A population of ~300 ruffs is housed at the Max Planck Institute for Biological Intelligence, Seewiesen, Germany (for details see Supplementary Methods S1). The population was founded in 2018 with 23 ruffs obtained from Dutch breeders, 5 ruffs from German breeders and 194 ruffs from Simon Fraser University, Vancouver, Canada. The latter were individuals from a captive-bred population founded with 110 ruffs hatched from wild eggs collected in Finland in 1985, 1989 and 1990, plus two Faeder males brought from the Netherlands in 2006 (Lank et al., 2013, 1995). Each individual of the population has been genotyped for its morph using a set of six single-nucleotide polymorphism markers located in the inversion region (Giraldo-Deck et al., 2020).

In May and June 2021, we collected sperm by abdominal massage (see a detailed protocol of sperm collection – including video - and sample preparation) or by electro‐stimulation (Lierz et al., 2013; Bulla et al., 2024). The length (15 mm) and diameter (4 mm) of the electro-stimulation probe, as well as the electric current and the number of electric impulses was adapted to the sampled individuals. The probe was inserted into the urodeum of the cloaca to stimulate the ampullae ductus deferens using three 1 s electric current impulses. The voltage was increased gradually from 0.09 V to a maximum of 1.0 V until contractions of the cloaca and the muscles of the tail were observed; each impulse was followed by a 2-3 s break. In June, we only sampled sperm using abdominal massage. The morphs were sampled in a haphazard order to avoid a potential confounding effect of sampling order. In both months, we attempted to obtain sperm by abdominal massage from all males. We obtained at least one ejaculate sample from 92 males (59 Independents, 25 Satellites and 8 Faeders).

Ejaculates (~0.5–3μl) were pipetted from the cloaca and immediately diluted and gently mixed in 50μl of preheated (40°C) Dulbecco’s Modified Eagle’s Medium (Advanced D-MEM, Invitrogen™). To record sperm swimming speed, we then pipetted an aliquot of 2.5μl onto a standard 20μm two-chamber count slide (Leja, The Netherlands) placed on a thermal plate (Tokai Hit, Tokai Hit Co., Ltd.) kept at 40°C. When sperm densities on the slide were too high, we took a new aliquot and further diluted the sample. We then recorded sperm movements. By limiting the time between ejaculate collection and recording, we avoided that changes in sperm swimming speed over time (Cramer et al., 2016) would influence the results. For morphology measurements, we pipetted an aliquot of 20μl into 50μl of a phosphate buffered saline solution (PBS; Sigma P-4417) containing 1% formalin.

For each sperm sample we recorded sperm swimming speed at 25 frames per second for approximately 45s in eight different fields of the Leja slide under a 100x magnification, using phase contrast and a digital camera (UI-1540-C, Olympus) mounted on a microscope (CX41, Olympus) fitted with a thermal plate (Tokai Hit, Tokai Hit Co., Ltd.) kept at a constant temperature of 40°C. We confirmed that this temperature was appropriate based on cloacal temperature measurements of five female ruffs. For each female, we inserted a high-resolution temperature-probe connected to a MSR145 data logger (MSR Electronics GmbH, https://www.msr.ch/en/) into the cloaca, and logged the temperature every 5s for 3 min, i.e., until an asymptotic value was reached. Cloacal temperatures ranged between 40.6°C and 42.3°C, and are similar to the body temperatures measured in shorebirds (Charadriiformes; range: 40.4-41.8°C, N = 10 species) and in passerines (39.2-43.5°C, N = 16 species; (McNab, 1966).

