Coevolution of religious and political authority in Austronesian societies

Authority, an institutionalized form of social power, is one of the defining features of the large-scale societies that evolved during the Holocene. Religious and political authority have deep histories in human societies and are clearly interdependent, but the nature of their relationship and its evolution over time is contested. We purpose-built an ethnographic dataset of 97 Austronesian societies and used phylogenetic methods to address two long-standing questions about the evolution of religious and political authority: first, how these two institutions have coevolved, and second, whether religious and political authority have tended to become more or less differentiated. We found evidence for mutual interdependence between religious and political authority but no evidence for or against a long-term pattern of differentiation or unification in systems of religious and political authority. Our results provide insight into how political and religious authority have worked synergistically over millennia during the evolution of large-scale societies.


Dynamic co-evolutionary model
Adapting previous work 3 , we model the coevolution of political and religious authority as a multivariate stochastic differential equation, similar to a multivariate Ornstein-Uhlenbeck (OU) process. OU processes are mean-reverting stationary Gauss-Markov processes, whereby a trait changes due to both deterministic reversion towards some central value and stochastic Gaussian noise. In an evolutionary context, these deterministic and stochastic components of the OU process are often referred to as 'selection' and 'drift', respectively.
Using an OU process, we model the evolutionary history of political and religious authority on the Austronesian language phylogeny as a time series. We allow the deterministic dynamics of the OU process ('selection') to play out over the length of each tree segment, and add the stochastic drift components ('drift') to the end of each segment as independent samples from a standard normal distribution. The differential equation is as follows: As outlined in Driver et al. 4 and Ringen et al. 3 , the solution to this differential equation for any time interval − 0 is: where A # = A ⊗ I + I ⊗ A with⊗ denoting the Kronecker-product, I is an identity matrix, row() is an operation that takes elements of a matrix row-wise and puts them in a column vector, and irow() is the inverse of the row operation.
We map this model onto the Austronesian language phylogeny with the rstan R package 5 .
Following the algorithm in Ringen et al 3 ., we divide the evolutionary history of each lineage into tree segments, where each tree segment begins with the parent node and ends with the child node or tip, and calculate the length of each segment . We then initialise the ancestral trait values for political and religious authority and, for each segment, solve the above equation for ( ). We repeat the above steps for all taxa on the phylogeny. To account for phylogenetic uncertainty, we combined the posteriors across 100 models fitted to 100 randomly-drawn posterior phylogenetic trees. To account for spatial autocorrelation in the locations of Austronesians societies, we include a Gaussian process over longitude and latitude coordinates. Standard MCMC diagnostics (̂ ≤ 1.05) and trace plots suggested that the model converged normally.
In our dynamic co-evolutionary model with two latent variables, political authority 1 and religious authority 2 , we can calculate the equilibrium trait values for both latent variables as: In the main text, we report the standardised difference in the equilibrium value for one trait, given an absolute deviation increase in the other trait ( ).
For further details about this dynamic co-evolutionary model, see Ringen et al. 3