2.1. Lack of cultural and contextual knowledge

Cross-cultural research has often relied on the ‘universality assumption’ (Kline et al., Reference Kline, Shamsudheen and Broesch2018), the belief that a lack of behavioural variation across cultural contexts necessarily means that a theory or construct is universal. Conversely, the presence of behavioural variation is considered evidence of cultural variation. However, for theories and constructs to be truly universal, their meaning must be consistent across cultures, and culture itself is the ultimate source of meaning (Pepitone & Triandis, Reference Pepitone and Triandis1987). For example, men in Fiji and Scotland both wear skirt-like garments on the lower halves of their bodies as part of their formal wear, but the superficial similarities of these outfits do not signify a shared cultural meaning or heritage. Similarity alone does not equate to cultural equivalence. On the other hand, the church and state organizations of Fiji and Scotland differ from each other along many dimensions, despite sharing a recent history of English colonial interference disrupting formerly autonomous rule. This example illustrates how over-reliance on surface similarities, without interrogating the deeper origins of cultural norms, can lead to misinterpretations of cross-cultural similarities and differences. This is exacerbated by the tendency of cross-cultural developmental science to overlook the critical insights provided by ethnography and observation, which can offer an in-depth understanding of cultural contexts and how we should study constructs.

Ethnographic approaches involve collecting in-depth qualitative data through interviews, surveys, and observations to better understand the cultural context in which behaviours develop and occur. These observations provide valuable cultural insights into children’s development. Constructs such as parenting styles, goals, or cognitive processes may hold different meanings across cultures, potentially leading to misunderstandings or misinterpretations of data if these are not examined and understood within their cultural contexts. For example, Keller et al. (Reference Keller, Lamm, Abels, Yovsi, Borke, Jensen, Papaligoura, Holub, Tomiyama, Su, Wang and Chaudhary2006) found that while certain western cultures (e.g., Germany, the USA, and Greece) emphasise independence in child-rearing, non-western cultures (e.g., rural Cameroon and rural India) prioritise interdependence and social cohesion, highlighting the importance of cultural context in understanding developmental behaviours. Without context, researchers might interpret the lack of independent behaviours in children in non-western cultures as a sign of delay or dependency, but they are in fact in line with cultural values in the population. By integrating ethnographic methods and gaining personal experience within the community of interest, researchers can better develop experimental protocols and interpret results. This approach allows for a more nuanced understanding of development across diverse cultural settings and enhances the validity of experimental measures.

Working with a community to better understand how and why children respond to experimental protocols can provide a richer understanding of constructs of interest. For example, in addition to the earlier examples of innovation and pattern recognition, the value of community input was clearly demonstrated in a cross-cultural study of mirror self-recognition (MSR) in young children (Broesch et al., Reference Broesch, Callaghan, Henrich, Murphy and Rochat2010). In the MSR task, children are surreptitiously marked on the face or body and then shown their reflection in a mirror. The underlying assumption is that if a child recognises the reflection as themselves, they will attempt to remove the mark upon discovery. The majority of children in Kenya did not pass the task in the classic sense (by removing the mark when faced with the mirror image); instead they ‘froze’ and did not respond to the protocol, which could have been misunderstood as their having ‘failed’ the task. However, through discussions with local adults, researchers identified that the children’s freezing behaviour could be understood as hesitation about removing the mark. The children were acutely aware that the experimenter had intentionally placed the mark on them and did not want to disobey an authority figure. This is a poignant example of how a result that initially did not align with expectations revealed the importance and value of community input and contextual understanding in interpreting behaviour.

