EFTA01202924.pdf

1 1 Lower Body Symmetry and Running Performance in Elite Jamaican Track and Field 2 Athletes 3 Short Title: Symmetry in Elite Jamaican Athletes 4 5 Robert Trivers's, Bernhard Fink'-, Mark Russell3, Kristofor McCarty*, Bruce James5, & Brian G. 6 Patestis* 7 8 'Graduate Program in Ecology and Evolution, Rutgers University, New Brunswick, New Jersey, 9 08901, USA. 10 2Courant Research Centre Evolution of Social Behavior & Institute of Psychology, University of 11 Gottingen, Kellnenveg 6, 37077 Gottingen, Germany. 12 ;Faculty of Health and Life Sciences, Northumbria University, Newcastle-upon-Tyne, NEI 8ST, 13 UK. 14 *Department of Psychology, Faculty of Health and Life Sciences, Northumbria University, 15 Newcastle-upon-Tyne, NE1 8ST, UK. 16 5MVP Track and Field Club, 237 Old Hope Road, University of Technology, Kingston 6, 17 Jamaica. 18 *Brian G. Palestis, Department of Biological Sciences, Wagner College, Staten Island, New 19 York, 10301, USA. 20 21 *Correspondence to: 22 23 EFTA01202924  2 1 Abstract 2 In a study of degree of lower body symmetry in 73 elite Jamaican track and field athletes we 3 show that both their knees and ankles (but not their feet) are — on average — significantly more 4 symmetrical than those of 116 similarly aged controls from the rural Jamaican countryside. 5 Within the elite athletes, events ranged from the 100 to the 800m, and knee and ankle asymmetry 6 was lower for those running the 100m dashes than those running the longer events with turns. 7 Nevertheless, across all events those with more symmetrical knees and ankles (but not feet) had 8 better results compared to international standards. Regression models considering lower body 9 symmetry combined with gender, age and weight explain 27 to 28% of the variation in 10 performance among athletes, with symmetry related to about 5% of this variation. Within 100m 11 sprinters, the results suggest that those with more symmetrical knees and ankles ran faster. 12 Altogether, our work confirms earlier findings that knee and probably ankle symmetry are 13 positively associated with sprinting performance, while extending these findings to elite athletes. 14 EFTA01202925  3 1 Introduction 2 Trivers et al. [1] showed that Jamaican eight year olds of both sexes with more 3 symmetrical knees were better sprinters 14 years later in both the 100 and the 200m dashes, 4 while symmetry of the upper body and feet did not predict sprinting speed, and ankle symmetry 5 sometimes seemed to have a minor positive effect on sprinting speed. There are two separate 6 reasons why we might expect symmetry of lower body traits to be positively associated with 7 sprinting speed. Symmetry is inherently more efficient in races—it is less physically demanding 8 and thus saves energy. This is presumably why the lower body asymmetry of Jamaican children 9 8.2 years of age is 1/3rd that of upper body asymmetry 12). Walking and running are inherently 10 symmetrical and should favor symmetrical traits associated with them while upper body 11 movements may or may not be symmetrical (vide, laterality) [3,4]. 12 On the other hand, it has been known since the 1950's from experimental work on 13 Drosophila that stress during early development is associated with greater adult bodily 14 asymmetry, as is genetic inability to deal with the stress (e.g., inbred vs. outbred). This led to the 15 notion that fluctuating asymmetry (FA) — deviations from bilateral symmetry in paired traits, 16 randomly distributed to the left and right side — is a measure of developmental instability, i.e., 17 the inability of an organism genetically to buffer the system against stressors to achieve the 18 optimal state, namely symmetry itself [5]. The key is that if a population shows true FA then it 19 can be presumed to be attempting to be symmetrical, so that failure to do so is a measure of 20 failure to reach—in the face of developmental perturbations—the phenotype that the genotype is 21 aiming for. Once it was discovered that females in a variety of birds, insects and mammals 22 preferred symmetrical males as mating partners. even in species lacking male parental EFTA01202926  4 1 investment, it became obvious that body symmetry must be positively associated with genetic 2 quality—a resistance from disease, for example [6]. and many other advantageous traits [71. 3 To test for associations between symmetry and athletic performance, we compared lower 4 body symmetry (knees, ankles and feet) in a control sample of Jamaicans measured in the 5 countryside, carefully matched to a sample of elite Jamaican athletes measured in Kingston (all 6 members of the MVP Track and Field Association). Among the elite athletes we also obtained 7 performance—best times in preferred events against world records—to see if variation in peak 8 sprinting times was in part predicted by variation in lower body symmetry. 