Linguistic diversity and theory in psychological science

Abstract

When it comes to understanding the human mind and behaviour, theory is often generated from and tested via samples that are not representative of the population as a whole. Over twenty years ago, this problem was brought into sharp focus with a number of authors calling for research in the cognitive and psychological sciences to be more aware of diversity and representativeness (e.g. Atran, Medin, & Ross, 2005;Norenzayan & Heine, 2005;Rozin, 2001) -a point also more recently recognised in comparative cognition (Andrews & Monsó, 2025). 'WEIRD' -Western, Educated, Industrialised, Rich, Democratic (Henrich, Heine & Norenzayan, 2010) -was coined as an acronym to raise awareness that most findings in psychological science and related fields are gleaned from studies with limited participant pools. Over a decade later, the extent to which such concerns have resulted in changes to practice can be questioned (e.g. Apicella, Norenzayen, & Henrich, 2020), and a coherent framework to embed linguistic and cultural diversity in psychological theorizing and empirical testing has yet to be established.In tandem with the recognition of a problem with representativeness in psychological science, two other major issues have emerged as ongoing concerns. First, there is the 'reproducibility crisis', pulling together a range of methodological and statistical issues, including worries regarding sample sizes, statistical reporting, questionable research practices (e.g. p-hacking) and publication bias (e.g. Open Science Collaboration, 2015). Second, a range of concerns have been expressed regarding the very nature of theorizing in the psychological sciences. Some have drawn attention to the lack of specificity of theories as one of the origins of the replication crisis itself, calling for more formal theories to be de rigour in the discipline (e.g. Oberauer & Lewandowsky, 2019;van Dongen et al., 2025), while others point the finger at the absence of an overarching theory formulation framework that may help to constrain theories to become more appropriately testable (e.g. van Rooij & Baggio, 2021,) akin to the 'levels of explanation' approach proposed by Marr (1982).In this article I consider how these three major concerns interconnect in consideration of one of the most essential elements of WEIRDness -the failure to embrace linguistic diversity in both theoretical and empirical psychological science.There are over 7,000 spoken and signed languages across the globe (Ethnologue, 2026), In contrast to the assumption from generative linguistics that languages share universal features (e.g. Chomsky, 1986;Pinker, 1996), a body of (largely) anthropological linguistic studies has shown that languages vary in significant ways, with diversity found at almost every level of linguistic organisation (Evans & Levinson, 2009). For example, languages differ across dimensions including sound systems, phonology, morphology, grammar, orthography, and semantics, leading Evans and Levinson (2009) to conclude that diversity is the most remarkable property of language. If the goal of the cognitive and psychological sciences is to identify and study universal psychological processes, akin to Newton's universal laws of physics (Kroupin, Davis & Henrich, 2025), then such linguistic diversity may challenge the very nature of the focus on universals (see Levinson, 2012 for discussion).In light of this linguistic diversity, one can ask how the cognitive and psychological sciences are doing prior and since Evans and Levinson (2009) in terms of sampling speakers of diverse languages. The answer is not very well. The cognitive and psychological sciences have been dominated with publication in English and testing of English-speaking participants. Arnett (2008) reported that the vast majority of articles sampled from each of six major (APA) psychology journals (across disciplines) within the preceding 5-year period reported data from English-speaking samples (ranging from 74-91% of articles across the six journals). Moreover, these English-speaking samples were from just five countries: United States, UK, Canada, Australia, and New Zealand. Thalmayer, Toscanelli and Arnett (2021) note that this state of affairs has barely changed in the intervening 10-year period, with a continued preponderance of samples (and authors) from English-speaking countries.This 'Anglocentric' bias may partly be due to the fact that English is spoken by a large percentage of the world's population (around one sixth speak English to a degree of proficiency; Ethnologue), as well as the continued dominance of psychology research in English speaking countries (Thalmayer et al., 2021). If English can be regarded as prototypical of the world's language, this bias might not be such a problem. But as Blasi et al. (2022) point out, English is atypical of the world's languages with respect to many key properties, including its writing system, grammaticalized categories, morphology, phonology and syntax (see Blasi et al., 2022 for discussion). This Anglocentric bias is also important for the generation of theory in the psychological and cognitive sciences, given the dominance of psychologists working in English (Arnett, 2008;Thalmayer et al., 2021). A lack of awareness of linguistic diversity can bring with it the assumption that the distinctions in one's language are universal categorical distinctions when they are not, affecting the construction of theory. Indeed, there are many cases where central psychological theoretical constructs may be a direct reflection of the language of the psychologists doing the theorizing rather than universal concepts sitting outside of language (Brick, Hood, Ekroll & de-Wit, 2022;Wierzbicka, 2013). For example, it is tempting to think that psychology is the science of the 'mind' and that 'mental life is a uniform concept across languages and cultures. But as Weisman et al. (2022) show, although it can be argued that a mind-body distinction may be universal, exactly how components of mental life (especially social-emotional components) are conceptualised differs substantially across languages and cultures (see also Lillard, 1998). To give a second example, the idea that there is a set of basic emotions with privileged status in human mental life and social interactions (e.g. as proposed by Ekman, 1992) has been challenged on the basis that the "enterprise is bedevilled by an overreliance on English affective terms" (Ortony, 2022, p. 41). Wierzbicka (1999) discusses the failure of high-frequency English "basic" emotion words to translate directly into other languages, noting for example that