There are also cases where ethnographic follow-up could have significantly enriched our understanding of children’s development. For example, in a pilot study of an economic game where children could either share or keep sweets for themselves (Kline, Reference Kline2008, field notes), researchers observed that children in a Yasawa (Fiji) village frequently collected large amounts of sweets for themselves during the game. In many contexts, this would have been interpreted as a self-serving behaviour. However, the researchers also noted that immediately after the game, the children distributed the sweets to other children, kin, and neighbouring households (Kline, Reference Kline2008, field notes). Similarly, Wiessner (Reference Wiessner2009) found that Ju/’hoan Bushmen were more selfish in a dictator game and an ultimatum game because the anonymity of the games removed cultural institutions that governed sharing and reciprocity behaviour. These observations suggest that a sole interpretation based on the participants’ responses to the experimental protocol does not fully capture the meaning of their sharing behaviour. Including observational measures as part of the study design could offer a more accurate understanding of what the experimental data reveals within its broader cultural context.

Developmental scientists aim to extract meaningful insights from the historical, social, and ecological contexts in which development occurs, where complex interactions between these levels impacts development. This task becomes even more challenging when considering developmental changes, as children, their social partners, and their interactions are in a constant state of flux. The challenge is further compounded for researchers with limited experience or knowledge of the specific region or culture they are studying. Numerous assumptions shape the development of a research programme, questions, and constructs, and researchers may not always be aware of these biases – such as by providing a task that is over-reliant on 2D materials or interpreting ‘freezing’ as non-responsiveness. For developmental scientists conducting cross-cultural research, it is crucial to identify and avoid these biases and assumptions. Developing culturally grounded and valid measures requires careful planning, systematic checking, refinement, and thoughtful implementation – a process that often contrasts with the fast-paced nature of academic research (Rafiq et al., Reference Rafiq, Jordan, McNutt, Neelo, Attias, Boersma, Palmer, Ruesink and Abrahms2024). Unfortunately, the academic environment today often prioritises speed and quantity over depth and quality, fostering competitive, high-pressure research environments (Fernandez-Cano, Reference Fernandez-Cano2021; Frith, Reference Frith2020). Unreflective application of western-constructed and validated protocols to non-western cultures can lead to difficulties in inference and interpretation (Kline et al., Reference Kline, Shamsudheen and Broesch2018).

2.2. Dissociating developmental and cultural theory and methods

Contextual theoretical approaches, such as cultural-historical psychology, bioecological theory, and mediational theories of mind (Bronfenbrenner, Reference Bronfenbrenner1979; Cole, Reference Cole1996; Rogoff, Reference Rogoff2003), emphasise the importance of examining the child at various contextual levels. It is surprising that we continue to run experiments and develop measures with only an individual child’s response in mind. Children do not, and cannot, develop in isolation and, like others in the field, we suggest that if we are to have an accurate understanding of childhood, we must take time to consider the cultural context when developing research projects.

Despite a history of research showing that nature and nurture are intertwined, many well-intentioned researchers (ourselves included) have tended to approach culture as a thing to remove, isolate, control for, or work around. The goal is often to develop ‘clean’ tasks and protocols to examine social-cognitive abilities as isolated from cultural influences. Sometimes these tasks are developed to be performed with just one individual child (see Broesch et al., Reference Broesch, Callaghan, Henrich, Murphy and Rochat2010; Callaghan et al., Reference Callaghan, Rochat, Lillard, Claux, Odden, Itakura, Tapanya and Singh2005), and sometimes dyadically (see Corbit et al., Reference Corbit, Brunschwig and Callaghan2021), with a caregiver (see Clegg et al., Reference Clegg, Wen, DeBaylo, Alcott, Keltner and Legare2021), or in a group. These tasks may then be performed by a local individual with the objective of removing any barriers to understanding the task (e.g. language – verbal and non-verbal), or social barriers (e.g. performing a task with a stranger). It is possible that with the use of this widely accepted approach, we are only removing a language barrier (Stibbard-Hawkes et al., Reference Stibbard-Hawkes, Abarbanell, Mabulla, Endeko, Legare and Apicella2024). For instance, in our research in Vanuatu (Smit et al., Reference Smit, Broesch, Siegel and Mistlberger2019), although we asked a local adult to be the experimenter for a task, this did not guarantee that the researchers had a fluid comprehension of the language spoken by the local experimenter and the child. In one case, it took a week to accurately translate seemingly straightforward questions about sleep practices (e.g. ‘What time do you go to sleep most nights?’). Issues like these could potentially be avoided with clearer communication, but power dynamics and other motivations often prevent smooth and clear communication about such concerns (Urassa et al., Reference Urassa, Lawson, Wamoyi, Gurmu, Gibson, Madhivanan and Placek2021). Language barriers specific to a particular construct are just one of the many cultural layers that may unknowingly influence our results. One needs only to consider a contextual model to construe the endless possible influences that lead to individual and group decision-making.