9 In this paper we show that elite athletes have markedly more symmetrical knees than 10 countryside controls and also more symmetrical ankles, while symmetry of feet does not differ 11 between the two. We further demonstrate that within elite sprinters, performance shows striking 12 and consistent positive associations with both knee and ankle symmetry. To wit, taking all 13 athletes together, knee symmetry predicts sprinting ability while ankle symmetry almost does so. 14 By adding covariates, such as weight and age, knee and ankle symmetry remain as marginally 15 significant predictors of athletic success at the highest level of elite performance. Considering 16 only the 100m sprinters (n = 32) we find positive correlations between degree of knee and ankle 17 symmetry and world performance. It also seems noteworthy that overall symmetry decreases 18 among those who run 200m, 400m or 800m races, as if adapted to or caused by frequent left- 19 hand turns. 20 21 22 Methods EFTA01202927  5 1 Subjects 2 We recruited two groups of subjects: elite Jamaican track and field athletes and controls 3 from the Jamaican countryside. The elite athletes were all members of the MVP Track and Field 4 Club working out of University of Technology in Kingston. We were able to get a nearly 5 complete sample of club membership (73 of 77). There are exclusive criteria for being members 6 of MVP and it is highly sought after by up and coming Jamaican athletes. International 7 Association of Athletics Federations (IAAF) scores allow comparisons across disciplines and 8 genders [8] (www.iaaf.org) and we used these scores, which we calculated based on personal 9 bests, as our measure of performance. Approximately 90% of MVP club members have IAAF 10 scores above 1000, and 6 of our subjects have remarkably high scores (>1200). Although their 11 training all included sprinting, the athletes' best events varied, and in addition to comparing 12 athletes and controls, we also compared symmetry of athletes whose specialties differed. 13 Control subjects consisted of individuals recruited in the countryside through word of 14 mouth by someone who knew the community very well and had worked with the Jamaican 15 Symmetry Project [2] since its inception in 1996. This person was given an exact list of the ages 16 and sexes of the elite athletes we would be measuring in Kingston, so she could recruit a 17 matching sample, which she did exactly. We then invited additional subjects who were within 18 the age range of the elite athletes to appear and be measured, except that we discouraged all 19 overweight individuals (which were overwhelmingly women) since we doubted there were many 20 overweight elite athletes. The final number of control subjects was 116. 21 Elite athletes and controls were of very similar age (athletes: mean +/-SD = 23.0 +/-3.2; 22 controls 23.0 +/-3.6; tip= 0.14, p = 0.89). There is also no difference in variance in age between EFTA01202928  6 1 groups (Levene's test, F = 1.68, p = 0.20) and the ranges are as follows: elite 17.13 — 31.95; 2 control 17.21 — 32.46. 3 Measurement oftraits 4 Before traveling to Jamaica the three measurers met, and each measurer was exclusively 5 allocated one trait to eliminate between-measurer effects. There were two sessions, each of 6 approximately 6 hours, in which the measurement points for knees, ankles and feet were agreed 7 upon and preliminary repeatability for each trait was established. Measurement points were the 8 same as adopted in the Jamaican Symmetry Project in 1996 [2] and thus match the landmarks 9 used in Trivers et al. [1]. 10 When conducting the study in Jamaica, we measured the traits twice per side for each 11 individual (second foot measurements missing for one subject). Repeated measurement is 12 standard in FA studies, because the differences between sides are often so small that they can be 13 similar in magnitude to measurement error [9-11]. Only by measuring each side at least twice 14 can one demonstrate that the differences between sides reflect actual asymmetry, rather than 15 measurement error. The calculated average of the two measurements were used in the subsequent 16 statistical analysis. 