When does matriliny fail? The frequencies and causes of transitions to and from matriliny estimated from a de novo coding of a cross-cultural sample

Abstract

The question of when and why societies have transitioned away from matriliny to other types of kinship systems-and when and why they transition towards matriliny-has a long history in anthropology, one that is heavily engaged with both evolutionary theory and cross-cultural research methods. This article presents tabulations from a new coding of ethnographic documents from the Standard Cross-Cultural Sample (SCCS), tallying claims of transitions in kinship systems both away from and to matriliny using various levels of stringency. We then use our counts as the outcome variables in a set of Bayesian analyses that simultaneously estimate the probability of a transition occurring given societal covariates alongside the conditional probability of detecting a transition given the volume of ethnographic data available to code. Our goal is to estimate the cross-cultural and comparative frequency of transitions away from and to matriliny, as well as to explore potential causes underlying these patterns. We find that transitions away from matriliny have been significantly more common than 'reverse transitions' to matriliny. Our evidence suggests that both rates may be, in part, an artefact of the colonial and globalizing period during which the data comprising much of the current ethnographic record were recorded. Analyses of the correlates of transitions away from matriliny are consistent with several of the key causal arguments made by anthropologists over the past century, especially with respect to subsistence transition (to pastoralism, intensive agriculture and market economies), social complexity and colonialism, highlighting the importance of ecological factors in such transitions. This article is part of the theme issue 'The evolution of female-biased kinship in humans and other mammals'.

Keywords: SCCS; cross-cultural research; eHRAF; kinship; matriliny; subsistence.

Figure 1.

Estimates of occurrence and detection of transitions away from matriliny: Bayesian double-logit model using the level of coding restriction that excludes claims citing no supporting evidence. Estimated values can be found in electronic supplementary material table S4.1.4. The y-axis provides an abbreviated version of the variable name followed by the SCCS variable number, with the sample size in parentheses (not all SCCS societies are coded for all variables). Each row represents a separate model. Variables are arranged in three categories, then sorted from top to bottom by estimated effect size. Full variable names are provided in electronic supplementary material, table S2.1. The left-hand panel displays the estimated odds ratio and 95% confidence interval for the societal covariate. The vertical bar indicates an odds ratio of 1, or no effect. Cases whose error bars are entirely above or below this line are significantly different from an odds ratio of 1 and are coloured (lighter) orange, with point estimates marked with triangles; other cases are coloured (darker) blue. For formatting purposes, error bars are truncated at 10. The right-hand panel provides the predicted probabilities of the model estimates. The blue lines extending from the y-axis to the points indicate the estimated baseline probability of a transition away from matriliny in the absence of the covariate. The orange lines extending from the points indicate the additional effect of the covariate (or reduction where this line extends to the left of the point). Orange triangular points indicate that the covariate effect is significant. The x marks indicate the estimated conditional probability of detecting a transition (assuming that one occurred) in an ethnographic corpus with mean number of pages and publication year span. (Online version in colour.)

Figure 2.

Estimates of occurrence and detection of transitions away from matriliny: Bayesian double-logit model using the level of coding restriction that excludes cases of limited social scope. Estimated values can be found in electronic supplementary material, table S4.1.5. The y-axis provides an abbreviated version of the variable name followed by the SCCS variable number, with the sample size in parentheses (not all SCCS societies are coded for all variables). Each row represents a separate model. Variables are arranged in three categories, then sorted from top to bottom by estimated effect size. Full variable names are provided in electronic supplementary material, table S2.1. The left-hand panel displays the estimated odds ratio and 95% confidence interval for the societal covariate. The vertical bar indicates an odds ratio of 1, or no effect. Cases whose error bars are entirely above or below this line are significantly different from an odds ratio of 1 and are coloured (lighter) orange, with point estimates marked with triangles; other cases are coloured (darker) blue. For formatting purposes, error bars are truncated at 10. The right-hand panel provides the predicted probabilities of the model estimates. The blue lines extending from the y-axis to the points indicate the estimated baseline probability of a transition away from matriliny in the absence of the covariate. The orange lines extending from the points indicate the additional effect of the covariate (or reduction where this line extends to the left of the point). Orange triangular points indicate the covariate effect is significant. The x marks indicate the estimated conditional probability of detecting a transition (assuming that one occurred) in an ethnographic corpus with mean number of pages and publication year span. (Online version in colour.)