Male sexual ornament size but not asymmetry re ects condition in stalk-eyed ies

Transcription

1 Male sexual ornament size but not asymmetry re ects condition in stalk-eyed ies Patrice David 1,3, Andrew Hingle 1, Duncan Greig 1,4, Adam Rutherford 1,2, Andrew Pomiankowski 1 and Kevin Fowler 1* 1 The Galton Laboratory, Department of Biology, University College London, 4 Stephenson Way, London NW1 2HE, UK 2 Developmental Biology Unit, Institute of Child Health, University College London Medical School, London WC1N 1EH, UK 3 Gënëtique et Environnment, Universite Montpellier II, Montpellier Cëdex 05, France 4 Yeast Genetics, Institute of Molecular Medicine, University of Oxford, Oxford OX3 9DU, UK Models of sexual selection predict that females use ornament size to evaluate male condition. It has also been suggested that ornament asymmetry provides females with accurate information about condition. To test these ideas we experimentally manipulated condition in the stalk-eyed y, Cyrtodiopsis dalmanni, by varying the amount of food available to developing larvae. Males of this species have greatly exaggerated eyestalk length and females prefer to mate with males with wider eyespans. Our experiments show that male ornaments (eyestalks) display a disproportionate sensitivity to condition compared with the homologous character in females, and to non-sexual traits (wing dimensions). In contrast, in neither sex did asymmetry re ect condition either in sexual ornaments or in non-sexual traits. We conclude that ornament size is likely to play a far greater role in sexual selection as an indicator of individual condition than does asymmetry. Keywords: condition; uctuating asymmetry; handicap; ornament; sexual selection; stalk-eyed ies 1. INTRODUCTION Male animals have evolved exaggerated sexual ornaments to attract mates (Darwin 1871; Andersson 1994). Various models of sexual selection predict that because ornaments are costly, ornament size covaries positively with condition, allowing females to use ornament size to evaluate male quality (Andersson 1986; Pomiankowski 1987; Grafen 1990; Iwasa et al. 1991). It also has been proposed that uctuating asymmetry (FA), the random nondirectional deviation from perfect bilateral symmetry (Van Valen 1962), is used by females to assess male quality. This proposal is based on the assumption that trait asymmetry is a sensitive measure of developmental stress and has led to the prediction that ornament FA covaries negatively with condition (MÖller & Pomiankowski 1993; Watson & Thornhill 1994; MÖller & Swaddle 1998). These predictions relate to costly, sexual ornaments that have been subject to exaggeration by sexual selection. Non-sexual traits are thought to lack or show much weaker condition dependence in their size and asymmetry. However, much of the evidence for these proposed relationships comes from correlational studies of stressful conditions and ornament size or asymmetry (reviewed in Andersson 1994; MÖller & Swaddle 1998). The key problem with correlational studies is that their results may be confounded by uncontrolled factors. One solution is to use experimental manipulations so that stress can be * Author for correspondence (k.fowler@ucl.ac.uk). applied in a systematic and controlled fashion. Several studies have used this approach to show that ornament size often has condition-dependent expression (Andersson 1994). However, the only two experimental studies of ornament asymmetry have yielded contrasting results. In barn swallows, Hirundo rustica, ornament asymmetry was negatively a ected by parasite load (MÖller 1992), but in the horned beetle, Onthophagus taurus, only ornament size and not asymmetry responded to nutritional stress (Hunt & Simmons 1997). Here we experimentally evaluate the e ect of variation in condition in laboratory populations of Malaysian stalkeyed ies, Cyrtodiopsis dalmanni. Males of this species have greatly exaggerated eyestalk length and females prefer to mate with males with wider eyespans (Wilkinson & Reillo 1994). Females possess much shorter eyestalks. We manipulated larval food density as a measure of condition and examined the consequences for the size and asymmetry of the ornamental trait in males, the homologous trait in females, and a pair of wing traits that are not under sexual selection. We used individuals from two independent populations derived from di erent geographical locations in Malaysia. More than a thousand individuals were assayed for each trait, an order of magnitude larger than previous studies. Measures were made for an unusually broad range of ve stress levels, spanning a 20-fold change in intensity. Our experimental approach is powerful because it allows the simultaneous comparison of responses of sexual and non-sexual traits, under controlled conditions, with a high degree of replication. 265, 2211^ & 1998 The Royal Society Received 5 August 1998 Accepted 24 August 1998