17 To preserve anonymity each participant received a unique ID at arrival, printed on A6 18 cards, and only this number was used to assign his/her measurements. Measurements of the three 19 lower body traits were all performed using a digital Vernier caliper (Preisser products, Germany) 20 to an accuracy of 0.01 mm. To eliminate potential measurement biases, the calipers were closed 21 after each measurement. In addition to the three lower body traits, weight and height were 22 measured and age was self-reported. Collecting all measurement took approximately 10 minutes EFTA01202929  7 1 per participant and there was a time span of some 3-5 minutes between first and second 2 measurements of a trait, in which participants completed measurements at another station before 3 they returned for the second measurement. Each measurement was called out to an assistant 4 sitting next to the investigator to ensure that the investigator could not memorize them. 5 6 The size of the left and right knee of each participant was measured as defined by the 7 largest breadth, measured between the medial and lateral epicondyles of the femur. The 8 investigator measuring ankles first located the widest point of the ankle by hand before putting 9 the caliper in place. For both knee and ankle measurements, participants were seated on a chair 10 with legs bent by 90° and measurements were taken by an investigator sitting in front of the 11 participant. Somatometric landmarks were palpated and measurements were taken with constant 12 pressure of the caliper to minimize soft tissue related measurement error. Foot length was 13 measured bilaterally using standardized A3 graphical paper (resolution of I mm) that was then 14 measured with calipers. In a seated position, participants were instructed to place both feet on to 15 the paper in a manner which aligned the ptemion to the head of the 2nd metatarsal axes of each 16 foot in parallel. Using a leveling device, markings were placed at the ptemion and at the tip of 17 toe one of each foot. To ensure consistency of measurement, markings were placed at the tip of 18 toe one irrespective of whether toes two through five would have elicited a greater foot length. 19 The perpendicular distance between two parallel lines extrapolated from the original foot 20 markings was deemed foot length. 21 22 Statistical analysis EFTA01202930  8 1 Most analyses were performed using IBM SPSS Statistics 21. The presence of fluctuating 2 and directional asymmetry (FA and DA) were assessed using the Excel template at 3 www.biology.ualberta.ca/palmeriasytn/FA/FA-Refs.htmfttools. With this template, the presence 4 of significant FA (FA > measurement error) is demonstrated by a significant F-test for a sides X 5 individuals interaction in a mixed model ANOVA [9,10]. DA is indicated by a significant effect 6 of sides in the ANOVA model. 7 Significant DA was present in feet and ankles (see Measurements in the main text). To 8 ensure that statistical tests are comparing FA and are not biased by directional differences 9 between sides, we used the index of FA developed by Graham et al. [11]. This index is based on 10 unrotated factors extracted from the covariance matrix of a principal component analysis of the 11 mean right side measurements on a trait and the mean left side measurements. This first factor 12 represents the covariance between sides, plus half the FA and half the remaining measurement 13 error. The second factor represents half the FA and half the remaining measurement error. This 14 second factor is therefore an index of FA and does not include DA, which would be extracted 15 with the covariance between sides. When analyzing overall asymmetry, including DA, we use 16 relative asymmetry (absolute asymmetry in a trait divided by trait size). 17 Unsigned FA (absolute value of right — left) has an asymmetrical frequency distribution 18 with only half of a bell curve. However, FA should not then be transformed to achieve normality, 19 because F-tests are robust and transformation may change the underlying relationships between 20 FA and other variables [10, 12]. Comparisons of FA are comparisons of variance in a trait, and 21 comparisons of the means of two half-normal distributions gives an unbiased estimate of 22 differences in variance [9, 10]. Parametric correlation and regression analyses are also robust in 23 this case [12]. To avoid inflating Type I error by conducting multiple tests, multivariate analyses EFTA01202931  9 1 were used whenever appropriate. When the dependent variables were FA in individual traits, the 2 three traits were tested simultaneously using multivariate GLM in SPSS (MANOVA, 3 MANCOVA). When testing dependent variable composite FA (sum of the FA index for the three 4 traits) we used univariate GLMs (ANOVA, ANCOVA). When the FA values were the 5 independent variables and athletic performance (indexed by IAAF scores) the dependent 6 variable, we used multiple regression. 