Wing length canalisation and behaviour across birds: a phylogenetic meta-analysis of variance

Klaus Reinhold, Alfredo Sanchez-Tojar

Stronger stabilising selection is expected to lead to a decrease in trait variation (i.e., in higher canalisation). We examined this prediction across species by investigating individual variation in wing length across measured as the coefficient of variation (CV). We hypothesised that species that heavily rely on aerial feeding as well as long-distance migratory species should show higher canalisation (lower CV) in wing length than non-aerial feeders and non-migratory species. We collected species- and population-specific summary statistics on wing length for males (k = 340 CV values) and females (k = 310) from the literature (172 species, 314 populations) and analysed them using recently developed meta-analytic metrics for integrating phenotypic variance estimates. Our phylogenetic multilevel meta-analysis showed that heterogeneity among CV values was relatively low, indicating generalisability and replicability of the overall CV value (2.6%). Nonetheless, about a third of the heterogeneity was associated with phylogenetic relationships (23%) and differences among species (8%), a precondition to test for species-specific drivers of those differences. Although not all pairwise comparisons were statistically significant, all our analyses both at the species and the population level robustly and consistently showed higher canalisation in aerial (CV = 2.5% [2.2 to 2.9]) compared to non-aerial feeders (CV = 2.7% [2.4 to 3.0]), and in migratory (CV = 2.5% [2.2 to 2.9]) compared to non-migratory species (CV = 2.8% [2.4 to 3.2]). We conclude that wing length in bird species relying on their wings more heavily is likely under stronger (stabilising) selection, which in turn would have led to the observed higher canalisation on this trait for those species. Our study showcases how to combine already available descriptive statistics for phenotypic traits with underused meta-analysis of variance approaches to test often-neglected evolutionary predictions at the variance level. We hope to inspire others to expand our phylogenetic study to more species and life-history traits as well as to other wing traits of higher dimensionality such as wing area and shape, and more generally, to study canalisation across species and traits.