Epistemology and methodology are deeply intertwined. Mainstream developmental psychology has failed to describe and understand children’s development in non-western populations (Nielsen et al., Reference Nielsen, Haun, Kaertner, Legare, Kärtner and Legare2017; Sanches de Oliveira & Baggs, Reference Sanches de Oliveira and Baggs2023). In fact, western psychology has a long history of describing development as a series of universal processes within an individual child that is independent of context.

Developmental theories are based on epistemological assumptions that guide the way we conduct research and interpret the behaviours we study. However, these underlying assumptions are often unexamined or unrecognised. Many methods used in developmental psychology are based on western standards, making them inadequate for capturing developmental processes even in contexts in which one is familiar (see Dahl, Reference Dahl2017). Developmental science operates within a broad set of concepts, making assumptions about the developing child and their developmental context, forming the conceptual framework that guides day-to-day research activities. For example, we assume that the child is born into a context in which they have parents and siblings who will scaffold the child for success in formal educational settings.

It is essential to articulate and examine the presuppositions or philosophical assumptions upon which theories, research, and methodologies are based. For instance, Stanton (Reference Stanton2013) pointed out that Indigenous epistemologies are very different from those valued by mainstream society. Specifically, Indigenous communities believe that experiences are best shared in a dynamic, interactive, and face-to-face context; however, mainstream academic epistemologies value professional contexts that use written forms. Moreover, Indigenous epistemologies value subjectivity and multiple perspectives (Guimarães, Reference Guimarães2020), which contrasts with the academic focus on reliability, validity, and consistency. A move away from Eurocentric epistemology would contribute to the development of methodologies that are more relevant to the cultural-specific contexts and would enhance construct validity in cross-cultural research.

2.3. Lack of causal frameworks

Though cross-cultural developmental research has emphasised experimental and reductionist methods, along with standardised protocols, it often fails to adopt robust causal frameworks. Most research questions in the behavioural sciences are causal in nature (Potochnik, Reference Potochnik2017; Rohrer, Reference Rohrer2024). There has been a recent increase in calls for more rigorous consideration of causal assumptions (the underlying assumptions when determining cause-and-effect relationships) and analyses (McElreath, Reference McElreath2022), including in cross-cultural research (Deffner et al., Reference Deffner, Rohrer and McElreath2022), to improve study validity. Causal analysis can benefit from integrating both quantitative and qualitative methods (Blersch et al., Reference Blersch, Franchuk, Lucas, Nord, Varsanyi and Bonnell2022). We encourage researchers to adopt a formal causal inference approach throughout the research process, from study design and participant recruitment (Greenland, Reference Greenland2022) to data analysis and interpretation (McElreath, Reference McElreath2022).

Recently, researchers have argued that cross-cultural scientists need to carefully consider a formalised causal inference approach to dealing with key steps in the research process (Deffner et al., Reference Deffner, Rohrer and McElreath2022; McElreath, Reference McElreath2022). By ‘formalised’ we mean explicitly identifying the effect of interest (often referred to as the theoretical estimand) (Lundberg et al., Reference Lundberg, Johnson and Stewart2021) and the empirical estimand (the data generated by specific tasks) (Chatton & Rohrer, Reference Chatton and Rohrer2024), and describing the relationship between the two using Directed Acyclic Graphs (DAGs). This allows us to pre-emptively address the assumptions for potential claims of validity (Rohrer, Reference Rohrer2024). In simple terms, a formalised causal inference approach makes clear what effect we are trying to examine and under what conditions we can confidently claim to have found an effect. While many studies may claim to have found a relationship between two variables, without a formal causal inference framework, we cannot be certain that the observed effect is genuine (Rohrer, Reference Rohrer2024).