7 Significance tests were conducted with an alpha level of 0.05, and, whenever appropriate, 8 effect sizes are given in addition to p-values. Tests are two-tailed except when testing for 9 relationships between FA and racing performance. In these situations we use a directed test, a 10 compromise between one and two-tailed testing for strongly directional hypotheses [13, 14]. Our 11 previous work [1] showed a positive relationship between symmetry (low FA) and sprinting and 12 there is no reason to expect that greater symmetry would result in poorer performance, therefore 13 we have a strong directional hypothesis. Our other comparisons are not as obviously directional: 14 effects of weight, age, and gender on FA and on performance and effects of athletic training on 15 FA. 16 The dataset will be deposited in the Dryad repository. 17 Eth lea! statement 18 This research was approved by the Institutional Review Board of Rutgers University, 19 protocol # 14-425M. All participants were informed on the purpose of the study and gave written 20 consent. Permission of legal guardians was also obtained for subjects below 18 years of age. 21 EFTA01202932  10 1 2 Results 3 Measurements 4 Mean values (+/-SD) for the three traits are as follows (all n = 189): right knee 96.80mm 5 +1-7.63; left knee 96.66 +/-7.94; right ankle 67.99 +/-5.40; left ankle 68.70 +/-5.48; right foot 6 259.89 +/-17.23; left foot 261.03 +/-17.49. A Mixed Model Sides X Individuals ANOVA (see 7 Methods) demonstrates significant FA in all 3 traits (FA> measurement error: knees F188,378 = 8 1.96; ankles F188,378= 2.91; feet F188.376= 5.55; all p < 0.0001) and significant directional 9 asymmetry (DA) in ankles (Fuss= 34.76, p < 0.0001) and feet (F1,188 = 23.00, p < 0.0001) but 10 not knees (Fun = 1.47, p = 0.23). The left foot and ankle tended to be larger than the right: 11 64.0% of subjects had larger left ankles and 61.9% had larger left feet, with little variation 12 among groups. The significant p-values are so low that they are still at the 13 < 0.0001 level even 13 with any correction for multiple tests. Because of the presence of DA, we use FA values 14 corrected for DA (see Methods). 15 Athletes vs. controls 16 Mean values of the FA index (corrected for DA) for athletes and controls are presented in 17 Figure 1. The FA index was compared between elite athletes and controls in MANOVAs with 18 dependent variables knee, ankle and foot asymmetry, and group (elite vs. control) and gender as 19 the factors. There are significant differences between athletes and controls in knee and ankle 20 asymmetry, but not in foot asymmetry (Table I). There were no significant sex differences, but a 21 significant gender x group interaction for knees because female controls had particularly high 22 mean knee FA (mean +/-SD = 1.04 +/-0.82, n = 56; compare to Figure 1). Using composite EFTA01202933  11 1 asymmetry (sum of knee, foot, ankle) rather than individual traits again demonstrates significant 2 differences between athletes and controls and no significant effect of gender, but also no 3 significant gender x group interactions. 4 Including weight and age as covariates does not change the key results (Table 1). Again 5 significant differences between athletes and controls are present for composite FA and separately 6 for knees and ankles, but not feet. The gender x group interaction is now significant for both 7 knees and feet, rather than just knees, but not when using composite FA. There is also a positive 8 relationship between weight and foot FA, but the overall relationship with weight is marginal 9 and there are no significant relationships between age and FA (Table 1). 10 Within-athlete con:part:cons 11 Relationships between FA and performance of the athletes, measured by IAAF scores, 12 were tested using multiple regression. For FA we used the residuals of a regression between FA 13 corrected for DA and discipline, to control for effects of particular events on symmetry (see 14 below). There is a significant negative relationship between composite FA and performance (i.e., 15 more symmetrical athletes perform better; Table 2). Among individual traits, a significant 16 relationship is present for knees, with ankles marginal (directional p = 0.087; Table 2). There is 17 also a significant effect of gender. Even without considering covariates age and weight, the 18 regression model explains 12 to 13% of the variance in performance (adjusted r2 = 0.12 when 19 using composite FA and 0.13 when using individual traits). Partial r2 values indicate that knee or 20 composite FA, depending on the model, are each related to approximately 5% of the variation in 21 performance (Table 2). EFTA01202934  12 1 When age and weight are added in as covariates, there is a strong positive relationship 2 between age and performance, and the regression model explains 27 to 28% of the variance in 3 performance (adjusted r2 = 0.27 when using composite FA and 0.28 when using individual traits; 4 Table 2). The relationship with composite FA remains significant and is related to 5% of the 5 variation in performance, but relationships with weight and with individual traits are