Neighbours and relatives: accounting for spatial distribution when testing causal hypotheses in cultural evolution

Abstract

Many important and interesting hypotheses about cultural evolution are evaluated using cross-cultural correlations: if knowing one particular feature of a culture (e.g. environmental conditions such as temperature, humidity or parasite load) allows you to predict other features (e.g. language features, religious beliefs, cuisine), it is often interpreted as indicating a causal link between the two (e.g. hotter climates carry greater disease risk, which encourages belief in supernatural forces and favours the use of antimicrobial ingredients in food preparation; dry climates make the production of distinct tones more difficult). However, testing such hypotheses from cross-cultural comparisons requires us to take proximity of cultures into account: nearby cultures share many aspects of their environment and are more likely to be similar in many culturally inherited traits. This can generate indirect associations between environment and culture which could be misinterpreted as signals of a direct causal link. Evaluating examples of cross-cultural correlations from the literature, we show that significant correlations interpreted as causal relationships can often be explained as a result of similarity between neighbouring cultures. We discuss some strategies for sorting the explanatory wheat from the co-varying chaff, distinguishing incidental correlations from causal relationships.

Did alcohol facilitate the evolution of complex societies?

“…Model 2 adds the rst potential confounders (spatial proximity and common descent), since it is known that geographically close and historically related societies tend to share many aspects of culture, environment and demography that are likely to affect both the presence of alcohol and the level of political complexity 24,71 .…”

The world has gone mad

“…A few papers are out (e.g. Bromham & Yaxley, 2023 ; Major-Smith, 2023 ) and many more on the verge of publication; I can assure you it is nearly there. We are also just starting to compile a special collection on cultural evolution of the arts, guest edited by Oleg Sobchug and Mason Youngblood.…”

Editorial: The adaptive value of languages: non-linguistic causes of language diversity, volume II

“…Although in practice many researchers do not draw a clear distinction between statistical tests and causality, it should be clear that correlation and statistical inference do not mean causality within cross-linguistic or cross-cultural datasets (Bromham and Yaxley, 2023 , inter alia). Regardless, recent studies investigating phonological diversity and non-linguistic factors bring together multiple lines of research to investigate, and thus try to validate, correlation patterns in terms of causality (Bromham and Yaxley, 2023 ; cf. Hernan and Robins, 2020 ).…”

“…In sum, investigating language evolution requires “new” methods for studying causal associations between linguistic and non-linguistic variables. While researchers strive for methods for dealing with statistical biases including phylogenetic, spatial, and diachronic autocorrelation (Moran et al, 2021 ; Bromham and Yaxley, 2023 ), the current state of the art uses multifaceted strains of correlational evidence to try to support causality ( Maddieson and Benedict ). Additional experimental findings, “big” data, and new approaches to estimate causal effects from observational data, i.e., causal inference or causal networks (Roberts et al, 2020 ), are the current avenues aimed at fruitful progress.…”