To illustrate, consider intelligence as it is commonly operationalized and measured using non-verbal intelligence tasks. A more precise theoretical estimand of intelligence can be stated as the ‘expected score of a child in the Raven’s Progressive Matrices task, averaged over all school-going children in rural Thailand and Laos’. Using this approach allows us to circumvent the limitations of interpreting models (here, for intelligence) which are poorly formalised (van Hoogdalem & Bosman, Reference van Hoogdalem and Bosman2023), with vast variations of empirical evidence across different contexts. Following Deffner et al. (Reference Deffner, Rohrer and McElreath2022), we argue that a formalised framework can be a valuable tool to help describe and approach concerns in a more systematic way.

2.4. Superficial, short-term partnerships and collaborations

Cross-cultural research almost always requires extensive and multifaceted collaboration. Increasingly, publications and guidelines are emphasising the importance of effective, ethical, and equitable partnerships in developmental cross-cultural research (Broesch et al., Reference Broesch, Crittenden, Beheim, Blackwell, Bunce, Colleran, Hagel, Kline, McElreath, Nelson, Pisor, Prall, Pretelli, Purzycki, Quinn, Ross, Scelza, Starkweather, Stieglitz and Mulder2020, Reference Broesch, Lew-Levy, Kärtner, Kanngiesser and Kline2023; Burger et al., Reference Burger, Chen, Erut, Fong, Rawlings and Legare2023; Spake et al., Reference Spake, Hassan, Schaffnit, Alam, Amoah, Badjie, Cerami, Crampin, Dube, Kaye, Kotch, Liew, McLean, Munthali-Mkandawire, Mwalwanda, Petersen, Prentice, Zohora, Watts and Shaver2024; Urassa et al., Reference Urassa, Lawson, Wamoyi, Gurmu, Gibson, Madhivanan and Placek2021). However, many cross-cultural collaborations are short term, reflecting the traditional experimental model followed by mainstream psychology. Within these teams, collaborations are often vertical, with western researchers dominating discussions and leading projects, often in their native languages. Psychology researchers from non-western regions frequently report experiencing bias from their western counterparts, particularly when they conduct research in their own regions or are affiliated with non-western institutions (Raval et al., Reference Raval, Baiden, Espinosa-Hernandez, Magis-Weinberg, Nguyen, Titzmann and Zheng2023). These dynamics contribute to the lack of construct validity in cross-cultural research, as western theories and methods remain unchallenged, even when they fail to adequately explain observed phenomena. Our goal is to approach collaboration development through the lens of construct validity, ensuring that all perspectives are equally valued and incorporated.

Optimal collaboration is particularly important for improving construct validity in cross-cultural research because it introduces expertise in terms of cultural knowledge, academic disciplines, methodological development, and the implementation and dissemination of research to diverse networks. For example, input from researchers and stakeholders from the study of cultural context can help develop protocols that are perceived and administered as intended, reducing the influence of implicit biases from researchers from different cultural contexts (Hruschka et al., Reference Hruschka, Munira, Jesmin, Hackman and Tiokhin2018). Contribution from various disciplines can provide complementary perspectives or approaches. Researchers using experimental methods in cross-cultural research can gain valuable insights from anthropologists, who use qualitative methods to deepen their understanding of the study population, and vice versa (Weisman & Luhrmann, Reference Weisman and Luhrmann2020).

2.5. Culturally inappropriate tools and tests

Improving the validity and rigour of any study depends on a well-designed research plan and proper training of the research team, particularly those administering and coding tasks. This is especially important in international and cross-cultural developmental research, which often involves diverse groups of researchers across different geographical locations. In such studies, research questions must be carefully refined using developmental and cultural theory. Additionally, protocols must be translated, developed, rigorously piloted (with pilot data carefully interpreted), and administered in a way that ensures participants perceive them and respond as intended.