A few weeks later, a single person (JA) analyzed each recorded field using the CEROS computer-assisted sperm analysis system (Hamilton Thorne Inc.). JA visually inspected the tracked objects, excluded non-sperm objects and static spermatozoa from the analysis (Cramer et al., 2016; Laskemoen et al., 2010; Opatová et al., 2016), and noted the quality of the recording (e.g., presence of feces). We recorded a median of 192 sperm cells per sample (mean = 204, range: 5 – 562; N = 134 recordings from a total of 92 males, 46 recorded in May and 88 in June; Independent: mean = 202, N = 83; Satellite: mean = 205, N = 37; Faeder: mean = 218, N = 14), which represented a median of 78% of motile sperm in a sample (mean = 69%, range: 12 – 92%; Independent: mean = 70%; Satellite: mean = 69%; Faeder: mean = 61%). For each sample the software estimated the mean curvilinear, straight-line, and average-path velocity (for details see Supplementary Methods S2, Figures S1, S2; for within-male seasonal repeatability see Supplementary Figures S3, S4, Table S1).

Within 5 days of sample collection, we pipetted (i) 20μl of the PBS-1% formalin fixed sample into 50μl of 5% formalin for long-term storage and (ii) smeared 10μl onto a microscope slide, and let it dry at room temperature or on a heating block set at 37°C. The next day, each slide was rinsed gently with deionized water to remove salt crystals and dried at room temperature.

To each dried slide we added a drop of Bisbenzimide (Hoechst 33342; Molecular Probes) to stain the sperm nucleus, and a drop of MitoTracker™ Green FM Dye (Invitrogen™ M7514) to stain the sperm midpiece (mitochondria; Figure 2 ). Within 48h of staining, we inspected the microscope slides under 200x magnification with a Zeiss Axio Imager.M2 light microscope fitted with a Zeiss Axiocam 512 color camera (12 megapixel; 4250 × 2838, pixel size of 3.1μm × 3.1μm) and DAPI 465nm and green 519nm filters using Zeiss ZEN blue 3.1 imaging software. For each male we used a single sample (for 30 males from May, for 62 males from June) and photographed at least 10 intact, normal-looking spermatozoa under 400x magnification (objective size 40, ocular size 10). We then selected the 10 best, single-sperm images per male for measurements. We then randomized and renamed all pictures, such that we measured sperm blind to the morph and the identity of the individuals.

For each sperm, we measured the length of the acrosome, the nucleus, the midpiece and the tail to the nearest 0.1μm using the open software Sperm Sizer 1.6.6 (McDiarmid et al., 2021; https://github.com/wyrli/sperm-sizer; see our Sperm Sizer protocol (Bulla et al., 2024). To assess within- and between-observer repeatability of Sperm Sizer measurements, we selected 40 single-sperm pictures and remeasured them, once by the same, once by a different person. The measurements were highly repeatable, both within- and between persons (mean = 97%; Supplementary Table S2). Nevertheless, all measurements used in the analyses were taken by the same person (KT). Each measured sperm and sperm part were numbered and referenced in the database and hence can be linked to the original picture as well as to the pictures of each part generated by Sperm Sizer, which contain the lines and measurement values (see Data in Bulla et al., 2024). This ensures transparency and allows re-measurement of the same sperm by the same or a different person.

We calculated (i) total sperm length as the sum of all parts, (ii) head length as the sum of acrosome and nucleus length, (iii) flagellum length as the sum of midpiece and tail length, (iv) relative midpiece length as midpiece length divided by total length and (v) relative flagellum length as flagellum length divided by total length. We then computed coefficients of variation within each male for each absolute trait as the standard deviation divided by the mean. Note that estimating coefficients of variation from only five sperm per male gave similar values and previous studies from different species also used ten sperm per male to estimate the coefficient of variation in sperm traits (Kleven et al., 2008; Lifjeld et al., 2010).

In the Supplement, we report correlations between sperm components (Supplementary Methods S3, Figure S5), and estimates of within-male repeatability (Supplementary Figure S3, Table S1).

