Figure 1.

Geography of twinning rate and norms about the treatment of twins. (a) National twinning rate per 1000 births (adjusted for average maternal age) in 76 countries; data from Smits and Monden (2011). (b) A closer look at Africa. (c) Percentage of land area historically held by predominantly non-twin-killing groups, a proxy for geminophilous norms; data from Fenske and Wang (2023, fig. 4), who retrieved the information from Murdock (1959). Naive country-level regressions suggest that there are 3.84 (95% confidence interval, CI, 0.72, 6.96; adjusted R² = 0.12; N = 36; P = 0.017) more twin births per mille in countries where non-twin-killing groups make up the entire population, relative to countries where such geminophilous norms are not documented. Such an analysis, however, treats an absence of evidence of geminophilous norms, as evidence of absence. Restricting the sample to countries for which at least 50% of territory is unambiguously coded as historically populated by either twin-killing or non-twin-killing groups, the coefficient increases to 4.17, but the confidence region expands (95% CI, −3.02, 11.36; adjusted R² = 0.02; N = 21; P = 0.241) to include the value of 0, owing to the smaller sample of countries. Finer-scale models are needed to make such comparative analyses rigorous, as simple regressions are subject to ecological confounding.