Translated materials must be appropriate and accessible for all intended recipients, including participants, community members, stakeholders, and research assistants. Ensuring this accessibility should be an integral part of the research process. This is particularly critical when working with children, as complex instructions or language can easily be misunderstood. One common cause of errors in cross-cultural research is poorly translated materials. To avoid misinterpretation, it is essential that words and phrases accurately capture the intended constructs; this requires the investment of careful attention and time. This diligence ensures that any variation in performance reflects true cultural differences, rather than issues with construct validity (Amir & McAuliffe, Reference Amir and McAuliffe2020; Burger et al., Reference Burger, Chen, Erut, Fong, Rawlings and Legare2023; Holding et al., Reference Holding, Anum, van de Vijver, Vokhiwa, Bugase, Hossen, Makasi, Baiden, Kimbute, Bangre, Hasan, Nanga, Sefenu, A-Hayat, Khan, Oduro, Rashid, Samad, Singlovic and Gomes2018; Pamei et al., Reference Pamei, Cheah and McBride2023). This concern differs from protocols that are simply poorly worded; the prevalence of English in academia can introduce systematic bias, especially when research begins with English-language materials that are shaped by American or British cultural assumptions.

In many studies, children are rewarded for participation, with common rewards in a western context including stickers, sweets, toys, and sometimes small amounts of money in exchange for participation. Rewards and payments are not always appropriate across different developmental populations due to contextual variations. In Brazil, for example, monetary rewards are forbidden in any kind of scientific research by the ethical legislation, and other types of reward are assessed on a case-by-case basis. Meanwhile in the USA, monetary rewards to children are relatively common. Researchers can either standardise or match rewards across populations or offer culturally specific ones, which can mean providing markedly different rewards for participants in different cultural contexts.

Providing both the same and different rewards across diverse populations can evoke varying levels of motivation or different perceptions, leading to disparities in engagement and outcomes of the research. In some cultures, monetary rewards of varying amounts might be highly motivating, while in others, social recognition or communal benefits are more effective. Additionally, the same reward may elicit different levels of motivation across populations of children, (e.g., sweets in one culture may not be as novel and motivating as in another, where they are not commonly found). Ignoring these differences can lead to unintended consequences, such as demotivation or misinterpretation of the rewards’ value, ultimately compromising the validity and reliability of the findings.

A high-profile example of this is the marshmallow task (Mischel et al., Reference Mischel, Shoda and Peake1988), which measures delayed gratification by testing whether children will wait to eat one marshmallow for the promise of a second marshmallow. Willingness to wait has been associated with better outcomes later in school, relationships, and health (Michaelson & Munakata, Reference Michaelson and Munakata2020). One study found that Japanese children delayed gratification for a marshmallow longer than American children. Yet when the reward was changed to a wrapped gift, American children delayed gratification longer than Japanese children (Yanaoka et al., Reference Yanaoka, Michaelson, Guild, Dostart, Yonehiro, Saito and Munakata2022). Waiting to eat is emphasised more in Japan than it is in the USA, whereas waiting to open gifts is emphasised more in the USA than it is in Japan. This highlights the importance of considering the cultural context for different rewards and incentives in study design to increase construct validity.

3.1. Including ethnographic and observational approaches

To enhance the construct validity of our experimental protocols, we recommend combining experimental methods with ethnographic and observational approaches. Collecting both quantitative and qualitative data enables researchers to identify patterns and develop contextually grounded interpretations (Dahl, Reference Dahl2017; Rogoff et al., Reference Rogoff, Dahl and Callanan2018; Xu, Reference Xu2017). We suggest three possible approaches for designing experiments that integrate both observational and ethnographic data.

1. 1) The first approach involves using an experimental protocol followed by observations. This method helps the researcher to understand cultural pathways within universal developmental processes (Greenfield et al., Reference Greenfield, Keller, Fuligni and Maynard2003). Experimental results can facilitate direct cross-cultural comparisons, allowing researchers to assess whether a phenomenon occurs similarly across cultures. This approach offers a way to compare performance in a controlled way. However, on its own, it may not capture cultural dynamics fully. While experimental results can confirm or reject a hypothesis, it may not be clear how meaningful a behaviour observed in an experiment is within a specific cultural context. Therefore, we recommend pairing experiments with observational and interview data to provide more meaningful insights.

