The mites came from a laboratory-adapted base population (> 1000 individuals) collected near Stirling in central Scotland . In the laboratory, mites were maintained at 22–23 °C and > 90% relative humidity and fed with dried yeast twice a week. Obtaining virgin individuals for the experiments was started by collecting quiescent (immobile) mites in the second or final juvenile stage (protonymph or tritonymphs, respectively) from the stock population. Two male morphs occur in S. berlesei: aggressive heteromorphs (fighters) and nonaggressive homeomorphs (scramblers). Because male morph determination takes place at the early tritonymphal stage  and is based on population density cues  and/or body size and weight [52, 53], selecting a particular stage prior to isolation helped to control the proportion of fighters among adults. The mites were reared individually in glass tubes (Ø 8 mm) with bases made with plaster of paris and charcoal and closed with cotton stoppers. All tubes were kept on Petri dishes (Ø 100 mm) that were covered with filter paper moistened once a day with tap water. The filter paper was replaced once a week. The same conditions were maintained throughout the experiments described below.
(a) Effect of mating frequency on female fitness
All females and males used in the following experiment were obtained by isolating quiescent protonymphs and tritonymphs, the last juvenile stage, from the stock population in individual vials. Between 12 and 24 h, after the final moulting, all females were randomly subdivided into two treatments with different mating frequencies. In the high mating frequency (HMF) treatment, each female (n = 59) had access to a male for 20 h a day; each male was transferred every day and mated with another female from the low mating frequency (LMF, n = 59) treatment for 4 h a day. Females were always paired with the same male throughout the experiment. To control the effect of male morph on female fitness, I used both fighters (eclosed from protonymphs; n = 29) and scrambler males (eclosed from tritonymphs; n = 30). Dead males were replaced with new ones of the same morph and age; mating with more than one male was recorded. To avoid adverse impacts of manipulation on female survival, females were transferred to new vials only twice - on the 6th and 11th days of the experiment.
Females were checked daily for survival for the entire duration of the experiment (18 days). Survival analyses were performed using a mixed-effect Cox model with male identity as a random factor and treatment, male morph, interaction between treatment and male morph as well as mating with more than one male as a fixed factors (function coxme, R package coxme ). The results from the mixed-effect Cox model revealed a significant interaction between mating frequency treatment and male morph. To investigate this interaction in detail, I conducted a separate Cox proportional hazards regression model for fighter and scrambler morphs (function coxph, R package survival [55, 56]).
As a measure of female fecundity, I used the number of eggs laid between the 6th and 11th days of the experiment, as a preliminary survey showed that the number of eggs laid during this period is highly correlated (r = 0.94, n = 41, p < 0.001) with lifetime fecundity. Female fecundity was analysed using a generalised linear mixed model (GLMM) with a negative binomial error distribution (function glmmadmb, R package glmmADMB [57, 58]) to account for overdispersion. Male identity was introduced into the model as a random factor, and treatment, male morph, female longevity and the treatment * male morph interaction were fixed factors. Females that died before the 6th day of the experiment were excluded from the fecundity analysis. The goodness-of-fit of the models was evaluated using the Akaike information criterion (AIC)  in both the mixed-effect Cox model and GLMM.
(b) Manipulation of juvenile diet quality
I started the following experiments by collecting the quiescent protonymphs from the stock population and housing them individually. Individuals were randomly subdivided into two diet treatments: low-quality diet and high-quality (control) diet. Mites from the high-quality treatment were fed with dried yeast ad libitum; the low-quality treatment contained filter paper, which is considered to be a reduced-quality food  and one grain of dried yeast (dimensions ca. 0.25x0.5x0.25 mm). Between 12 and 24 h before adults emerged, all individuals from both treatments were fed with the same amount of dry yeast (3-4 grains) to avoid the effect of starvation on the traits measured. Virgin individuals 12 to 24 h after the final moulting were used for experiments to avoid possible differences due to age.
