Polygyny carries costs in both sexes in Trinidadian guppies

11 According to sexual selection theory, males should benefit more from mating with multiple 12 partners than females do, as male investment into offspring production is typically lower. For 13 females, empirical evidence indeed often shows diminishing returns or even costs of mating 14 multiply. For males, the assumption often seems to be “the more, the better” – i.e., a steady 15 increase of male reproductive success with mate number – but experimental tests of it are 16 rare. Here we used a laboratory experiment with Trinidadian guppies (Poecilia reticulata), 17 known for being promiscuous, to assess how pairing males weekly with 4 vs. 7 females 18 affects both sexes’ reproductive performance (n = 32 polygynous males and 170 monogamous 19 females). Increased polygyny delayed females’ reproductive onset by 9% and tripled their risk 20 of reproductive failure. High-polygyny males fathered offspring with 49% more females and 21 had 73% higher daily reproductive output. Y et, they needed 19% longer to initiate pregnancy, 22 and only accumulated more offspring than low-polygyny males after two months. This study 23 suggests that male mating performance is not unlimited. Especially when high extrinsic 24 mortality selects for fast reproduction, less polygyny might be advantageous, and the strength 25 of sexual selection perhaps more similar between the sexes than often assumed. 26 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 3

27 Sexual selection theory posits that males – i.e. the sex with the smaller, more numerous 28 gametes – benefit more from mating with additional partners than females do [1]. The slope 29 of reproductive success on mate number, called the “Bateman gradient”, is thus expected to 30 be steeper in males, and sexual selection consequently stronger [2-4]. Across the animal 31 kingdom, empirical evidence indeed supports these expectations [5]. 32 The costs and benefits of multiple mating are relatively well-studied in females. Polyandry 33 (i.e. female matings with multiple males) has been found to be common in a wide range of 34 taxonomic groups [e.g. 6, 7-9], despite diminishing returns or even costs as the number of 35 mating partners increases [e.g. 10, 11-13]. By contrast, polygyny, where one male mates with 36 multiple females, is generally assumed to benefit males at a cost to females [14, 15]. Even at 37 high mating rates, males are expected to gain from additional copulations without saturation, 38 because mating with more females should equal more offspring per male [2, 14, 15]. When 39 costs of polygyny to males have been studied, it was mainly with respect to increased male-40 male competition [e.g. 16, 17] or reduced amounts of paternal care given to offspring [e.g. 18, 41 19, 20]. 42 However, polygyny might be costly for males even when male-male competition and male 43 parental investment are absent. Although sperm has long been viewed as an unlimited 44 resource [2, 21], a male’s number of ejaculates is limited [22]. Empirical evidence of sperm 45 depletion, such as reduced ejaculate size and fertilisation rates after successive matings, exists 46 e.g. for Soay sheep [23], Medaka fish [24], and Lepidopterans [25]. Also ejaculate 47 components other than sperm can become depleted, depressing female fecundity [26]. Even 48 without sperm or ejaculate depletion, male mating performance can decrease over successive 49 matings [27]. Hence, if males spread themselves too thinly by doing a bad job of mating with 50 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 4 too many females, the costs to their females might translate into costs to themselves, and their 51 reproductive success might be lower than that of less polygynous males. 52 Here we test whether polygyny imposes costs on male or female reproduction in Trinidadian 53 guppies (Poecilia reticulata). Guppies are internally fertilising, live-bearing fish whose 54 promiscuous mating system makes them an excellent system for studying sexual selection 55 [28, 29]. Neither sex provides parental care, removing one of the primary costs of – and thus 56 limiting factors to – polygyny. We conducted a laboratory experiment, pairing males weekly 57 with 4 vs. 7 females in sequential monogamy. By assigning males a set level of polygyny, we 58 remove the confounding effect of male quality that, in observational settings, can prevent 59 lower-quality males from achieving higher degrees of polygyny, potentially masking its costs. 60 We measured female age at first reproduction and first litter size and computed male and 61 female daily and cumulative reproductive output. We found that increased polygyny was 62 costly for both sexes. High-polygyny males were slower and less successful at initiating 63 pregnancy, thereby delaying high-polygyny females’ reproductive onset and tripling their risk 64 of reproductive failure. 65 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 5