As Kline et al. (Reference Kline, Shamsudheen and Broesch2018) suggest, it is possible to observe the same behaviour in different societies, yet the underlying mechanisms driving those behaviours may differ, or different behaviours in separate populations may stem from the same underlying mechanism. This principle is central to evolutionary anthropology, which seeks to explain how one ‘human nature’ can result in diverse cultural practices. An example is the ‘polygyny threshold’ model, which assumes that specific ecological conditions make polygyny more likely than monogamy (see Ross et al., Reference Ross, Mulder, Oh, Bowles, Beheim, Bunce, Caudell, Clark, Colleran, Cortez, Draper, Greaves, Gurven, Headland, Headland, Hill, Hewlett, Kaplan, Koster and Ziker2018 for a review and critique). Similarly, in developmental research, ‘statistical learning’ of language follows a comparable approach – one set of rules for all developing children can lead to highly diverse outcomes, depending on the cultural context. Because developmental processes and outcomes are not 1:1 pairs, it is crucial to investigate both processes and outcomes across different measures before drawing developmental conclusions from cross-cultural comparisons.

1. 2) A second approach involves conducting an experiment to challenge or confirm ecological validity. If an experiment is not yielding the expected results, researchers can pair those findings with observations and interviews to demonstrate that the experiment is not accurately capturing the intended construct. Lew-Levy et al. (Reference Lew-Levy, Pope, Haun, Kline and Broesch2021) applied this approach when examining tool innovation using the hook task with Congolese BaYaka forager children and Bondongo fisher–farmer children. They used structured observations and interviews to explain why their experimental findings did not support their predictions in terms of a methodological mismatch.

2. 3) The third approach involves starting with observations (either ethnographic or quantitative) and subsequently develop hypotheses and design a culturally specific experiment to test the dynamics observed. This approach is particularly useful for deepening cultural understanding or ‘keeping the tension’ between fieldwork and ethnography (Astuti, Reference Astuti2017). Experiments in this approach are effective for eliciting implicit knowledge that may not be easily verbalized (e.g., sharing norms) or that may be too complex for young children to verbalize. Lew-Levy et al. (Reference Lew-Levy, Andersen, Lavi and Riede2022) employed this approach to examine the developmental trajectory of intra- and interethnic sharing in the Republic of Congo, using a dictator game and basing predictions about the point at which these sharing norms would stabilise on ethnographic insights. While observational and ethnographic data can be time-intensive, integrating them into the research design helps prevent researchers from engaging in ‘helicopter research’, where they remain detached from the community.

3.2. Employing evidence-based, flexible causal frameworks

When adopting a formal causal inference approach, we suggest that cross-cultural science can learn from efforts made in other fields, such as health research (Haber et al., Reference Haber, Wieten, Rohrer, Arah, Tennant, Stuart, Murray, Pilleron, Lam, Riederer, Howcutt, Simmons, Leyrat, Schoenegger, Booman, Kang Dufour, O’Donoghue, Baglini, Do and Fox2022), sociology (Lundberg et al., Reference Lundberg, Johnson and Stewart2021), and personality psychology (Rohrer, Reference Rohrer2024). Drawing on Deffner et al. (Reference Deffner, Rohrer and McElreath2022), we present a brief set of guidelines for applying a causal framework to cultural developmental research.

Multisite studies should begin by clearly defining how the populations differ based on the specific study objectives or goals. The first step is to clarify what the theoretical estimand is, which is the key focus when conducting empirical analysis. The theoretical estimand is the unit-specific quantity for each target population. For example, we might estimate that the probability of 5–10-year-old children solving a task in a given population is 60%. Once the theoretical estimand is set, we need to attach it to the empirical estimand, which allows us to estimate the probability that all 5–10-year-old children in the study population will solve the task based on data collected from our sample (since we typically cannot collect data from every individual in a population).

