Parasites and politics: why cross-cultural studies must control for relatedness, proximity and covariation

Source: Royal Society

A growing number of studies seek to identify predictors of broad-scale patterns in human cultural diversity, but three sources of non-independence in human cultural variables can bias the results of cross-cultural studies. First, related cultures tend to have many traits in common, regardless of whether those traits are functionally linked. Second, societies in geographical proximity will share many aspects of culture, environment and demography. Third, many cultural traits covary, leading to spurious relationships between traits. Here, we demonstrate tractable methods for dealing with all three sources of bias. We use cross-cultural analyses of proposed associations between human cultural traits and parasite load to illustrate the potential problems of failing to correct for these three forms of statistical non-independence. Associations between parasite stress and sociosexuality, authoritarianism, democracy and language diversity are weak or absent once relatedness and proximity are taken into account, and parasite load has no more power to explain variation in traditionalism, religiosity and collectivism than other measures of biodiversity, climate or population size do. Without correction for statistical non-independence and covariation in cross-cultural analyses, we risk misinterpreting associations between culture and environment.

1. Introduction

The search for meaningful predictors of variation in human cultural traits and the diversity of human societies has a long history [1,2]. Modern analyses of broad-scale cultural diversity use an array of statistical analyses to extract patterns from global data, such as the effect of primary productivity on hunter–gatherer populations [3], the interaction between agricultural practises and religious beliefs [4] and the influence of rivers on language diversity [5]. However, many such analyses fail to account for one or more sources of statistical non-independence inherent in large observational datasets, which can lead to spurious relationships between traits and environments. Non-independence violates fundamental statistical assumptions and can lead to inflated degrees of freedom, incorrect parameter estimates, and the false impression of direct, causal relationships between variables that are only indirectly or incidentally linked. We can consider the problems of statistical non-independence in cross-cultural studies in three broad categories: phylogenetic non-independence, spatial autocorrelation and covariation among variables.

1.1. Galton's problem or non-independence due to evolutionary relationships

Human populations are related by descent, so closely related societies share many cultural traits that they have inherited from their shared ancestors. Societies that share a more recent common ancestor are likely to be more similar in many aspects of culture, including religious beliefs, material culture and social norms, than they are to more distantly related societies [6–8].