66 Experimental fish 67 The fish used here were first-generation (F1) laboratory-born offspring of 297 founders 68 collected from a high-predation stream (Guanapo River) in Trinidad, and brought to Lund 69 University, Sweden, in October 2024. They were housed in a temperature-controlled room 70 (26.5 ± 0.2ºC [mean ± SD]) in four independent recirculation systems with continuous 71 filtration, ultraviolet sterilisation, and a 12L:12D light cycle. Founder females were kept in 72 individual 1.1-litre tanks and checked daily for fry. F1 fry were reared in 1.1-litre tanks in 73 sibling groups (maximum ten fry per tank). At age 30-35 days, F1 juveniles were 74 anaesthetised, sexed, and moved to individual 1.1-litre tanks. All fish were fed ad libitum with 75 liver paste and live brine shrimp nauplii (Artemia spp.). 76 Mating trials 77 From the sexed F1 juveniles, we selected 170 females and 32 males. They had 75 founder 78 mothers, with 2.3 ± 1.0 F1 females per founder mother (range 1-6) and a different founder 79 mother for every F1 male. We assigned fish to one of two experimental levels of polygyny. At 80 the low level, a male was paired with four females (n = 18 males, n = 18 x 4 = 72 females), at 81 the high level, with seven females (n = 14 males, n = 14 x 7 = 98 females). We thus had 32 82 mating groups, each consisting of one male and either four or seven females, respectively. 83 Males had different founder mothers than their females. One individual initially presumed 84 female was later found to be male, and one female died early, reducing the number of females 85 in two high-polygyny groups to six, and the total female number to 168. 86 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript Within each mating group, we paired fish according to a two-week schedule, such that each 87 female was mated once a week, and males had regular recovery periods (Fig. 1). The schedule 88 alternated both the females’ mating order and day-night cycle of mating opportunities, to 89 equalise the likelihood of successful insemination. We set up mating groups in a staggered 90 fashion over 37 days (Supporting table 1). 91 Figure 1. Schedule of mating trials at two experimental levels of polygyny. Mating groups contained one male and either four females (low polygyny, n = 18 groups) or seven females (high polygyny, n = 14 groups). All fish were initially virgins. Females were paired with their assigned male once a week, either during daytime (approx. 8 hours) or overnight (approx. 16 hours). The schedule was repeated until females either gave birth or reached 100 days of age. Females were housed in individual tanks; males were moved between female tanks, with recovery periods in their own tank. Hence, matings happened in a sequentially monogamous way and always in the females’ home tank. 92 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 7 Females were mated first at age 42.7 ± 4.8 days (range 34-64) and males at age 51.3 ± 7.1 93 days (range 21-59). We ensured that all males were sexually mature, as indicated by a fully 94 developed gonopodium. Low-polygyny females were, on average, 4.2 days younger on their 95 first mating opportunity than high-polygyny females. In part, this difference (t = -6.2, df = 96 154.8, p = 6.3e-09) arose because fewer days are needed to mate one male with four rather 97 than seven females (see Fig. 1), but logistic difficulties contributed to it. To account for a 98 potential bias introduced by variation in female age at first mating, we included it as a 99 predictor in statistical models. 100 We mated females until they gave birth or reached 100 days of age without reproducing. In 101 four mating groups, the male died before all females either reproduced or turned 100 days old. 102 The resultant lack of mating opportunities for these 16 females (8 from high-, 8 from low-103 polygyny groups) could affect their reproductive performance. “Death of male” was thus a 104 predictor in statistical models. Non-reproductive females aged 100 days were kept isolated for 105 12 days. If by day 112 they still had not given birth, we declared them non-reproductive with 106 respect to their assigned male. To investigate whether they were infertile, we then paired them 107 with two additional males. These were unrelated to the females, the females’ first males, and 108 each other. None of them had previously been part of mating trials. Every week, we paired 109 non-reproductive females with their two males, sequentially, for 24 hours per male. Females 110 that were non-reproductive by the age of 165 days were considered infertile. 