All analyses were performed in R (R-Core-Team, 2022) using the ‘lm’ function to fit linear models and the ‘lmer’ function of the ‘lme4’ R-package to fit linear mixed-effects models (Gelman and Hill, 2007; Gelman and Su, 2021). We used the ‘sim’ function from the ‘arm’ R-package and a flat prior distribution (Gelman and Hill, 2007; Gelman and Su, 2021) to create a sample of 5,000 simulated values for each model parameter (i.e. posterior distribution). We report effect sizes and model predictions as medians, and the uncertainty of the estimates and predictions as Bayesian 95% credible intervals represented by the 2.5 and 97.5 percentiles (95% CI) from the posterior distribution of 5,000 simulated or predicted values. Estimates with credible intervals that include zero are statistically unclear (Dushoff et al., 2019). We graphically inspected the goodness of fit, and the distribution of the residuals. Unless stated differently, plots were created with ‘ggplot’ function from the ‘ggplo2’ R-package (Wickham, 2016).

To investigate whether sperm traits differ between the morphs we fitted linear models with each sperm trait as a response variable, and with male morph as a predictor (three-level factor). For analyses of velocity, we primarily used June values because the breeding season was at its peak in June and we obtained data from all but four males (for those four we used May values). We added the number of tracked sperm cells per sample (ln-transformed) as a covariate to the model to control for potential effects of sperm density or sample quality. For analyses of morphology, we used average male values. Alternative models using all velocity recordings, using individual sperm measurements, or controlling for sampling month or aviary gave similar results (Supplementary Figure S6). Because correlations between sperm traits and inbreeding were weak (mean r = -0.09, range: -0.26 to 0.07; Supplementary Figure S7, Methods S4), we did not include the inbreeding coefficient in the main models. Furthermore, the results were not confounded by relatedness (estimated based on microsatellite genotyping data) of the sampled individuals (see Supplementary Methods S4, Table S3, Figure S8).

To investigate whether sperm morphology explained variation in sperm swimming speed, we fitted a set of linear models with each velocity measure as the dependent variable (using predominantly June values), and with the male average of each sperm morphology trait as a predictor. All models were controlled for the number of tracked sperm (ln-transformed) and morph (three-level factor). As the relationship between sperm swimming speed and morphology might be non-linear (Bennison et al., 2016), we also fitted a set of linear models that included the quadratic term of each morphology trait (2^nd order polynomial). We then tested whether the quadratic term improved the model fit, using Akaike’s Information Criterion corrected for sample size (Anderson, 2008), with the ‘AICc’ function from the ‘MuMIn’ R-package (Bartoń, 2022). Because this was not the case for any trait (Supplementary Figure S4), we report results from the simple model without the quadratic term. To investigate whether the effect of one sperm component (head, midpiece and tail length) was influenced by the other components, we also ran a multivariate model that contained the three sperm components as predictors. Correlations between the three predictors were low (Supplementary Figure S5). The effect sizes and uncertainties for head, midpiece and tail length were similar in the multivariate and the univariate models (Supplementary Figure S9) and we report the results from the univariate models in the main text.

Sperm swimming speed in vitro in a standard medium, measured as curvilinear velocity, did not differ between the three morphs ( Figure 3 , Table 1 ; Supplementary Figure S6; Bulla et al., 2024), but measured as straight-line or average-path velocity, Faeder sperm moved slower than the sperm of both Independents and Satellites, with the effect being clearer for the straight-line velocity (Faeder relative to Independent estimate [95%CI]: –0.62 [-1.24 - -0.02], relative to Satellite: -0.78 [-1.42 - -0.13]) than for the average-path velocity (-0.36 [-0.89 - 0.16] and -0.46 [-1.03 - 0.09]; Table 1 ). Note, however, that the posterior distribution of the parameters indicates a 92% probability for Faeder relative to Independent and a 95% probability for Faeder relative to Satellite that the true mean difference in the average-path velocity is less than zero. Independents and Satellites did not differ in sperm velocity ( Figure 3 ; Supplementary Figure S6, Table 1 ).

