American Society of Naturalists

A membership society whose goal is to advance and to diffuse knowledge of organic evolution and other broad biological principles so as to enhance the conceptual unification of the biological sciences.

“Common field data limitations can substantially bias sexual selection metrics”

Posted on

Emily R. A. Cramer, Sara A. Kaiser, Michael S. Webster, and T. Brandt Ryder (Aug 2020)

Estimating sexual selection metrics from standard field data may result in substantial levels of bias

Read the Article

Comparing sexual selection across species and populations is key to test the evolutionary causes and consequences of this process. However, a paper appearing in The American Naturalist shows that metrics used in such comparisons are likely to be strongly biased in typical field studies, particularly for metrics that require estimating the number of copulation partners. Most field studies infer the number of copulation partners an individual has by using genetic tools to assign parentage of sampled offspring. Copulations that fail to fertilize eggs are not detected, so mating partners are consistently under-estimated. Furthermore, this approach creates an artificially high correlation between the number of detected mating partners and the number of offspring. Researchers were already concerned about such bias, and several empirical studies on captive individuals, where copulations could be directly observed, supported that concern. However, the extent of the issue had not been thoroughly explored. In this study, researchers examine how data limitations inherent in studies of wild bird populations affect the accuracy of the four most widely used sexual selection metrics. The authors generate 39,000 computer-simulated populations of socially monogamous breeding pairs, belonging to 15 species with varying levels of extra-pair paternity to evaluate several types of field data limitations, using a range of biologically relevant values. They found substantial bias when copulations are inferred from parentage outcomes rather than being observed directly. The degree of bias differs among species and due to factors such as male infertility, nest predation, and incomplete sampling of extra-pair offspring. In addition to providing code for other researchers to assess data limitations within their own study populations, these authors suggest using this tool to make informed choices when selecting and interpreting sexual selection metrics.


Abstract

Sexual selection studies widely estimate several metrics, but values may be inaccurate because standard field methods for studying wild populations produce limited data (e.g., incomplete sampling, inability to observe copulations directly). We compared four selection metrics (Bateman gradient, opportunity for sexual selection, opportunity for selection, and s′max) estimated with simulated complete and simulated limited data for 15 socially monogamous songbird species with extra-pair paternity (4-54% extra-pair offspring). Inferring copulation success from offspring parentage creates non-independence between these variables and systematically underestimates copulation success. We found that this introduces substantial bias for the Bateman gradient, opportunity for sexual selection, and s′max. Notably, 47.5% of detected Bateman gradients were significantly positive for females, suggesting selection on females to copulate with multiple males, though the true Bateman gradient was zero. Bias generally increased with the extent of other sources of data limitations tested (nest predation, male infertility, and unsampled floater males). Incomplete offspring sampling introduced bias for all metrics except the Bateman gradient, while incomplete sampling of extra-pair sires did not introduce additional bias when sires were a random subset of breeding males. Overall, our findings demonstrate how biases due to field data limitations can strongly impact the study of sexual selection.