2 2212 P. David and others Male ornament size and asymmetry in stalk-eyed ies We tested whether the size of exaggerated male sexual ornaments (eyestalks) is more sensitive to condition than the size of the homologous trait in females or other nonsexual traits. We then tested whether FA in sexual ornaments showed similar greater sensitivity to condition. Finally, we compared whether di erences in condition were better re ected by trait size than by trait asymmetry. 2. MATERIALS AND METHODS (a) Manipulation of condition Two laboratory populations of C. dalmanni (L and G), with di erent geographic origins in Malaysia, were used to provide independent tests of condition dependence. Both stocks have been maintained in cage culture at 25 8C on a 12 h:12 h light : dark regime with corn food medium replaced on a regular cycle. Since foundation in 1993, the population size was kept high (at least 200 individuals) to minimize inbreeding. We experimentally varied food availability among developing ies to manipulate condition, because this stress is easily controlled and quanti ed, and is likely to be important under natural conditions. Each day, groups of 15 randomly chosen females were allowed to lay eggs for 5 h on damp lter paper. Batches of 13 eggs were placed on one of ve levels of corn (0.03, 0.06, 0.1, 0.25 and 0.5 g per egg) and were cultured until adult ies emerged. Su cient eggs were collected to allow the same number of replicates of each food level to be set up on each day. Over a four-week period, the procedure was repeated seven times (L) and ve times (G). All emerging ies were collected and frozen (840 L and 470 G ies). Later they were sexed before dissection of heads and wings. Although we could not control for maternal genotype, the contribution of di erent females was not a source of bias because maternal genotypes contributing to the stock of experimental eggs were numerous (several hundreds) and eggs were distributed randomly across food levels. Owing to egg^adult mortality, variable numbers of adult ies emerged and we used the number of emerged ies per gram of food as the density estimate (this gave slightly stronger regressions than the number of eggs per gram of food). (b) Measurements Measurements were taken on all experimental ies, using a monocular microscope attached to a camera lucida and digitizing tablet. Eyestalk length and two wing dimensions (corresponding to length and width) were estimated as distances between pairs of easily identi ed landmarks ( gure 1). Two replicates of each measurement were taken on di erent days; a replicate included left and right measures of each individual. Sample sizes were not the same for di erent traits because occasionally specimens exhibited damage to heads or wings. As the experiment could not be set up in a single day, uncontrolled environmental variation could have obscured larval density e ects on trait size and asymmetry. We tested models incorporating day and density e ects. For trait-size measurements, day e ects and day density interactions were usually signi cant, although very small compared to density e ects. For asymmetry measurements, they were never signi cant, with the exception of wing width asymmetry in L females. Therefore density e ects were analysed after pooling all experimental days into a single sample. Figure 1. Stalk-eyed y measurements of (a) eyestalk length and (b) two wing dimensions corresponding to length (x) and width ( y). Our measure of eyestalk used the median bristle rather than the stalk-tip as a landmark because it could be scored more reliably. (c) Statistical analysis of trait size Trait size was calculated as the sum of left and right measurements. The repeatability of measurements was assessed as the correlation coe cient between replicates. Condition dependence of trait size was measured using linear regression of trait size (averaged across replicates) on density. Sex di erences in the e ect of density were estimated as the sex density interactions in a linear model incorporating sex and density e ects. (d) Statistical analysis of trait asymmetry Measurement error is an important