, including cross-cultural research on loneliness and investigations into value systems, perceptions of climate change, and family dynamics.
Through this work, we’ve learned quite a bit about how scientists can conduct high-quality, culturally grounded research beyond the West. Here are seven key lessons we would like to share.
1. The world does not speak English
Researchers from North America and Western Europe often default to running studies in English, even when targeting non-English-speaking countries in the Global South. While English may work in former British colonies, like India, South Africa, or Kenya, it falls short elsewhere, especially in Asia and Latin America.
This is a problem, because people from non-English speaking countries who read and write in English are special, not representative. Most likely, they’ve been economically privileged enough to learn the language in good schools or colleges.
Another problem with running studies in English is that even highly proficient speakers can experience the foreign-language effect, where they feel more emotionally detached from the material simply because it’s not in their native language. People find it easier and more natural to complete tasks in their native language, which can influence how they respond.
For some, this is a matter of dignity. In Turkey, for instance, some of Besample’s respondents declined to participate in studies conducted in English, because the lack of a Turkish translation felt disrespectful. Kazakhstan presented an even greater challenge; respondents dropped out at the consent stage if it was presented in dense legal English—they treat formal agreements very seriously, and the inability to fully understand the text was perceived as a risk.
At Besample, we recommend translating studies into local languages for countries where the average English proficiency is below 50 percent, which helps boost sample size and representativeness. Today, it’s easy to generate a first draft of translations with AI, and local experts can help proofread and ensure that materials are appropriate given the cultural context.
2. Standard attention checks create more noise than signal
As the popularity of online studies grew in the scientific community, researchers became more concerned with whether participants were actually paying attention. So they developed attention checks—questions to assess whether a participant was reading the materials carefully—to filter out low-quality responses.
In studies with non-Western populations, researchers frequently exclude those who fail attention checks, assuming this reflects a participant’s lack of focus and indicates low-quality data. However, in global research, such failures are often a matter of a language barrier or less familiarity with participating in studies rather than a lack of engagement or incoherent data.
For example, in an analysis I coauthored on the data quality of online research platforms, we found that Besample respondents with lower English proficiency could fail attention checks but provide high-quality data to other questions. These findings suggest that attention-check performance is not a definitive proxy for data quality.
Before making the easy call to dismiss responses, scientists should view attention-check failures through a linguistic lens—or, even better, design localized quality checks based on consistency and accuracy prior to launching a study in a new context.
3. Questions about race and ethnicity are not universal
Another trap researchers can fall into is using Western-centric templates for questions on race and ethnicity. Though it’s common to think about one’s identity in terms of race in places like the United States or United Kingdom, in other countries it’s not. And when it comes to ethnicity, there are over 800 distinct ethnicities (or over 7,000 if you use ethnolinguistic groups as a taxonomy).
To better understand the diversity of our participants, we surveyed nearly 1,000 people from 11 countries: Brazil, Colombia, Indonesia, Japan, Mexico, Morocco, Nigeria, Spain, Ukraine, the United Kingdom, and the United States. We asked respondents about their self-identification regarding race and ethnicity, providing both open-ended and multiple-choice questions. Open-ended responses showed that many participants described themselves by ethnic group or geographic region rather than conventional racial classifications.
For the multiple-choice questions, people in countries such as Ukraine, Morocco, and Indonesia selected commonly used categories 50 percent less often and were more inclined to select a category when presented with options from an extended ethnic scale, because it included ethnic identifiers that standard questions leave out.
When launching a study in a new, culturally distinct place, failing to adapt questions around race and ethnicity can lead to meaningless or even misleading data.
4. Simple questions about gender aren’t so simple
Cultural nuances can also alter the meaning of even the most basic-seeming gender identity terms. A couple of years ago, when we were building an early pool of respondents, we noticed a puzzling pattern in Indonesia: In a demographic pre-screener, an unexpected number of people were selecting “other” when asked whether they identified as pria (male), perempuan (female), or lainnya (other). We were curious to learn why.
We knew that Indonesia was among a few world societies where, in certain provinces, more than two gender categories were accepted. But most of our respondents were not from those provinces. When we added an open text field to the “other” option, we started to see an influx of typed-in responses, all similar to laki-laki. Young Indonesian men were just not calling themselves pria because that was a term for older, adult men. For younger males, there was a different word: laki-laki. While distinctly identifying as male, young men were selecting “other” to type in this exact word.