not 6 significant (both knees and ankles are marginal; Table 2). The effect of gender is no longer 7 significant, and likely resulted in part from a selection effect based on age: generally the most 8 successful athletes are those most likely to continue competing as they age, and this may be 9 especially true of women. Examination of the historical records of those six males and six 10 females with the top ten finishes per sex (two of whom are in our sample) reveals that the age 11 effect is probably not only due to selection—there is also typically a within-individual 12 improvement with age for several years, but then a leveling off and a decline late in an athlete's 13 career, circa 28 years (www.iaatorglathletes). 14 We also conducted analyses restricted to those whose best event is the 100m sprint, 15 because we predicted the relationship between FA and performance to be particularly important 16 for sprinting events and our sample for sprinters, although small (n = 32), is larger than for any 17 other events. Here we use the index of FA corrected for DA, but not the residuals with event, 18 because we are testing within one event. A regression with dependent variable IAAF score and 19 independent variables gender, knee, ankle, and foot FA is not significant (F4,27 = 1.97, p = 0.13, 20 adjusted r2 = 0.11) but suggests relationships with knee and ankle FA (p = -0.33, directional p = 21 0.044 and p = -0.33, p = 0.045, respectively), but not foot FA (l3 = 0.18, p = 0..46) or gender (p 22 = 0.10, two-tailed p = 0.56). The overall regression model is significant when covariates age and 23 weight are added (F6,25 = 2.84, p = 0.030, adjusted r2 = 0.26), but the apparent relationships with EFTA01202935  13 1 knee and ankle FA are no longer significant (knees: p = -0.11, directional p = 0.34; ankles: p = - 2 0.24, p = 0.093). For foot FA, p = 0.13, p = 0.55. Two-tailed p-values for the other variables are 3 all greater than 0.05, but less than 0.10. Relationships with composite FA are not significant 4 (directional p = 0.31 without covariates, 0.43 with covariates), because, within sprinters, foot FA trends opposite the predicted direction and therefore when knee and ankle FA are combined with 6 foot FA the relationships become weaker. When using composite FA, significant effects of 7 gender (p = 0.62, two-tailed p = 0.017), age (p = 0.39, p = 0.024), and weight (13 = 0.55, p = 8 0.034) are all present and 25% of the variation in sprinting performance is explained (F4,27 = 9 3.56, p = 0.019, adjusted r2 = 0.25). IAAF scores should not show significant differences 10 between males and females, because they are calculated to allow comparison of relative 11 performance regardless of gender. The significant relationship we find is likely a result of our 12 sample happening to contain particularly successful female sprinters. 13 In addition to testing for relationships between FA and performance, we tested whether 14 symmetry among athletes varied with their primary event, because events differ in the stresses 15 they produce and may cause both fluctuating and directional asymmetry. Because of small 16 sample sizes for several events, the events were combined into three categories: straight sprints 17 (100m, n = 32), longer races with turns (200m, 400m, 800m; n = 29), and events with jumping or 18 leaping (100 and 110m hurdles, 400m hurdles, long jump, high jump; n = 11) [the one shot- 19 putter was excluded]. An ANOVA test revealed significant variation among these categories in 20 relative composite FA (not corrected for DA): F20 = 3.67, p = 0.031. A Tukey's HSD post-hoc 21 test revealed that longer distance runners were significantly more asymmetrical than sprinters. 22 Means for sprinting and jumping events were similar. In a MANOVA, there were no significant 23 differences among groups for the individual traits, but the trends for knees and especially ankles EFTA01202936  14 1 match the trend for composite FA (Fig. 2). If we use the FA index corrected for DA the patterns 2 are similar, but are not significant (composite FA index, F2,69 = 2.37, p = 0.10). However, here 3 we specifically want to test the effects of athletic stresses, which are often asymmetric, on 4 overall symmetry and therefore DA should be included. Above we were interested instead in the 5 relationships between developmental stability (indexed by FA) and performance, and thus 6 needed to factor out DA. 7 8 Discussion 9 Jamaicans are the elite sprinters of the world. Why? If symmetry of knees and ankles is a 10 factor, why should Jamaicans be especially symmetrical (there is no knowledge of whether they 11 actually are)? One possibility is heterozygosity for genes important to sprinting. The slave trade 12 greatly increased heterozygosity on the West African side by mixing genes up and down the 13 West coast of Africa from Senegal to Nigeria [15, 16]. Recently a mtDNA haplotype has been 14 isolated that correlates with success in African American—but not Jamaican—sprinters [17]. 