111 Measurement of phenotypic traits 112 We measured female age at first parturition and the number of offspring in first litters. Female 113 tanks were checked daily for fry. Fry born within one day of each other were considered to be 114 part of the same litter. We calculated female daily reproductive output by dividing first litter 115 size by the interval between a female’s first mating opportunity and first parturition. A male’s 116 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 8 daily reproductive output was the sum of the daily reproductive outputs of his assigned 117 females. We identified the mating opportunity a female’s pregnancy was likely initiated at by 118 subtracting the shortest observed gestation length (22 days) from her first parturition date, 119 followed by finding the closest mating opportunity before that date. 120 Statistical analysis 121 We used generalised linear mixed models (GLMMs) to analyse the interval between females’ 122 first mating opportunity and first birth, the size of the first litter, and female daily reproductive 123 output. The fixed effects were level of polygyny (factor: low vs. high), female age at first 124 mating, female pairing order in week 1 of mating trials, premature death of male (factor: no 125 vs. yes), and the recirculation system a female was housed in (factor: 1-4). We fitted random 126 intercepts for male identity and female’s mother’s identity, to account for non-independence 127 of data among females grouped by these factors. 128 We compared models fitted with different error structure to find the best-fitting one for each 129 trait. For the mating-to-birth interval, the best model had Gaussian errors, an identity link 130 function, and base-e log-transformed response values; for first litter size, Poisson errors, a log 131 link, and untransformed response values; for female daily reproductive output, Gaussian 132 errors, an identity link, and a square-root-transformed response. 133 We performed statistical analyses in R v. 4.4.2 [30]. GLMMs were fitted using “glmmTMB” 134 [31]. We tested the significance of random effects with log-likelihood ratio tests and assessed 135 model fit using DHARMa [32]. Scatterplots were prepared using “beeswarm” [33]. Values are 136 given as mean ± SD. 137 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 9

138 Effects of polygyny on female reproduction 139 Increased polygyny delayed females’ reproductive onset (Fig. 2a, Table 1). Females in the 140 high-polygyny treatment had 9.0% longer intervals between their first mating opportunity and 141 first birth (42.9 ± 11.4 days) than those in the low-polygyny treatment (39.4 ± 8.8 days). 142 Mating-to-birth intervals were also longer for younger females (Fig. 2b) and when a female’s 143 assigned male died prematurely (Table 1). They did not depend on the pairing order of 144 females, nor on their placing inside the laboratory (Table 1). Male identity explained a 145 significant amount of variation in mating-to-birth intervals, while the identity of the females’ 146 mother did not (Table 1). 147 High-polygyny females had larger first litters than low-polygyny females (Fig. 2c) but the 148 same daily reproductive output (Fig. 2d, Table 1). On average, first litters contained 30.8% 149 more offspring in high-polygyny groups (6.4 ± 3.7 vs. 4.9 ± 2.5; Fig. 2c). Given that litter size 150 was positively correlated with female age at first parturition (Pearson’s r = 0.49, t145 = 6.8, p = 151 2.04e-10), this difference likely arose from high-polygyny females being older and thus 152 presumably larger when producing their first litters. All other predictors of first litter size 153 were nonsignificant, except for the females’ mothers’ identity (Table 1). However, despite 154 their larger first litters, high-polygyny females did not have a higher daily rate of offspring 155 production: 0.15 ± 0.07 vs. 0.13 ± 0.06 offspring per day (Fig. 2d, Table). No other predictor 156 influenced that rate either (Table 1). 157 The high-polygyny treatment increased females’ risk of reproductive failure (Fig. 2e). Only 158 21 out of 168 females (12.5%) did not reproduce, but rates of non-reproduction were over 159 three times greater in the high- (17.7%) compared to the low-polygyny group (5.6%; 160 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 10 Pearson’s /g2031/g2869 /g2870 = 4.5, p = 0.0339; Fig. 2e). Notably, non-reproductive females were not infertile: 161 upon being paired with additional males, 20 out of 21 females (95.2%) gave birth. 162 Figure 2. Effects of experimentally increased polygyny on female reproductive performance. High-polygyny females needed more time to conceive than low-polygyny females (a) – an interval that also increased the younger a female was when mated for the first time (b). High-polygyny females had larger first litters (c) but equal daily reproductive output (d). They were, however, more likely to remain non-reproductive (e). In (e) the sample size is 72 and 96 females per group, respectively, and error bars are binomial standard errors (± 1 SE) computed according to Zar [34]. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 11 Effects of polygyny on male reproduction 163 At first glance, being more polygynous boosted males’ reproductive success. Males in high-164 polygyny groups fathered offspring with 49.4% more females (5.6 ± 1.3) than low-polygyny 165 males (3.8 ± 0.5; Wilcoxon rank sum test: W = 31, p = 9.33e-05; Fig. 3a). They also had a 166 73.0% higher total daily reproductive output (W = 38, p = 0.0005; Fig. 3b), receiving 0.83 ± 167 0.31 compared to 0.48 ± 0.14 offspring per day from all their females combined. This was 168 despite a 3.2 times higher fraction of non-reproductive females among high- than low-169 polygyny males (17.5 ± 18.7% vs. 5.6 ± 13.7%; W = 71, p = 0.0161). 170 However, when taking the aspect of time into account, increased polygyny was costly for 171 males. High-polygyny males needed 18.5% more mating opportunities – and thus 18.5% 172 more time – to initiate pregnancy in a female (3.7 ± 1.7 vs. 3.1 ± 1.3 mating opportunities; 173 Welch two-sample t-test: t142.3 = -2.4, p = 0.0201; Fig. 3c). Additionally, the benefits of 174 increased polygyny accrued relatively late. The total number of offspring a high-polygyny 175 male sired only surpassed that of low-polygyny males 65 days after a male’s first mating (W = 176 36, p = 0.0007); after 35 and 50 days, respectively, the difference was nonsignificant (both p > 177 0.34; Fig. 3d). 178 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 12 Figure 3. Effects of experimentally increased polygyny on male reproductive performance. High-polygyny males gained offspring from a higher absolute number of females (a) and had a higher daily reproductive output (b). However, they needed more mating opportunities to impregnate their females (c) and only accumulated more offspring than low-polygyny males 65 days after their first mating opportunity (d). 179 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 13 Mating-to-birth interval No. offspring in first litter Daily reproductive output Fixed effects: Estimate (S.E.) z-value p-value Estimate (S.E.) z-value p-value Estimate (S.E.) z-value p-value (Intercept) 4.34 (0.24) 18.00 < 2e-16 1.59 (0.46) 3.44 0.0006 0.28 (0.07) 3.78 0.0002 No. female partners (4 vs. 7) 0.22 (0.06) 3.53 0.0004 0.30 (0.11) 2.61 0.0091 0.02 (0.02) 0.91 0.36 Female age at first mating -0.02 (0.01) -2.78 0.0054 0.00 (0.01) 0.04 0.97 0.00 (0.00) 1.00 0.32 Female pairing order -0.02 (0.01) -1.56 0.12 -0.03 (0.03) -1.13 0.26 -0.00 (0.00) -0.38 0.70 Death of male (no vs. yes) 0.20 (0.09) 2.25 0.0242 0.18 (0.17) 1.03 0.30 0.00 (0.03) 0.01 0.99 Recirculation system 2 0.08 (0.06) 1.34 0.18 -0.08 (0.14) -0.54 0.59 -0.03 (0.02) -1.30 0.20 Recirculation system 3 -0.04 (0.06) -0.73 0.47 0.02 (0.13) 0.12 0.90 0.01 (0.02) 0.32 0.75 Recirculation system 4 -0.00 (0.05) -0.08 0.94 0.13 (0.13) 1.02 0.31 0.01 (0.02) 0.68 0.50 Random effects: Var (S.D.) /g2257/g2778 /g2779 p-value Var (S.D.) /g2257/g2778 /g2779 p-value Var (S.D.) /g2257/g2778 /g2779 p-value Male identity 0.014 (0.119) 13.41 0.0003 0.019 (0.139) 1.72 0.19 0.000 (0.000) 0.0 1.00 Female’s mother’s identity 0.007 (0.084) 2.39 0.12 0.054 (0.233) 7.23 0.0072 0.000 (0.000) 0.0 1.00 Residual 0.030 (0.174) 0.008 (0.091) Table 1. Outcome of GLMMs for reproductive traits of female guppies whose partners experienced either a lower or higher level of 180 polygyny. Number of females = 147 (lower level of polygyny: 68, higher level of polygyny: 79). 181 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 14

182 Our findings indicate that polygyny imposes direct costs on female reproduction, and that 183 these translate into reproductive costs to males. Female guppies in the high-polygyny 184 treatment conceived more slowly and were three times more likely to remain non-185 reproductive. High-polygyny males, for their part, were slower at initiating pregnancy, had 186 higher reproductive failure, and only outperformed low-polygyny males after two months. 187 While females in our study undoubtedly faced higher costs than males given their inability to 188 seek compensatory matings elsewhere, we show that polygyny carries direct fitness costs for 189 male guppies, too. We can rule out that high-polygyny males were accidentally paired with 190 less fertile females, as non-reproductive females successfully reproduced with other males. 191 Costs of polygyny are rarely considered in species without parental care. We are not aware of 192 prior work in guppies that quantified how polygyny affects male or female reproduction. But 193 sexual interactions have been found to be costly to males in other respects. Large guppy males 194 show higher mortality when housed in mixed-sex compared to all-male groups [35]. Increased 195 male reproductive effort decreases males’ foraging rates, growth, and survival [36, 37]. Sperm 196 production is plastic in guppies, decreasing under food limitation [38] and increasing in the 197 presence of females [39-41], suggesting that its costs are substantial (although see [42]). Data 198 on how male mating history affects female reproductive performance are sparse; guppy 199 females primarily experience polyandry [28, 29], where matings with multiple males likely 200 mitigate some of the costs associated with polygyny. 