Tools like DAGs are helpful to linking theoretical and empirical estimands (for an introduction and application to DAGs, see Rohrer, Reference Rohrer2018; Wysocki et al., Reference Wysocki, Lawson and Rhemtulla2022). DAGs are a graphical and mathematical tool to help researchers identify a priori causal assumptions between variables with three major structures – a confounder (a variable that independently influences both the predictor and outcome variables, X <- Cf -> Y), a mediator (a variable on the predictor–outcome causal path that influences the outcome variable, X -> M -> Y), and a collider (a variable mutually affected by both the predictor and the outcome, X -> Co <- Y). DAGs also contain nodes which capture whether the populations or contexts of interest differ from each other in relevant aspects. In the previous example of intelligence, a selection node S in a DAG can be used to represent the assumption that regions differ in their school starting age. If we want to study the effect of schooling on intelligence then we can have a variable, school starting age (instead of current age), as a selection node. Thinking ahead about such potential nodes facilitates the principled selection of variables for cross-cultural studies. Thus, DAGs help researchers understand and explain the data, as well as identifying the potential presence of selection bias (Hernán et al., Reference Hernán, Hernández-Díaz and Robins2004). Since cultures differ across many dimensions, using a formal causal framework enables researchers to identify appropriate study populations, select specific variables to measure (rather than capturing everything possible), and pinpoint causal relationships, leading to more accurate conclusions (Deffner et al., Reference Deffner, Rohrer and McElreath2022).

We also recognise the tension between proposing technical (statistical) solutions and advocating for an inclusive approach to cross-cultural research. Complex statistical techniques often require tools and training that are inaccessible to many researchers worldwide. As the expectation of sophisticated statistical approaches in research grows, many departments in wealthy, western institutions (in disciplines such as psychology or anthropology) have begun hiring dedicated statisticians or offering consultancy services to assist researchers. This problem is further exacerbated by the spread of generative-AI models (e.g., large language models) which often require exceptional amounts of computational resources. However, many researchers may not have access to training, appropriate software, or statistical support. These disparities can create inequities in publishing opportunities, as researchers from wealthier institutions often have an advantage when it comes to submitting to mainstream journals. This highlights a significant issue of academic inequality that we must continue to acknowledge and address (IJzerman et al., Reference IJzerman, Dutra, Silan, Adetula, Basnight-Brown and Forscher2021).

We note, however, that implementing a causal approach doesn’t necessarily require advanced training or resources. The first step is careful, transparent thinking about which variables influence the outcome measure. Descriptive information of relevant environmental and cultural variables, alongside rich ethnographic information, is critical to strong causal frameworks with clearly defined study populations. In cases where complex statistical and causal approaches are unavailable – or even when they are – we encourage researchers to include a ‘Constraint on Generalisability Statement’. This statement should explicitly identify and justify the target populations in empirical papers (Simons et al., Reference Simons, Shoda and Lindsay2017; Tiokhin et al., Reference Tiokhin, Hackman, Munira, Jesmin and Hruschka2019). The inclusion of this statement will afford more accurate inference and improve study replicability.

3.3. Conducting community-engaged and collaborative research

3.4. Culture-specific refinement and training

Culturally appropriate research design and thorough training of the research team are essential for enhancing validity and rigour in cross-cultural studies. Many principal investigators (PIs) and labs already have effective workflows for training and research design, with published guidelines available to help improve these methods (Burger et al., Reference Burger, Chen, Erut, Fong, Rawlings and Legare2023; Spake et al., Reference Spake, Hassan, Schaffnit, Alam, Amoah, Badjie, Cerami, Crampin, Dube, Kaye, Kotch, Liew, McLean, Munthali-Mkandawire, Mwalwanda, Petersen, Prentice, Zohora, Watts and Shaver2024). Our aim here is to offer additional suggestions that can be integrated into these workflows specifically to strengthen construct validity.