15 Since there is a general (if often weak) positive relationship between heterozygosity and body 16 symmetry [18] we are eager to do targeted studies of genomics on areas associated with 17 sprinting, including energy substrate utilization, muscle fibre-type distribution and body 18 composition analyses (with specific reference to the shape and size of the glutei maximi). Fast 19 twitch (anaerobic) muscle fibres are characterized by specific adaptations which benefit the 20 performances of explosive high-intensity actions such as those involved in sprinting. Notably, 21 West Africans appear to have a higher fast twitch muscle fibre content than do comparable 22 Europeans (67.5% vs 59% in one sample [19], as cited in [20]). EFTA01202937  15 1 An interesting problem arises. How much is genetic quality revealed by overall body 2 symmetry and how much is it associated with symmetry of particular body parts? To give a 3 (partly) counterintuitive example, over-all body symmetry, based on 10 measurements of race 4 horses, is positively correlated with performance in competitive races, but both lower body 5 asymmetry, especially of the knees, is a correlate and so also are features of the head [21], as if 6 the symmetry of the apparatus controlling the running (brain, head, sensory organs) is as 7 important as the parts doing the running (themselves already selected for symmetry as shown in 8 the Jamaican children, see [1]). Moller and colleagues [22] also demonstrated a link between 9 lower-body symmetry and locomotion (in chickens), but did not measure upper-body traits. 10 Manning and Pickup [23] showed for humans that degree of nostril symmetry positively predicts 11 running performance in middle-distance runners, which is plausible since middle distance 12 running relies on oxygen much more than do sprints and symmetrical organs of air intake should 13 maximize consumption. Foot asymmetry predicts physical aggressiveness in boys [24] and lower 14 body symmetry predicts the same for college undergraduates [25], perhaps because stability is 15 especially important in fights. Ear asymmetry, in turn, predicts tendency for women and girls to 16 cradle a baby or doll, respectively, on the right side [3]. By contrast, Tomkinson, Popovie and 17 Martin [26] failed to find any significant differences between elite football (soccer) and 18 basketball players, nor between elite (national league) and sub-elite (state leagues) categories of 19 each. This is not surprising given their methodology. Numerous traits, soft body and hard, were 20 measured for symmetry in each individual, and then averaged together so that a single average 21 value was typically used. In our own work, symmetry of one part of the body (knees) suggests 22 quality of associated parts as well, not just the bones of the knee but the attached cartilage, 23 muscle, associated structures and developmental control. We know nothing about the details of EFTA01202938  16 1 this. The only part of the sprinting apparatus thought in advance to be important that we did not 2 measure was the buttocks, hips, and glutei maximi (the largest and strongest muscles in the 3 human body). Muscular strength measurements and imaging techniques such as dual-energy X- 4 ray absorptiometry may allow a chance for progress on this (possibly key) variable. 5 A second problem has to do with cause and effect. If we find that those elite sprinters 6 with the best times also have the most symmetrical knees, is it because those with symmetrical 7 knees do well in advance or is it that intensive training leads to both success and more 8 symmetrical knees? Of course it is likely that cause and effect go in both directions, and there is 9 a weak trend toward decreased knee FA with age in the athletes (r = -0.15, p = 0.22). However, 10 the fact that 8 year old knee FA predicts sprinting ability 14 years later [1] and that there are no 11 statistically significant changes in symmetry with age in elite athletes, even though the older will 12 have exposed themselves to symmetrical forces during training longer than the younger, suggests 13 that knee symmetry is by no means a mere reflection of training. 14 In addition to the positive effects of symmetry on running performance that we 15 demonstrate, we also find differences in symmetry between running events. Runners specializing 16 in longer races are less symmetrical than sprinters, and this difference is particularly noticeable 17 for the ankles. Why this is the case is unclear, but may relate to the asymmetrical stresses of 18 events requiring turns and an increased prevalence of rearfoot striking in longer events, which 19 would alter the distribution of stresses acting on the joints [27]. 