Over the past 50 years, a wealth of testable, often conflicting, hypotheses has been generated about the evolution of offspring sex ratio manipulation by mothers. Several of these hypotheses have received support in studies of invertebrates and some vertebrate taxa. However, their success in explaining sex ratios in mammalian taxa, and especially in primates, has been mixed. Here, we assess the predictions of four different hypotheses about the evolution of biased offspring sex ratios in the well-studied baboons of the Amboseli basin in Kenya: the Trivers-Willard, female rank enhancement, local resource competition, and local resource enhancement hypotheses. Using the largest sample size ever analyzed in a primate population (n = 1372 offspring), we test the predictions of each hypothesis. Overall, we find no support for adaptive biasing of sex ratios. Offspring sex is not consistently related to maternal dominance rank or biased towards the dispersing sex, nor it is predicted by group size, population growth rates, or their interaction with maternal rank. Because our sample size confers power to detect even subtle biases in sex ratio, including modulation by environmental heterogeneity, these results suggest that adaptive biasing of offspring sex does not occur in this population.

The Amboseli Baboon Research Project (ABRP) is a long-term, longitudinal study of non-provisioned, individually recognized, wild baboons living in and around Amboseli National Park, Kenya. Demographic, behavioral, and environmental data have been collected on a near-daily basis since the inception of the project in 1971. Critical for the analyses presented here, ABRP has data on offspring conception, birth, and death dates as well as data on female dominance rank (see below) from 1971 to 2020. Additional description of the study population and its history can be found in Zipple et al. (in review).

The data files can be opened with any program that reads CSV (comma-separated values) files. If users wish to replicate our analyses exactly, they can use the R code, which must be used within the environment of the R statistical package.