Sperm morphology differed between the three morphs, with exception of midpiece, by less than one standard deviation ( Figure 3 ; Supplementary Figure S6, Table 1 ). Sperm of Independents were the shortest with the shortest midpiece and tail, and hence flagellum (midpiece and tail length combined). Acrosome and nucleus length, and hence head length, as well as tail did not differ between Faeders and Independents. Satellite sperm had the shortest nucleus and head, but the longest tail. The midpiece was longest in Faeders and 1.19 standard deviations longer than the midpiece from Independents (95%CI: 0.46 - 1.88; Figure 3 , Table 1 ). Total sperm length and flagellum length were similar for Faeders and Satellites ( Figure 3 ; Supplementary Figure S6, Table 1 ). Consequently, midpiece length relative to total sperm length was largest for Faeders, but similar for Satellites and Independents; flagellum length relative to total sperm length was largest in Satellites and similar in the other two morphs (Supplementary Figure S6, Table 1 ). The coefficients of variation in sperm traits were noisy with unclear trends and sperm of Independents was not less variable than that of the Faeders and Satellites ( Figure 3 ; Supplementary Figure S8, Table 1 ).

We thus demonstrate quantitative difference in sperm traits, but also a lack of major differences, i.e., sperm polymorphism, among the three male morphs. The two genes related to spermatogenesis that reside within the inversion thus have at most limited impact on the measured sperm traits. Such finding might be unsurprising given that tens to hundreds of genes influence spermatogenesis (Linn et al., 2021). Furthermore, the lack of major differences in sperm traits accords with previous studies showing no clear and consistent differences in sperm traits between males that use alternative mating tactics (Dougherty et al., 2022; Kustra and Alonzo, 2020).

Although one expects a steeper selection gradient for Faeder and Satellite males that have a lower probability to copulate multiple times with the same female (reviewed in Dougherty et al., 2022; Kustra and Alonzo, 2020), selection will act on all male morphs to optimize sperm traits. Additionally, males of all morphs mate with females of all morphs (Lank et al., 2002; Thuman, 2003), matching the size of sperm storage tubules in the female population may normalize selection on sperm size. Indeed, independent of the strength of sexual selection, Faeder-beneficial mutations are orders of magnitude less likely to arise than Independent-beneficial mutations, given a Faeder population size of 1% and the need for the beneficial mutation to arise within the inversion, being less than 0.5% of the genome. Moreover, outside the inversion, the genetic background has been selected to function best in the most abundant Independents, making evolution of special adaptations more difficult in the rare morphs (Giraldo-Deck et al., 2022).

Our results thus highlight that greater sperm competitiveness is an unlikely mechanism maintaining Faeders or Satellite inversions in the ruff population. The measured sperm traits seem to give the Faeder males little advantage over Independent males. However, there is a strong mating skew with only few Independents obtaining most copulations on a given lek (Tolliver et al., 2023; Vervoort and Kempenaers, 2019; Widemo and Owens, 1995), such that the average reproductive success of an Independent will be low. Thus, the rare Faeder males need moderate mating success through sneaking, facilitated by their female mimicking appearance and behavior (Jukema and Piersma, 2006) to compensate. Moreover, if Faeders copulate less frequently (no data), they may produce larger ejaculates and hence a rare copulation may be more likely to be successful everything else being equal (“more tickets in the lottery”). Possibly, Faeders do better, because morphs differ in other, non-measured, but fitness-relevant sperm traits. For example, the longer midpiece of Faeder sperm might correlate with longer sperm survival inside the female reproductive tract. However, this would need further study. To emphasize, despite the impressive work on genetics and genomics of ruff morphs (described in the introduction), we are still missing essential behavioral data from the wild, such as mating success and fitness measurements of individual males. Such lack of mating success data prohibits to make clear predictions about how differences in sperm competition strength and intensity between the three male morphs in ruffs influence their sperm traits.

