Generalised least squares regression analyses of the predictive effects of sperm length traits on the rate of extrapair paternity.

CVwm is the average coefficient of within-male variation in sperm length. CVbm is the coefficient of between-male variation is sperm length, adjusted for sample size (see Methods). The statistical analyses were based on transformed variables to approach normality (arcsine square-root for the proportion of extrapair young, log10 for sperm length, CVwm and CVbm). Slopes were tested against the prediction of 0 using a t-test. The λ-values express the degree of phylogenetic dependence of the associations, and the two superscripts indicate P-values of likelihood ratio tests of λ against models of λ = 0 (no phylogenetic dependence) and λ = 1 (full phylogenetic dependence), respectively. r2 values indicate the proportion of total variance explained.