201 The delayed conception and first birth of high-polygyny females suggest that polygyny might 202 be costly especially when extrinsic mortality is high. When mortality risks are high, anything 203 that lowers a male’s ability to fertilise eggs quickly should be selected against. After all, he 204 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 15 might die before encountering her again, or she might die before conceiving or before giving 205 birth. Guppies inhabiting downstream river sections, where population densities are kept low 206 by intense predation [43, 44], face substantially higher monthly mortality rates than their 207 conspecifics in upstream sections [45]. It might thus be expected that the optimum degree of 208 polygyny is lower in high- than in low-predation populations – thus adding another facet to 209 the many ways in which sexual selection dynamics differ between high- and low-predation 210 guppies [46-49]. Also irrespective of extrinsic mortality it might be advantageous to 211 reproduce early. There is evidence across various species that earlier-born offspring have 212 higher fitness than later-born offspring, for example in lizards [50], roe deer [51], and 213 mosquitofish [52]. Moreover, in a growing population, females gain more fitness from 214 offspring born earlier, as reproducing earlier allows them to contribute to population growth 215 sooner [53-55]. 216 Several limitations of our study should be addressed. First, it cannot pinpoint the mechanism 217 behind the delayed reproductive onset of high-polygyny females, which might be mating 218 fatigue, sperm limitation, or both. Future work could address this, perhaps by observing 219 mating behaviour or measuring sperm counts. Second, our experimental design simplified 220 natural conditions in three important ways: by imposing monogamy on females, considering 221 first litters only, and eliminating male-male competition. In the following, we explain why we 222 believe these simplifications do not interfere with our study’s goal and instead make our 223 estimated costs of polygyny conservative. Females being monogamous is an unlikely 224 scenario; in the field, both courtship displays and sneak mating attempts are common [28] and 225 levels of multiple paternity typically high [29]. Consequently, females that failed to get 226 pregnant quickly from their high-polygyny male would likely be inseminated by other males, 227 thereby increasing male costs of polygyny. The restriction to first litters might, by contrast, 228 not be overly unrealistic: under natural conditions, females might not live to produce a second 229 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 16 litter or might sire it with sperm from another male. Nonetheless, had our study included 230 second litters, then many low-polygyny females would have produced second litters before 231 high-polygyny females had their first, reducing the benefit of increased polygyny even after 232 65 days. Finally, eliminating male-male competition allowed us to quantify effects of 233 polygyny independent of variation in males’ competitive ability before or after copulation. 234 This, again, makes our study conservative: if a more polygynous mating strategy does not pay 235 off in a benign, competition-free scenario, it might not under more realistic conditions either. 236 Future experiments or modelling work could address costs of polygyny in the context of 237 competition between males. 238 An unexpected finding is the apparent bimodal distribution of the mating-to-birth interval, 239 visible as two parallel, downward-facing clouds in Fig. 2b. In the lower cloud, mating-to-birth 240 intervals were relatively short (~22-48 days); in the upper cloud, relatively long (~40-75 241 days). Within clouds, interval length depended on female age at first mating, decreasing the 242 closer that age was to about 50 days, when females apparently reached sexual maturity. Given 243 that guppy females, unlike males, lack external signs of maturity [56], identifying that age in 244 females is an interesting result in and of itself. If a female’s first mating happened around age 245 50 days and was successful, the mating-to-birth interval was shortest (min. 22 days), 246 consisting only of the gestation time. However, if the mating was (supposedly) unsuccessful, 247 the interval increased to about twice that length (~42-53 days). At this point, we do not know 248 whether this bimodality is an artefact of pairing females discontinuously (i.e. once a week); 249 more data will need to establish its veracity. However, it is faintly reminiscent of a 250 spontaneous ovarian cycle, for which circumstantial evidence exists in other poeciliids [57]. 