20 21 References EFTA01202939  17 1 1. Trivers R, Palestis BG, Manning JT (2013). The symmetry of children's knees is linked to 2 their adult sprinting speed and their willingness to sprint in a long-term Jamaican study. 3 PLoS ONE 8: e72244. 4 2. Trivers R, Manning IT, Thornhill R, Singh D, McGuire M (1999). Jamaican symmetry 5 project: long-term study of fluctuating asymmetry in rural Jamaican children. Hum Biol 71: 6 417-430. 7 3. Manning JT, Trivers, RL, Thomhill R, Singh D, Denman J, et al. (1997). Ear asymmetry and 8 left-side cradling. Evol Hum Behav 18: 1-14. 9 4. Manning JT, Trivers R, Thornhill R, Singh D (2000). The 2nd:4th digit ratio and hand 10 preference in Jamaican children. Laterality 5: 121-132. 11 5. Van Valen L (1962). A study of fluctuating asymmetry. Evolution 16: 125-142. 12 6. Moller AP (2006). A review of developmental instability, parasitism and disease: infection, 13 genetics and evolution. Infect Genet Evol 6: 133-140. 14 7. Moller AP, Swaddle JP (1997). Asymmetry, developmental stability and evolution. New 15 York: Oxford University Press. 16 8. Spiriev B (2011). IAAF Scoring Tables of Athletics. Monaco: IAAF. 17 9. Palmer AR (1994). Fluctuating asymmetry analyses: a primer. In: Markow T, editor. 18 Developmental Instability: Its Origins and Evolutionary Implications. Dordrecht: Kluwer. pp 19 335-364. 20 10. Palmer AR, Strobeck C (2003). Fluctuating asymmetry analyses revisited. In: Polak M, 21 editor. Developmental Instability: Causes and Consequences. New York: Oxford University 22 Press. pp 279-319. EFTA01202940  18 1 II. Graham JH, Emlen JM, Freeman DC, Leamy LJ, Kieser JA (1998). Directional asymmetry 2 and the measurement of developmental instability. Biol J Linn Soc 64: 1-16. 3 12. Gangestad SW, Thornhill R. (1998). The analysis of fluctuating asymmetry redux: the 4 robustness of parametric tests. Anim Behav 55: 497-501. 5 13. Rice WR, Gaines SD. (1994). `Heads I win, tails you lose': testing directional alternative 6 hypotheses in ecological and evolutionary research. Trends Ecol Evol 9: 235-237. 7 14. Ruxton GD, Neuhauser M. (2010). When should we use one-tailed hypothesis testing? 8 Methods Ecol Evol 1: 114-117. 9 15. Deason ML, Salas A, Newman SP, Macaulay V, Morrison EA, et al. (2012). Interdisciplinary 10 approaches to the demography of Jamaica. BMC Evol Biol 12: 24. 11 16. Simms TM, Rodriguez CE, Rodriguez R, Herrera RJ (2010). The genetic structure of 12 populations from Haiti and Jamaica reflect divergent demographic histories. Am J Phys 13 Anthropol 142: 49-66. 14 17. Deason M, Scott R, Irwin L, Macaulay V, Fuku N, et al. (2012). Importance of mitochondrial 15 haplotypes and maternal lineage in sprint performance among individuals of West African 16 ancestry. Scand J Med Sci Sports 22: 217-223. 17 18. Vollestad LA, Hindar K, Moller AP (1999). A meta-analysis of fluctuating asymmetry in 18 relation to heterozygosity. Heredity 83: 206-218. 19 19. Bouchard C, Dionne FT, Simoneau J-A, Boulay M (1992). Genetics of aerobic and anaerobic 20 performances. Exerc Sport Sci Rev 20: 27-58. 21 20. Holden C (2004). Peering under the hood of Africa's runners. Science 305: 637-639. 22 21. Manning JT, Ockenden L (1994). Fluctuating asymmetry in race horses. Nature 370: 185- 23 186. EFTA01202941  19 1 22. Moller AP, Sanotra GS, Vestergaard KS (1999). Developmental instability and light regime 2 in chickens (Gallus gallus). Appl Anim Behav Sci 62: 57-71. 3 23. Manning JT, Pickup LJ (1998). Symmetry and performance in middle distance runners. Int J 4 Sports Med 19: 1-5. 5 24. Manning JT, Wood D (1998). Fluctuating asymmetry and aggression in boys. Hum Nat 9: 6 53-65. 7 25. Furlow B, Gangestad SW, Armijo-Prewitt T (1998). Developmental stability and human 8 violence. Proc R Soc Lond B 265:1-6. 9 26. Tomkinson GR, Popovie N, Martin M (2003). Bilateral symmetry and the competitive 10 standard attained in elite and sub-elite sport. J Sports Sci 21: 201-211. 11 27. Kulmala JP, Avela J, Pasanen K, Parkkari J (2013). Forefoot strikers exhibit lower running- 12 induced knee loading than forefoot strikers. Med Sci Sports Exerc 45: 2306-2313. 13 14 15 Acknowledgments 16 We are most grateful for the personal interest of Jeffrey Epstein and Gordon Getty. We 17 thank Carla Hufschmidt and Tessa Cappelle for expert help in the field and John Graham for 18 advice on analysis. 