Sperm morphology had only weak and unclear effects on variation in sperm swimming speed ( Figure 4 ; Supplementary Figures S9, S10, Table S5). Males having longer sperm, sperm with a longer midpiece (absolute and relative) or sperm with a longer flagellum did not have sperm that swam faster (r varied between 0.05 and 0.18; Supplementary Figure S10). The lack of an association between sperm midpiece length and swimming speed may have several explanations. First, our method to measure sperm velocity might not reflect the swimming speed of the sperm within the seminal fluid or within the female reproductive tract. Note, however, that our preliminary work suggests that the way the sperm moved did not change after we placed the sperm into a solution that contained cloacal fluid of a ruff female. Also, our method is the same as the one used to measure swimming speed in passerine sperm, and we used the exact same equipment and observers as in some of the studies on passerines (e.g., Knief et al., 2017; Míčková et al., 2023; Opatová et al., 2016). Second, although the midpiece contains the energy-producing mitochondria, midpiece volume rather than length might be relevant for sperm swimming speed (Cardullo and Baltz, 1991; Mendonca et al., 2018), but see Cramer et al. (2022), or midpiece length might be associated with other sperm performance parameters such as longevity (which we did not measure). Third, it remains unclear whether oxidative phosphorylation in the midpiece is the primary source of energy for moving sperm (Turner, 2003).

Against the predictions (Humphries et al., 2008), sperm with longer heads (nucleus) or smaller relative flagellum length swam faster, when measured as curvilinear and average-path velocity (but not straight-line velocity), although these effects were small (< 0.25 of standard deviation, Figure 4 ; maximum r = 0.22, Supplementary Figure S10). Theory predicts that sperm with a shorter head relative to flagellum should swim faster because a relatively shorter head reduces drag (Humphries et al., 2008). The few existing within-species studies provide equivocal evidence for this prediction (Cramer et al., 2015, 2021; Helfenstein et al., 2009; Mossman et al., 2009), but the prediction has been supported in a between-species study in mammals (Tourmente et al., 2011) and in passerine birds (Lüpold et al., 2009). A recent study on passerines showed that sperm with a long acrosome, a short nucleus, a wide helical membrane, and a more pronounced waveform along the sperm head “core” swam faster (Støstad et al., 2018).

We observed that ruff sperm seemed to move in a different manner to that previously described for passerine and likely also for non-passerine sperm (video examples). Specifically, ruff sperm seem to ‘vibrate’ from side to side while slowly moving forward, whereas passerine sperm typically swim straight by rapidly rotating around their longitudinal axis (Ballowitz, 1888; Humphreys, 1972; Vernon and Woolley, 1999). We can exclude that the unusual head movements of ruff sperm were caused by a lack of depth in the standard 20μm two-chamber count Leja-slide, because using a deeper Leja-slide (100μm) resulted in similar movements (not recorded). However, our observations need further verification using other methods (Vernon and Woolley, 1999). If confirmed, the difference in movement may not be due to differences in general morphology, because ruff sperm have a helical, screw-like head ( Figure 2 ), similar to that observed in passerine sperm (Retzius, 1909). Previous studies described that non-passerine sperm moves with regular helical (dextral) waves (Ballowitz, 1888; Humphreys, 1972; Vernon and Woolley, 1999) or via counter-clockwise turning of the entire sperm body (Bird and Laguë, 1977). Given the few descriptions and the lack of video recordings of non-passerine sperm movements (e.g., Cheng et al., 2002; Denk et al., 2005; Dogliero et al., 2017; Fischer et al., 2014; Gloria et al., 2014), it remains to be seen whether the way ruff sperm propel is unique to this species, or typical for other Scolopacids, shorebirds or Charadriiformes in general. Clearly, the directional movement of ruff sperm is distinct from the smooth, snake-like, movements of sperm of non-avian taxa (e.g., human or sea-urchin Psammechinus miliaris (Cosson et al., 2015; Gillies et al., 2009; Gray, 1955; Saggiorato et al., 2017; Smith et al., 2009), but see squid sperm (Bishop, 1958)).

The screw-like head of ruff sperm is similar to the head shape reported for other Scolopaci (sandpipers), but differs from the straight heads observed in other Charadriiformes, e.g., genus Larus or Vanellus (Retzius, 1909). Our results confirm that ruff sperm is remarkably long (median and mean = 139.5 µm, range: 105.7 - 150 µm) compared to other shorebirds, with species means ranging between 57 and 100 µm, which has been interpreted as a sign that sperm competition was/is strong in ruffs (Johnson and Briskie, 1999).