251 In conclusion, our study shows that costs of polygyny are borne by females as well as males. 252 This implies that sexual selection operates on females, too – if polygyny comes with 253 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 17 reproductive costs in males, then presumably there is male mate choice. The appreciation of 254 sexual selection being important in females is growing, mainly driven by widespread positive 255 associations between polyandry and female reproductive success [58, 59]. We contribute to 256 this paradigm shift by highlighting that male mating performance is not unlimited, and the 257 association between polygyny and male reproductive success perhaps not unconditionally 258 positive. 259 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 18 Ethics 260 This research was approved by the Ethical Committee on Animal Experiments, Lund/Malmö, 261 Sweden (Dnr 5.8.18-13832/2024). 262 Data accessibility 263 All data and R code associated with this manuscript will be available as part of the Supporting 264 Information following manuscript acceptance. 265 Declaration of AI use 266 We have not used AI-assisted technologies in creating this article. 267 Author contributions 268 T.V .D.W.: investigation, formal analysis, writing – original draft. 269 F.D.G.: investigation, writing – review & editing, supervision. 270 T.P.: formal analysis, writing – review & editing. 271 A.F.: conceptualisation, formal analysis, data curation, writing – original draft, writing – 272 review & editing, visualisation, supervision, funding acquisition. 273 Conflict of interest declaration 274 We declare we have no competing interests. 275 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 19 Funding 276 This research was funded by Lund University (Start-up and Infrastructure Grant to A.F.) and 277 by the Royal Physiographic Society of Lund (Grant to A.F.). 278

279 We thank the core scientists of The Guppy Project, David Reznick, Joe Travis, Ron Bassar, 280 and Tim Coulson, for letting us stay in their field station in Trinidad and using its 281 infrastructure for capturing the founders of our laboratory population. We especially thank 282 Ryan Mohammed for his invaluable help with export permits and logistics, and Ignacio Paulin 283 and Gabriela Jeliazkov for help in the field. In Lund, we thank Madalena Madeira for help 284 with fish maintenance, and Olof Berglund for comments on an earlier version of the 285 manuscript. 286 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 20

287 1. L eh t on en, J ., Ba t eman grad ients from firs t principles . Na tur e Co mmunic a ti o ns, 2 022. 288 13(1). 289 2. Ba t e man, A. J ., Intra- s e x ual s election i n Dr os ophila. Her edi ty , 1948. 2 (3): p. 349- 368. 290 3. Arnold, S. J ., B at ema n's principl es and the measuremen t of s e x ual s election in pla n ts 291 and an imals . The A meric an N a tu r alist, 199 4. 144: p. S126- S14 9. 292 4. Henshaw , J ., A . K ahn, and K. F r itz sche, A rigor o us c ompar is on of se x ual sele c t ion 293 i n dex es via simulations of dive r s e mat ing sys t ems. P r oceeding s of t he N a tion a l 294 Ac a d emy o f S cience s of the U n i t ed Sta t es of A mer i c a, 2016. 113(3) : p. E 30 0-E308. 295 5. Janick e, T ., et al., Darwinian se x r oles c onfirmed across t he animal k ingdo m . Sc ien ce 296 Adv anc es, 2 016. 2 (2) : p. e15 00983-e150098 3. 297 6. G r i f fith, S., I.P . F . Owen s, and K . Thuman, Extr a pai r pa t ern i t y i n bir ds : a r ev iew of 298 i n t erspec ific variation and a d apt ive f u nction. Mol ecular E c ology , 2002. 11 (1 1 ): p. 299 2195-2212. 300 7. Bürkli, A. and J . Jo k ela, Increase i n mu l t iple p a t ernity across t he reproductive lif es pa n 301 i n a sperm-st oring, hermaphr odi tic fr esh wat er snail. Molecular E c ology , 20 17. 26(19) : 302 p. 5264-5278. 303 8. S o uls bur y , C., Genetic p at t erns of pa t e r nity and t es t es size in mammals. Pl os O ne, 304 2010. 5 (3): p. e95 81. 305 9. U ller , T . and M. O ls son, Multiple pa t er nity in r ept iles: patt ern s and pro c es s e s. 306 Molecular E c ology , 2008. 17 (11) : p. 2 566-2580. 307 10. L ang e, R., et al., F emale fitnes s opt imum at i n t er media t e mati ng r a t es under 308 traumatic mat ing. Plos O ne, 201 2. 7 ( 8) : p. e43234. 309 11. Y an, J .L., M.L. D obbin, and R. Duk a s, Se x ual conflic t and s e x ual network s in bed bugs: 310 the fit nes s c os t of tr aumatic ins eminat ion, f emale av oi da nc e and ma le mat e choice. 311 Pr oceeding s of t he R oy al S o ciety B , 2 024. 291(2 0 27): p. 20 232808 . 