19 Funding 20 We thank the Enhanced Education Foundation, the Ann and Gordon Getty Foundation, 21 the Biosocial Research Foundation, the Center for Human Evolutionary Studies (CHES) and the 22 German Science Foundation (DFG) for financial support. The funders had no role in study 23 design, data collection and analysis, decision to publish, or preparation of the manuscript. EFTA01202942  20 1 Tables 2 Table L Statistical comparisons of FA index between athletes and controls. Multivariate tests 3 compare three traits simultaneously (knee, ankle, foot), while univariate tests compare composite 4 FA (sum of knee, ankle, foot). Results are shown first with no covariates and then with weight 5 and age added as covariates. Statistically significant relationships (two-tailed tests, alpha = 0.05) 6 are indicated with asterisks. 7 a. Multivariate tests, no covariates: 8 Overall Multivariate Relationships Treatment Pillai's Trace F3.183 P Athlete vs. Control 0.083 5.55 0.001* Gender 0.013 0.83 0.48 Gender X A v. C 0.041 2.59 0.055 9 lo Relationships for Individual Traits Factor Trait Fuss P Partial etaz Athlete vs. Knee 10.37 0.002* 0.053 Control Athlete vs. Ankle 4.55 0.034* 0.024 Control Athlete vs. Foot 0.62 0.43 0.003 Control Gender Knee 2.24 0.13 0.012 Gender Ankle 0.26 0.61 0.001 Gender Foot 0.04 0.84 —0 Gender X Knee 4.60 0.033* 0.024 A. v. C. Gender X Ankle 0.001 0.97 —0 A. v. C. Gender X Foot 3.13 0.079 0.017 A. v. C. 11 12 13 EFTA01202943  21 1 b. Univariate test on composite FA, no covariates: 2 Overall Model: F3,185 = 4.86, p = 0.003*, adjusted r2 = 0.058 Factor Fuss P Partial eta2 Athlete vs. Control 13.24 <0.0001* 0.067 Gender 0.54 0.46 0.003 Gender X A. v. C. 0.07 0.80 —0 3 4 c. Multivariate tests, age and weight as covariates: 5 Overall Multivariate Relationships Treatment Pillai's Trace Finn P Athlete vs. Control 0.083 5.43 0.001* Gender 0.023 1.43 0.24 Gender X A v. C 0.046 2.94 0.035* Age 0.021 1.29 0.28 Weight 0.036 2.23 0.087 6 7 Relationships for Individual Traits Factor Trait Fuss P Partial eta2 Athlete vs. Knee 10.21 0.002* 0.053 Control Athlete vs. Ankle 4.52 0.035* 0.024 Control Athlete vs. Foot 0.49 0.48 0.003 Control Gender Knee 2.26 0.14 0.012 Gender Ankle 0.38 0.54 0.002 Gender Foot 1.74 0.19 0.009 Gender X Knee 4.34 0.039* 0.023 A. v. C. Gender X Ankle 0.002 0.96 —0 A. v. C. Gender X A. v. Foot 4.50 0.035* 0.024 C. Age Knee 0.13 0.72 0.001 Age Ankle 0.88 0.35 0.005 Age Foot 2.73 0.10 0.015 Weight Knee 0.04 0.84 —0 Weight Ankle 0.05 0.82 —0 Weight Foot 6.57 0.011* 0.035 8 EFTA01202944  22 1 d. Univariate test on composite FA, age and weight as covariates: 2 Overall Model: F5,183 = 3.46, p = 0.005*, adjusted r2 = 0.061 Factor Fug; P Partial eta2 Athlete vs. Control 12.82 <0.001* 0.065 Gender 1.63 0.20 0.009 Gender X A. v. C. 0.002 0.96 —0 Age 0.33 0.56 0.002 Weight 1.99 0.16 0.011 EFTA01202945  23 1 Table 2. Multiple regressions within athletes, testing for relationships with performance (IAAF 2 scores). FA values are residuals of the FA index on the athletes' primary events. P-values for the 3 predicted relationship between FA and performance are directional (see Methods). 4 a. Individual traits, no covariates 5 Overall Model: F4,68 = 3.38, p = 0.014*, adjusted T2 = 0.12 Factor Beta Partial r2 P Resid. Knee FA -0.22 0.05 0.040* Resid. Ankle FA -0.17 0.03 0.087 Resid. Foot FA -0.09 0.01 0.28 Gender 0.31 0.10 0.007* 6 7 b. Composite FA, no covariates 8 Overall Model: F2,70 = 6.20, p = 0.003*, adjusted r2 = 0.13 Factor Beta Partial r2 P Resid. Composite FA -0.22 0.05 0.031* Gender 0.33 0.11 0.004* 9 10 c. Individual traits, age and weight as covariates: 11 Overall Model: F6,66 = 5 43, p < 0.0001*, adjusted r2 = 0.27 Factor Beta Partial r2 P Resid. Knee FA -0.16 0.03 0.084 Resid. Ankle FA -0.16 0.03 0.081 Resid. Foot FA -0.06 0.01 0.34 Gender 0.21 0.02 0.20 EFTA01202946  24 Age 0.41 0.19 <0.0001* Weight -0.15 0.01 0.35 1 2 d. Composite FA, age and weight as covariates 3 Overall Model: F4,68 = 8.04, p <0.0001*, adjusted r2 — 0.28 Factor Beta Partial r2 P Resid. Composite FA -0.18 0.05 0.046* Gender 0.24 0.04 0.12 Age 0.42 0.20 <0.0001* Weight -0.12 0.01 0.44 EFTA01202947  25 1 Figure Captions 2 3 Figure 1. Boxplots for the three traits and their sum. FA is corrected for directional asymmetry 4 using Graham et al.'s index (see Methods). Elite athletes (n = 73) are represented by open bars 5 and controls (n = 116) by bars filled with diagonals. Circles represent outliers, with stars 6 indicating extreme outliers: color of circles and stars corresponds to coloring in bars. Controls 7 have significantly higher knee, ankle and composite FA than the athletes (see Table 1). 8 9 Figure 2. Boxplots for the three traits in athletes compared across events. These values reflect 10 overall asymmetry, including FA and DA, divided by trait size. Knees are represented by open 11 bars, ankles by bars filled with diagonals, and feet by bars filled with dots. Circles represent 12 outliers, with stars indicating extreme outliers; color of circles and stars corresponds to coloring 13 in bars. EFTA01202948  26 1 Figures 2 Figure 1 3 4 • Group Athletes Z Controls x (1) V C LL • • • 6 • • • 0