Statistical analyses
All analyses were carried out using the R statistical package (v. 3.5.2). The same model structure was followed for the analysis of all traits: the substrate cues (i.e., cues left on the patch by virgins or by mated females that were removed prior to the beginning of the mating sessions) and the female mating status (i.e., virgin or mated females present on the patch during the mating session) were fitted as fixed explanatory variables, whereas block (the day and time of the day at which the experiment was done) was fitted as a random explanatory variable (see Table S1).
Copulation duration was examined as “copulation duration of the first mating only” and “copulation duration across mating events”. In the analysis of the latter variable, the order of each copulation (i.e., whether it was the first, second, third mating, etc) was added as a covariate. All possible interactions between fixed factors were included.
The number of mating attempts and the number of mating events were analysed using a Poisson distribution (glmer , lme4 package; . The frequency of female acceptance was analysed using a binomial distribution (glmer , lme4 package), with the formulation of the dependent variable including the number of female rejections and acceptances within a cbind function. The duration of the first mating and the copulation duration across events were tested for normality and analysed using linear mixed-effect models (lmer , lme4 package; . Male survival was analysed using a Cox proportional hazards mixed-effect model (coxme , coxme package; , with the death of males being classified as natural or censored. This model included the number of matings as a covariate.
All maximal models were simplified by sequentially eliminating non-significant terms from the highest- to the simplest-order interaction . The significance of the explanatory variables was determined using Wald F tests, for continuous distributions and χ2 tests for discrete distributions .