312 12. Huchar d, E., et al., Con v enience polyandr y or con venience poly gyn y? cos tly se x under 313 f emale cont rol i n a promiscuous pr i mat e. Pr oceedings of the R oy al So c iet y B , 2012. 314 279(1732): p. 1 371- 137 9. 315 13. Maklak ov , A.A., T . Bilde, and Y . Lubin, Se xua l c on flict in t he wild: Elev at ed m ating ra t e 316 reduce s f emale lif etime repr oductive s uc ce s s . The A meric an N a t ur alist , 20 0 5. 317 165(Supplemen t) : p. S38-S45. 318 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 21 14. S h ust er , S.M. and M. J . W ade, Mating s y st ems and s trat e g i es . Mon ogr aph s i n Beh avior 319 and E colog y . 20 03, Princet on, New Je r se y , U S A : Princ et on Univ er s it y P r ess. 320 15. Emlen, S. T . and L . W . Oring, E c ology , s e x ual sel ection, and t he e volut i on of m ating 321 sy st ems . Scien c e, 19 77. 197( 4300): p. 215-223. 322 16. Jungwirth , A ., et al., P oly gyn y aff ects pat ernal care, but not sur vival , pair s s t abi lity , 323 and grou p t e nur e in a c ooper a tive cichlid. Beha v ior al Ec ology , 2016 . 27(2): p. 592-600 . 324 17. Dubuc, C., A. Ru i z-Lambides, and A. W idd i g , V ar ianc e in male lif etime r e p r oductive 325 s uc ce s s an d es timat ion o f the degree of polygyn y in a primat e. Beha v ior al Ec ology , 326 2014. 25(4): p. 878-8 89. 327 18. S chlich t, E . and B. K empenaer s , Origin and o ut come of soci a l po lygyn y in t he blue t it. 328 Ar dea, 2021. 109( 1): p. 91-11 8. 329 19. Desjar dins , J .K., et al., Co st s and benefits of polygyn y i n t he c ich lid N eolamp r ologus 330 pulc her . An i mal B eha v iour , 20 08. 75(5 ): p. 1771-1779 . 331 20. Mueller , S.D ., et al., R educed fit nes s of secon dary f emales in a poly gynous s peci es: a 332 32-yr s tudy of Sav an na h s par rows. B eha v ior al E c ology , 2 02 5. 36 (1): p . ar ae 093. 333 21. Dawkins, R., The sel f is h gene . 197 6, New Y ork: O xf or d Univer s it y Pr ess . 334 22. Dews b ury , D .A., Ejaculat e Cos t and M ale Choice. The Americ an Nat ur alist , 1982. 335 119(5): p. 601-6 10 . 336 23. Pr es t on, B. T ., et a l., D ominan t r ams lose out by s per m deple t ion. N a t ur e, 2001. 337 409(6821): p. 6 81- 682. 338 24. K ondo , Y . , M. K ohda, and S. Awat a, Male medak a con tinue t o ma t e with f e males 339 des pit e sperm depletion. R Soc Open Sc i, 2 025. 12 (1) : p. 2416 68. 340 25. T or r es-Vila, L .M. and M.D . J ennions, Male matin g his t ory and f emal e f ecu ndity i n the 341 L epidopt e r a: do male virg ins mak e be t t er p art ners? B e h a vi o r a l E c o l o g y a n d 342 S o c iobiology , 2 004. 57 (4): p. 318- 326 . 343 26. Abr aham, S., et al., Mal e ac ces sory gland dep letion i n a t ephri t id fly a f f ec t s f ema l e 344 f ec undit y i ndepen den tly of sperm d e plet ion. Beha v i or al Ec ology and So ci o biology , 345 2020. 74(5). 346 27. Abr egú, D . A., A. V . P er et ti , and M. G on zále z , Male perf ormanc e and asso c iated c os ts in 347 s uc ce s si ve sex ual enc ount ers in a poly gynous web wol f s pider . Act a E th ologi c a, 201 9. 348 22(3): p. 175-1 86. 349 28. Houde, A .E., S e x, co l or , an d mat e c ho i ce in guppies , ed. J .R. K r ebs and T .H . C lut ton-350 Br ock. 19 97, Princet on, New Jer se y : P r incet on Univer s ity Pr ess. 351 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 22 29. N e f f , B.D ., T .E. Pit cher , and I. W . Ramn ar i n e, In t er- popul a t ion vari atio n in m ult i pl e 352 pat erni t y and r epr oductive s k ew in the guppy . Molecula r E cology , 2 008. 17 (12): p. 353 2975-2984. 354 30. R Cor e T eam, R: A Lan gua ge and E n vir onmen t f or Statis tical Compu ting. 20 24, R 355 Fou nda tion f or S t a t ist ical Comp uting: Vienna, Aus tr ia. 356 31. Br ook s, M.E ., et al., glmmTMB balanc es s peed and fle xibili ty among pack a ge s f or 357 zero -infla t ed genera liz ed l inear mixed model ing. The R Journal, 2017. 9 (2) : p. 378-358 400. 359 32. Hart ig , F ., DH ARMa: residual dia g nosti cs f or hierar c hical (mult i -lev el/mixed) 360 regres s ion mo de ls . R pac kag e v er sion 0.4.1 ed, 20 21. 361 33. Ek lun d, A., bee s war m: the bee s warm p l ot , an al t ernativ e t o s tr ipc har t. 202 1. 362 34. Za r , J .H., Sampling a bin omial popu lat ion , in Bi os t at i s tical A nalys is , S . Fi sher and S.L. 363 S n av ely , Ed i t or s. 199 6, Pr en tice-H all , In c .: Ne w Jer se y , Upper Saddle Riv er , N J , U SA. p. 364 523. 365 35. Jerr y , M. and C . Br own, Fitnes s c o st s of se xual har as s men t - t he price of pe rsua s ion. 366 Et h o l o g y , 2 01 7. 123 (3): p. 242-250. 367 36. Jor dan, L. A. and R. C . B r ook s, The lif et i me c os t s of i n c r eas ed male repr oduc t i ve ef f ort: 368 Courtship, c opula tio n and the Coo lidg e eff e ct J ournal of E volut ion a r y Biol ogy , 20 10. 369 23(11): p. 2403- 2 409. 