Another example is hijra, a gender category found in India, Pakistan, Bangladesh, and Nepal. Hijra are neither “men” nor “women,” nor simply “transgender” in the Western sense. They are often recognized as a third gender, with long-standing social and ritual roles that stretch back hundreds of years. When people identify as nonbinary around the world, it often means something very different from what we’re used to in the Western context.
Asking about gender also requires more than just understanding the correct terminology. In many places, certain gender identities and sexual orientations face prejudice or are even punishable offences, so researchers need to be aware of this before launching their studies.
5. Standard questions about marital status are not sufficient
Marital status is deeply intertwined with family makeup, household decision-making, economic life, and a host of other things that social and behavioral scientists are interested in. But in some cultural contexts, asking about marital status is far more complicated than it might seem.
For instance, in South Africa, researchers may need to distinguish between a civil ceremony and a traditional union formalized through lobola, a practice involving the negotiated transfer of property (in cash or in kind) from the prospective husband or his family to the bride’s family.
Additionally, in the West, marital status is often interpreted as monogamy. But in countries such as Senegal, Guinea, Liberia, Mali, Burkina Faso, Benin, and Togo, polygynous unions account for 40–60 percent of reported marriages.
Turning to divorce—another key variable in studies of close relationships—Western researchers often assume that people are free to exit marriage through divorce. But in many countries, divorce is legally restricted, socially stigmatized, or practically inaccessible, making it a poor proxy for relationship quality or stability. A striking example is the Philippines, where divorce is not legally available for most citizens. Instead, marital dissolution occurs through Catholic annulment, a costly and rare process. As a result, less than two percent of the population has divorced, separated, or had a marriage annulled.
6. Households and household income look differently around the world
It is common for researchers in the United States to assess participants’ economic standing by asking about their household income. But the idea of the household—a cornerstone unit in Western economics and connected with income, consumption, and decision-making—is frequently different in non-Western cultures.
For example, in rural Malawi, households often diverge from the idealized Western model. Families can be spatially dispersed, household membership fluid, and major decisions follow matrilineal or patrilineal kinship lines. So in Malawi, kin social networks and resource flows often capture economic reality better than the Western household template.
Additionally, many people assess their income in monthly terms and in immediate purchasing power, rather than in abstract annual household income terms (and especially not in foreign currencies like dollars or euros).
Another financial difference worth pointing out is that many researchers assume they can pay online participants using bank transfers or PayPal. However, the ways people bank and handle everyday purchases is extremely variable. In many countries, cash and mobile transfers are the default rather than bank cards, and PayPal is virtually unknown. Bureaucratic hurdles can also make it highly inconvenient for people to receive a small $1–$3 reward for participating in a study via bank transfer. In some cases, this even requires visiting a bank in person to justify the transfer. This reality led us to build a network of more than 20 country- and region-specific partners, offering more than 45 different payment options to reward research participants.
7. Conventional technical criteria do not indicate data quality
Traditionally, certain technical criteria, such as a unique IP address, served as a proxy for data quality. If many people logged in from the same IP address, researchers assumed that the same participant was taking the survey multiple times, and the data should be thrown out. But this isn’t necessarily the case.
In regions with limited IP availability, such as parts of Africa and Asia, it’s common for multiple users to share the same IP address. Many mobile operators assign a single public IP to hundreds—or even thousands—of devices, especially in mobile broadband networks where dynamic IP allocation manages connectivity for large user bases.
Researchers have also traditionally assumed that respondents using a desktop or laptop computer will provide higher quality data, and have often optimized their studies for PCs. But in many non-Western countries access to PCs are limited—in some countries less than 10 percent of people have access to a computer or laptop. Limiting studies to PCs can create a barrier to data collection and bias samples toward higher-income, urban participants with better digital access.
But even if a survey is designed to work on a mobile device, people may lack reliable internet or have limited mobile data. These factors make it difficult for respondents to complete long surveys, especially those with large files or videos.
What appears to be “low-quality” data based on a technical criterion may actually reflect local infrastructural and socioeconomic realities. Relying on technical checks like PC-only access or unique IP addresses leads to biased samples, slower data collection, and a failure to capture the true diversity of non-Western regions.
* * *
These lessons detail some of the practical steps that researchers can take to move beyond W.E.I.R.D. populations. Taken together, these lessons also point to a broader shift that is needed if we want to make the social and behavioral sciences truly global: We need to have the courage to admit that things will likely go differently than we expect. And as we venture out to study the world, intellectual humility should be our starting point.