Whether a protocol is newly developed or has been adapted from an established one in another cultural context, it is essential for collaborators to prioritise input from local community members (Burger et al., Reference Burger, Chen, Erut, Fong, Rawlings and Legare2023). Protocols should be piloted extensively whenever possible, with pilot data shared with the collaborative team for review and refinement as needed. If translations are required, team members should ensure a rigorous process, such as using a multi-step translation and back-translation approach (Burger et al., Reference Burger, Chen, Erut, Fong, Rawlings and Legare2023; Holding et al., Reference Holding, Anum, van de Vijver, Vokhiwa, Bugase, Hossen, Makasi, Baiden, Kimbute, Bangre, Hasan, Nanga, Sefenu, A-Hayat, Khan, Oduro, Rashid, Samad, Singlovic and Gomes2018; Ozolins et al., Reference Ozolins, Hale, Cheng, Hyatt and Schofield2020). This cycle should be repeated until all parties are confident that the protocols are accurately understood as intended. The same careful attention should be given to disseminating findings, ensuring that published conclusions are culturally appropriate and informed by all collaborators to prevent misunderstandings or potentially harmful interpretations.

We emphasise that collecting pilot data is a critical part of research design and protocol development, requiring careful and systematic planning. Pilot studies – small-scale versions of the main study – can reveal potential strengths and weaknesses in protocols, highlight risks of failure in full-scale studies, and indicate whether instruments are eliciting the intended responses from participants (van Teijlingen & Hundley, Reference van Teijlingen and Hundley2002). Developmental research involves complexities that demand careful interpretation and handling.

Where possible, we recommend that PIs collect pilot data themselves to gain first-hand experience with the protocols and to understand participants’ perceptions and responses prior to full-scale data collection. Although the extent of PI involvement in data collection varies by institution, career stage, and study type, this hands-on step allows project leaders to appreciate key study details, such as task duration, participant reactions, adherence to protocol, and potential improvements for validity. If direct involvement is not feasible, PIs can review video recordings of the pilot data collection to gain similar insights.

Additionally, we suggest creating clear, accessible training videos for research assistants that supplement written protocol instructions. These videos should be in the appropriate language and provide visual guidelines on protocol administration, allowing assistants to watch them repeatedly for clarity and confidence (Heller Murray, Reference Heller Murray2024). Training videos can be especially useful for complex studies or for assistants with limited data-collection experience. Research assistants should also be encouraged to provide feedback on the training materials to refine and improve them. In previous cross-cultural studies, some authors of this manuscript have further required research assistants to record themselves administering tasks and receive feedback in order to minimize experimenter error.

Additionally, the research team should design reward systems that are perceived as fair and equitable across different populations. While the specific form of the reward may vary (e.g., monetary rewards, food, etc.), its value and impact should remain consistent. It is essential to use culturally appropriate incentives to ensure ethical and effective participation. The rewards system should reflect and respect the diverse values, norms, and socio-economic contexts of participants, and should be informed by community input and piloting. To assess the appropriateness of incentives and their effect on participant motivation, researchers should test different types of incentives during the pilot phase. This will allow them to identify potential issues with inappropriate incentives early on. By tailoring rewards to align with cultural expectations and values, and ensuring they are seen as fair and respectful, researchers can build trust, encourage meaningful participation, and maintain ethical standards, all while enhancing the validity and reliability of the data.

All these steps require significant time, and researchers should account for this in their project timeline. Rushing into data collection risks outcomes such as poorly designed protocols, low construct validity, and study failures. A key benefit of developing detailed, well-written protocols, easy-to-follow training videos, and conducting extensive piloting is that it supports the open science movement by increasing study replicability and transparency. Clear guidelines and transparent criteria for reward distribution are also crucial for maintaining equity in cross-cultural developmental research. By establishing these guidelines, researchers can ensure that the reward distribution process is well documented and consistently applied across diverse cultural contexts. This documentation ensures fairness and consistency, while also providing a clear record of the criteria used to justify specific incentives, ultimately supporting research integrity and validity.