To investigate whether diet treatment influenced the condition of an individual, the body length (excluding mouthparts) of virgin adult males was measured from digital images using ImageJ software, version 1.52a (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, https://imagej.nih.gov/ij/, 1997-2018). The effect of treatment on body length was analysed with the Mann-Whitney U test. Since the low-quality diet used in this part suppressed the expression of the fighter morph (90% of obtained males were scramblers), only scrambler males were used in all subsequent experiments.
(i) Effect of juvenile diet quality on male sexual behaviour and competitiveness
To assess how the diet regime influences male competitiveness and fertilisation success, I used a male sterilisation-based method [61, 62]. Virgin males from the stock population were irradiated with 120 Gy gamma rays (137Cs source giving approximately 3.55 Gy/min; Gammacell® 1000 Elite, BestTheratronics Ltd, Canada). After that dose, irradiated male sperm can fertilise eggs, but the eggs exhibit 100% embryonic mortality; thus, the eggs remain unhatched. Virgin females from the stock population (n = 78) were paired with males in the following order: irradiated male, focal male, irradiated male. The fertilisation success of the focal males was measured as the proportion of hatched eggs. Each male was kept with a female for 60 min, ensuring the occurrence at least one copulation event (personal observation). During the 60 min of observation, I recorded the number of copulation attempts, the time to start of the first copulation event and the time spent on copulation by a focal male. For the next three days, the females laid eggs, and after 5-7 days, the proportion of eggs that hatched was determined. Due to logistical constraints, the experiment was performed in two blocks with 18-20 replicates of each experimental treatment in each block.
The proportion of eggs sired by focal vs irradiated males was analysed using a generalised linear model (GLM) with quasibinomial errors to account for overdispersion . The number of larvae and unhatched eggs were treated as two-vector response variables . Three observations were removed from the analysis because the females did not lay any eggs.
Statistical analyses of male sexual behaviour were performed using the GLM with Poisson error for the number of copulations and linear model for the time to start of mating and mating duration. The time to start of mating was log-transformed to improve the normal distribution of the residuals. Distribution of residuals were checked with diagnostic plots and Shapiro–Wilk test. Due to missing values, one, five and one observations were removed from the analysis of the number of copulations, time to start of mating and mating duration, respectively. All models included treatment (low- vs. high-quality diet) and block as fixed factors.
(ii) Effect of juvenile diet quality on male-induced harm
To examine how the male condition affects female reproductive output and lifespan, I paired one virgin female from the stock population with one virgin male from the high-quality (n = 48) or low-quality (n = 45) food treatment in a 0.8-cm-diameter glass cell. On the second and sixth days of the experiment, I replaced the males with new ones from the same treatment. Allowing the female to copulate with more than one male decreased the possible effect of the male on fecundity due to occasionally observed male sterility.
The experiment was maintained for a 14-day period during which females were checked daily for survival. Survival analyses were performed using the Kaplan-Meier survival estimate and Cox proportional hazards model by comparing survival curves (function survfit and coxph, R package survival; Therneau 2015). Due to the logistical constraint and the time-consuming process of egg counting, the effect of diet quality treatment on female reproductive output was examined using the number of eggs laid between the 1st and 6th days of the experiment. This continuous 6-day recording period is considered to be representative of female fecundity rates due to high positive correlation (r = 0.887, n = 41, p < 0.001) with lifetime fecundity in the stock population of this species (results from preliminary experiments). Female fecundity data were tested for normality using the Shapiro–Wilk test and for equality of variance using Bartlett’s test separately for each treatment level. Given that the normality and equality of variance assumptions were met, the mean number of eggs laid by females was compared between treatments with a t-test.
(iii) Effect of juvenile diet quality on female resistance to male-induced harm
To determine the effects of treatment on female resistance to male-induced harm, I used a full-factorial experimental design. Females from the high- and the low-quality food treatments were mated with males from both experimental procedures for three days. Then, the males were removed, and the females were allowed to lay eggs for the next three days. Female fecundity was estimated based on the total number of eggs laid by females over that period. The differences in the total number of eggs laid by females were analysed by two-way ANOVA, as the normality and equality of variance assumptions were met, and with type III sums of squares due to differences in the number of repeats (27-29 females per experimental group).
All statistical analyses were performed in R version 3.4.1.