Within-pair paternity success relative to trill and male quality, with effects that remained significant after correcting for multiple testing in bold.

<p>Trill and male quality measures were the dependent variables, and the proportion of social offspring sired was the independent variable in mixed-effects models controlling for male identity, year, and (for trill measures only) trill type, pitch, trill duration, and the time of the trill in the song. Age, the only categorical dependent variable, was modeled with logistic regression; the estimated difference is the change in log-odds of being after second year with increasing proportion of social offspring sired. Following the recommendation of <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0059208#pone.0059208-Nakagawa2" target="_blank">[74]</a>, the effect size for age is a partial r using within-pair success as a categorical variable (all social offspring sired versus at least on social offspring sired by another male). For continuous variables, effect estimates are the slope, and the effect size is partial r. Abbreviations: obs., observations. CI, confidence interval. H:L, Heterophile:Lymphocyte ratio. PB, playback.