Russian does not have words that mean the same as "sadness" and "anger" in English. Thus, at the broadest level of theory, English-speaking psychologists make assumptions about central constructs that are not universally lexicalised/conceptualised across languages.Moreover, there is much evidence that the language one speaks is associated with behavioural differences across a broad range of domains. Three examples will make this clear. First, Chen (2013) investigated the relationship between the grammatical encoding of past and future (future time reference) in languages and a range of societallevel behaviours, finding that societies with languages that grammatically associate the future and the present (e.g. today it rains; it rains) are 31 percent more likely to have saved money in any given year, have accumulated 39 percent more wealth by retirement, are 24 percent less likely to smoke, are 29 percent more likely to be physically active, and are 13 percent less likely to be medically obese, compared to societies with languages that do not (e.g. today it rains; tomorrow it will rain). Second, languages encode number in different ways (see Xu, Liu, & Regier, 2020), with some languages possessing only approximate number terms (akin to vague quantifiers in English) while other have exact (recursive) number systems (e.g. English). It has been shown that the Pirahã, an Amazonian tribe with an absence of exact number terms in their language, are inaccurate in remembering large quantities compared to English speakers (Frank, Everett, Fedorenko, & Gibson, 2008;Gordon, 2004). Third, colour vocabularies exhibit considerable cross-linguistic variation, and a body of evidence indicates that speakers of different languages show different patterns of colour discrimination (see Lupyan, Rahman, Boroditsky & Clark, 2020 for a review), with effects emerge as early as 100ms after viewing a colour (e.g. Thierry, Athanasopoulos, Wiggett, Dering & Kuipers, 2009). While the relationship between language, cognition and culture are clearly complex (see Blasi et al., 2022 for discussion), the evidence that the language one speaks influences a wide range of behaviours is compelling.While most studies in psychology remain dominated by testing English speakers, it is important to note that the scientific study of language itself is not immune to this Anglocentric bias. Berghoff and Bylund (2025) document a lack of diversity of languages in a sample over 5500 published psycholinguistic studies over a twenty-year period, with 45% of published studied on English, and 76% overall comprising data from English and only nine other languages. Another key concern is that linguistics itself in many quarters lacks the methodological rigour that psychologists have been aspiring to for many years. The normative model in anthropological linguistics is to work with a small number of linguistic informants to glean information about language. This makes an implicit assumption that speakers of a language may make the same distinctions, and therefore that a few speakers of a language are representative of the population from which they are drawn. This assumption has recently been challenged with a largescale study of spatial demonstratives (e.g. this/that/here/there) across 29 languages. Coventry et al. (2023) tested the production of demonstratives experimentally with sample sizes with sufficient power to quantify variation both between and within languages. For all 29 languages tested, significant within-language variation was found, challenging the view that generalisations about a language can be made from a handful of linguistic informants. Variation within a population is something that is often overlooked in typical anthropological studies with small numbers of participants, and large-scale cross-linguistic experimental studies are relatively rare (see Majid et al., 2018;Coventry et al., 2023 for notable exceptions).One can ask what can be done to improve consideration of linguistic (and cultural) diversity in both psychological scientific theory and empirical testing. If the goal of the cognitive and psychological sciences is to identify what is and is not universal for our species, psychological science should move to conducting large-scale (suitably powered) studies across multiple languages and cultures as a matter of course. Practically this is challenging, given that research funding is still largely nationally determined, and the contact base for most researchers does not include extensive networks of researchers working on other languages and cultures. However, the Open Science Collaboration addressing issues of reproducibility across multiple labs and examples of such cross-linguistic collaborations across large numbers of languages and teams (e.g. Majid et al., 2018;Coventry et al., 2023) show that such endeavours are possible. Funders would do well to encourage more collaborative projects at the required scale.At a more modest scale, Majid (2023) suggests that culturally informed, theoretically motivated sampling can provide a useful model for research where larger scale sampling is not possible. To do so, researchers can select cultures that are maximally distinct for the research question at stake as a (more modest) test of universals, with iterative testing/replication across other languages that are similarly (maximally) distinct.Practically many researchers lack the knowledge of other languages and cultures to be able to select samples that are maximally distinct. Part of the issue here is that psychologists are simply not adequately trained in linguistic and cultural diversity. National and international bodies representing the discipline need to encourage university/school programmes for undergraduates and graduates to include training in this area.Finally, there is a need for journals and learned bodies to develop guidelines for reporting and to improve publication practices to encourage diversity in sampling. This should include explicit declarations regarding samples and assumptions at all stages of reporting. Blasi et al. (2022) note that English-language researchers do not have to mention (in titles, abstract, or limitations in discussions) that their findings pertain only to English, but this is not the case for researchers reporting results for other languages. Moreover, Kroupin Davis and Henrich (2025) call for an explicit statement regarding the environment/group tested and whether this group is universal or group specific. Publication outlets should also prioritise publishing findings from diverse samples recognising the importance of such endeavours to theory development in the psychological and cognitive sciences.