Virgin males were immediately aspirated and housed in vials of Lewis medium and excess live yeast grains at 18°C in their experimental triplets: ‘related familiar’, ‘related unfamiliar’, ‘unrelated familiar’ and ‘unrelated unfamiliar’. ‘Related familiar’ comprises three males collected from the same ‘single family’ vial. ‘Related unfamiliar’ comprises one male taken from each of the three ‘single family’ vials of the same family. ‘Unrelated familiar’ comprises three males taken from the same ‘mixed family’ vial. ‘Unrelated unfamiliar’ comprises one male taken from each of three ‘single family’ vials of three different families. No family contributed to more than one vial of each related familiar and related unfamiliar treatments, and families were randomly assigned so that each had an equal contribution to the unrelated familiar and unrelated unfamiliar treatments. Two days before the introduction of females, males were transferred to fresh vials and kept at 25°C. To produce virgin females, we reared eggs from the cage population at 18°C at standard density (approx. 250 flies per 75 ml bottle containing 45 ml of Lewis medium) in parallel with male collection, collected adult females under ice anaesthesia and aged them at 25°C in individual yeasted vials for 3 days before the start of the experiment.
We performed the experiment across two blocks, producing a combined total of 95 related familiar triplets (39 in block 1, 56 in block 2), 86 related unfamiliar triplets (37 in block 1, 49 in block 2), 96 unrelated familiar triplets (22 in block 1, 74 in block 2) and 86 unrelated unfamiliar triplets (33 in block 1, 53 in block 2). Differences in sample sizes across treatments are due to some flies escaping and stochastic variation in the number of adult males emerging in each family vial within the short collection period.
Behavioural observations
On day 1, we added a single virgin female to each male triplet in a randomly numbered vial to blind the observer to the treatment throughout data collection. On days 2, 3, 4, 5, 8, 9, 10, 11 and 12, we observed the vials during eight scans in the morning (only seven scans on day 2, block 1), 10 – 20 min apart and recorded whether any males were displaying aggression [37], courtship [38] and mating behaviours. Note that in Carazo et al. [22], triplets of males were replaced at regular intervals to prevent males co-ageing with the female. The set-up we used to generate unrelated familiar males prevented us from replacing males during the experiment, therefore males were allowed to age with the female in this study, and as such, we did not expect a similarly strong level of female harm as reported in [22].
Flies were transferred to fresh vials under light CO2 anaesthesia on days 3, 5, 8 and 11 in both blocks and additionally on day 15 in block 2, and the vials were retained to collect adult offspring (see Fitness measures). Vials were discontinued upon the female’s death, and we recorded the day of death up to day 15 in block 1 and up to day 19 in block 2 after which time any remaining females (6% across both blocks) were censored. We also censored vials in the event of male death (four related familiar vials, one related unfamiliar vial, one unrelated familiar vial, one unrelated unfamiliar vial) or flies escaping during handling (one related familiar vial, one related unfamiliar vial).
Fitness measures
Vials containing the offspring of experimental groups were reared at 25°C for 16 days, allowing sufficient time for offspring to develop to the pupal or adult stage, when they were then frozen. To account for different egg-to-adult development times between vials, we counted adult flies and pupae that had reached the P13 phanerocephalic pupal phase [39], identified by the black wing colour, and included both in offspring counts.
Statistical analysis
Survival models were performed using JMP [40]. All other models were performed using the MASS package [41] in R [42] using type III sums of squares to calculate p-values. For all analyses, we included block—and all interaction terms that include block—as fixed effects. In all cases, the interaction terms that include block were not significant (electronic supplementary material, table S2), so we removed these terms from the models and kept block as a fixed main effect. While we know which families contributed to the related familiar, related unfamiliar and unrelated unfamiliar treatments, our experimental design makes it impossible to know the family identities of flies in the unrelated familiar treatment. As our knowledge of family identity is confounded with treatment, we were not able to include family identity in the full model analysis.
For aggression and courtship, we analysed the number of scans per day in which the behaviour was observed with a binomial penalized quasi-likelihood GLMM [43], with relatedness, familiarity and their interaction, and block as fixed effects, and day within vial ID as a nested random effect. For mating rate, we analysed whether or not a mating was observed for each day using a binomial penalized quasi-likelihood GLMM with relatedness, familiarity and their interaction, and block as fixed effects, and vial as a nested random effect.
For female reproductive success, we analysed the total number of offspring produced during the experiment. Only 27 of the 357 females were still reproducing at the end of the experiment, and short-term reproductive success is known to be a strong predictor of long-term reproductive success in this species [44]. Therefore, our measurements of reproductive success during the experiment can be considered a very close estimate of lifetime reproductive success. Vials in which a male died