370 37. Mille r , L .K. and R . Br ook s, The eff ec t s of geno t y pe, age, and soc ial en viron me n t on 371 male ornamen tat io n, matin g behavi o r , and at tractivenes s. E v olution, 2005. 59(11) : p. 372 2414-2425. 373 38. G a sparini, C., et al., P at t e r n of i nbreeding de pr e s sion, conditi on depen dence, and 374 addit ive ge n e t ic variance in T rinid adi an guppy ejac u lat e trait s . E c ology & E v olut ion, 375 2013. 3 (15): p. 4940- 495 3. 376 39. Bo zyns ki, C . C . and N .R. Liley , The eff ec t of f emal e pr es enc e on s per miation, and of 377 male s e x ual ac t iv it y on "ready" s perm i n t he ma le guppy . Animal Beha v iou r , 2003. 378 65(1): p. 53-58. 379 40. Ca t t elan, S., et al., The e f f ec t of s per m producti on an d mat e avai la bi l it y on pat t erns of 380 alt erna t i ve mat i ng tactic s in the gu pp y . Animal Beha v iour , 2016. 112: p. 10 5- 110. 381 41. G a sparini, C., A. V . P er et t i, and A . Pilastr o , F e male pr es en ce influenc es s per m vel ocity 382 i n t he g uppy . B i o l ogy Let t er s , 20 09. 5 ( 6 ): p. 792-794. 383 42. Magris, M., et al., Quic k-change art ists: Ma l e guppies pay no cos t t o r e peatedly adjus t 384 their se xual s trat egie s. Beha vior al E c ology , 20 18. 29(5) : p. 1113-11 23. 385 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 23 43. G illiam, J .F ., D .F . F r a ser , and M. Alkins-K oo , Str uc t ure of a tr o pic al s tream fis h 386 community: A ro le f or biot ic int eracti ons. E colog y , 1993. 74(6): p. 185 6-18 70 . 387 44. R e z nick , D ., M. J . Butler , and H. R o dd, L if e - hist ory ev olut ion i n guppie s. VII. T he 388 compa rative eco logy of hig h- and l ow - predat ion en vironme n ts . The A merican 389 N a tu r alist, 2001 . 157 (2): p. 1 26-140. 390 45. R e z nick , D .N., et al., L if e-his t or y evolution i n gu ppies (P oeci lia r e t ic ulat a) 6. 391 Diff er ential mort al i t y as a mechanis m f or na tur al s election. E volut ion, 1996. 50(4) : p. 392 1651-1660. 393 46. Endler , J . A., A pr edat or's vie w of animal c ol or pa t t erns. E v olut ionary Bio logy , 1978. 394 11: p. 319-36 4. 395 47. Millar , N .P . , et al., Dis en t ang ling the sel ective f act ors that ac t on ma le colo ur i n wild 396 guppies . Oik o s, 2006 . 113(1) : p. 1-12 . 397 48. S c hwart z , A. K. a n d A.P . H endry , A t es t f or the p aralle l co- ev olution o f male c o l ou r an d 398 f emale pref erence in T r inidadi an gu p pies ( P oe c ilia reticula t a) . Ev olutionary E co l o gy 399 R es ear c h , 2007 . 9 (1 ): p. 71-90. 400 49. El g ee, K.E. , et al., Geographic v ar iati o n in sperm tr aits r eflec t s pr edation risk and 401 natural r at es of mult iple p at erni t y in the guppy . J ournal of E volut ionar y Biol ogy , 2010. 402 23(6): p. 1331-1 33 8. 403 50. Wap s tr a, E., et al., G iving offspring a head s t ar t in li f e: f iel d and experime n t al 404 ev idence f or s ele ction on ma t ernal ba s king b ehav iour in l iz ards. Jour nal of 405 E v olutionary Biol ogy , 20 1 0. 23 (3): p. 6 51-657. 406 51. Couls o n, T . , et al., Mod e ling ad apt iv e and nonadap tive respons e s of popula tions t o 407 envir onmen t al change. The A meric an Nat ur alis t, 2017 . 190(3): p. 3 13- 336. 408 52. R e z nick , D ., et al., On the vi r tue of bei ng the f ir st born: The inf luence of da te of birth 409 on fi tnes s in the mosquit of ish, Gambu s ia aff in is . Oi k o s, 2 006. 114 (1) : p. 135 -147. 410 53. F r ank , S .A ., N at ural se lection at multi ple s cale s . E v olution, 2025. 79( 7): p. 1 166-1184. 411 54. S t earns, S . C ., The e volution o f li f e his tories . 19 92, N e w Y ork : Ox f or d Univer s ity Pr ess, 412 U SA. 264. 413 55. Charleswo rth, B., E volution in a ge-s tr uc t ured pop ula tions. 2 ed. 1994, Ne w Y ork: 414 Cambridg e Univer s it y Pr ess. 415 56. P ot t er , T ., et al., E n v ir onmen t a l chang e, if unaccount ed, pr eve n t s det ection of cr y pt ic 416 ev olut ion in a wild popu lati on . Th e Americ an N a t ur alist , 2021. 197(1): p. 2 9-46. 417 57. S iciliano , M. J ., E vidence f or a s p on t an eous o v ar ian c ycle in f i sh of the ge nus 418 Xiphophor us. Biologic a l Bullet i n , 1972. 142(3): p. 480-4 88 . 419 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint van der Walle et al. – Manuscript 24 58. F r omon t e il, S., et al., Se xual selec t i on in f emal es and the evolut ion o f poly a ndr y . PLoS 420 Biology , 2023. 21(1 ): p. e3001 916. 421 59. T ang-Martine z, Z ., R ethinking Bat ema ns principles : Ch al lenging per s is t en t myths of 422 se x ually r el u c t an t f emales an d pr omisc uous males . J ournal of Se x R es ear c h , 2016 . 423 53(4-5): p. 532- 559. 424 425 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 10, 2026. ; https://doi.org/10.64898/2026.04.07.716995doi: bioRxiv preprint