source of bias in asymmetry studies, because it introduces random di erences between left (L) and right (R) sides that mimic FA (Palmer & Strobeck 1986; Merila«& Bjo«rklund 1995; Bjo«rklund & Merila«1997). The standard procedure to estimate the amount of error variance compared to FA is a two-way ANOVA described by Palmer (1994). This procedure is appropriate when L and R have independent measurement errors. We cannot be sure of this in our study. Owing to the large sample size, measurements were taken during a number of di erent sessions. Conditions such as the light, the calibration of the optic system, the precise position of the sample and the behaviour of the measurer may have varied from one session to the next. David et al. (1999) have shown that when session biases are suspected the best strategy is to measure left and right sides during the same session and to analyse the signed di erence of left minus right (L7R). To check for measurement error we performed a one-way ANOVA on (L7R), which partitions the total variance into variance between individuals and variance between replicates within individuals. The F-ratio between these two variances measures true FA relative to measurement error

3 Male ornament size and asymmetry in stalk-eyed ies P. David and others 2213 (David et al. 1999). F-ratios were high and very signi cant for each trait, population, and sex, indicating that measurement error did not obscure estimates of FA (proportions of total variance due to measurement error are in square brackets): male eyestalk, F 351,352ˆ11.50 [0.08] (L), F 155,156ˆ17.92 [0.05] (G); male wing length, F 417,418ˆ18.79 [0.05] (L), F 187,188ˆ 4.49 [0.18] (G); male wing width, F 429,430ˆ11.99 [0.08] (L), F 193,194ˆ7.97 [0.11] (G); female eyestalk, F 349,350ˆ17.65 [0.05] (L), F 209,210ˆ 9.60 [0.09] (G); female wing length, F 367,368ˆ12.46 [0.07] (L); F 248,249ˆ 4.48 [0.18] (G); female wing width, F 349,350ˆ5.70 [0.15] (L), F 255,256ˆ6.06 [0.14] (G); all p Individual signed asymmetry values were computed as left minus right, averaged over the two replicates. The e ect of taking the average is to reduce the measurement error variance by half (David et al. 1999). We tested whether the asymmetry values behaved as FA, rather than directional asymmetry or antisymmetry, following the procedure described in Palmer & Strobeck (1986). For all traits, sexes, populations and food treatments the distributions of signed asymmetries were tested for normality (Shapiro %, Wilk's W ) and a mean of zero (onesample t-test). A sequential Bonferroni correction was applied to p-values for each trait. Unsigned asymmetries could not be used directly in parametric tests because of their conspicuously nonnormal distribution. The transformation jleft7rightj 0.4 e ciently normalized our distributions, as suggested by Swaddle et al. (1994). 3. RESULTS (a) Condition and trait size Trait-size measurements had high repeatability (r40.99, p for all traits, sexes and populations). In each sex of both populations, eyestalk length and both wing dimensions decreased signi cantly when larval density increased, showing clear evidence of condition dependence ( gure 2). Variation in condition explained a large fraction of the variance in trait size (r 2ˆ40^68%). Although ies in population G were somewhat smaller for all traits than those in population L, both gave very similar responses to density ( gure 2). Sex di erences in the response of the eyestalk length to increased density were conspicuous (sex density interaction: F 1,713ˆ93.1 (L), F 1,363ˆ81.9 (G), both p ). The slope of the regression of eyestalk length on larval density in males was greater than twice its value for females ( gure 2a). Sex di erences in the response of wing length and wing width to larval density were much smaller, although signi cant in both populations (wing length, F 1,769ˆ18.3, p (L), F 1,435ˆ8.3, pˆ0.004 (G); wing width, F 1,797ˆ33.8, p (L), F 1,446ˆ7.7, pˆ0.006 (G); gure 2b,c). The sex di erence persisted when we examined relative eyestalk length. We used wing size as an index of body size. As wing length and wing width were highly correlated (rˆ0.92 (L) or 0.93 (G), both p50.001), their sum was taken as a single index of wing size. When density increased, the eyestalk to wing ratio decreased sharply in males (regression coe cient, bˆ , F 1,337ˆ133.0 (L); bˆ , F 1,144ˆ116.0 (G); both p50.001), but much less so in females (bˆ , F 1,322ˆ35.4 (L); bˆ , F 1,193ˆ61.1 (G); both p50.001). This implies Figure 2. Regression lines (least-squares) of (a) eyestalk length, (b) wing length, and (c) wing width on larval density, for males (solid line) and females (dashed line) in populations L (dark) and G (light). For the sake of clarity, individual data points are not shown. Statistical tests for the regressions: male eyestalk, F 1,360ˆ369.3 (L), F 1,155ˆ292.6 (G); male wing length, F 1,419ˆ452.2 (L), F 1,186ˆ395.0 (G); male wing width, F 1,428ˆ672.7 (L), F 1,192ˆ346.6 (G); female eyestalk, F 1,353ˆ236.2 (L), F 1,208ˆ414.1 (G); female wing length, F 1,350ˆ289.7 (L), F 1,249ˆ455.1 (G); female wing width, F 1,366ˆ331.4 (L), F 1,254ˆ491.4 (G); all p that male eyestalks are disproportionately a ected by condition, compared with female eyestalks. It is possible that the greater response in male eyestalks is due to allometry. Eyestalk length and wing size are linearly related, but they are not proportional to each other as their regression has a non-zero intercept (in each sex of both populations, all p50.001). So to test whether male eyestalks provide more information about condition than wing size alone we used the following linear model: E ˆ (W) (D) (W D), which partitions variation in eyestalk length (E) into e ects of wing size (W, wing length plus wing width), larval density (D), the interaction between these two variables (WD), a constant () and error (). The signi cance of each term was assessed by subtracting that term from the model and testing the change in explained variance using an F-test (Crawley 1993, p.196). The results are shown in table 1. In each sex of both populations, there was a signi cant wing size e ect (), re ecting the large positive allometry between wing size and eyespan. In males of both populations, eyestalk length responded more negatively to density than predicted using wing size alone, as a signi cant density e ect () was present in addition to the wing size e ect (). In females, the e ect of density was weak and inconsistent across populations (table 1). In addition, there was an interaction e ect in males (positive ); at higher densities, changes in wing size were associated with greater changes in eyestalk length. The same e ect was present in one population of females (L), but was of only marginal signi cance. (b) Condition and trait asymmetry Distributions of left7right di erences (averaged over the two replicates) were normal, with a few exceptions that conspicuously deviated from normality, owing to

4 2214 P. David and others Male ornament size and asymmetry in stalk-eyed ies Table 1. Parameter estimates and tests of the linear model of eyestalk length sample n (W) estimate (p) (D) estimate ( p) (W D) estimate (p) L males (50.001) (50.001) (50.001) L females (50.001) (0.02) (0.03) G males (50.001) (50.001) (50.001) G females (50.001) (0.30) (0.54) outliers. We therefore removed these individuals from the data set (eyestalk FA, 7 out of 1069 removed; wing length FA, 5 out of 1206 removed; wing width FA, none removed). All the remaining distributions were normal. Further FA analyses were conducted excluding outliers. However, repeated statistical tests including outliers made no di erences to the conclusions drawn. All means were not signi cantly di erent from zero, except for a single case of left-biased asymmetry (G females, 0.25 g food level, eyestalk asymmetry). Eyestalk FA, wing width FA and wing length FA each failed to show any response to larval density in either population and in either sex ( gure 3), with the exception of male wing length FA in population G, which, contrary to expectation, slightly decreased with density ( p ˆ 0.03). Size dependence could have obscured responses of trait FA to larval density. Eyestalk FA slightly increased with eyestalk length in males of both populations (b ˆ 0.024, F 1,350ˆ5.47, pˆ0.02 (L); bˆ0.025, F 1,155ˆ5.12, pˆ0.03 (G)). Likewise, wing length FA was dependent on size in males, negatively in population L (b ˆ , F 1,416ˆ4.78, pˆ0.03), and positively in population G (bˆ0.050, F 1,186ˆ5.25, pˆ0.02). But wing width FA in males, and all trait FAs in females, were not size dependent. Where size dependence was signi cant, the e ect of trait size was removed by using residuals of the regression on size. We repeated the original regressions using sizecorrected data, but there was still no evidence of condition dependence in any trait FA ( gure 3). Figure 3. Response of individual unsigned asymmetry (FA) to larval density for (a) eyestalk length, (b) wing length and (c) wing width, showing both sexes and populations (as in gure 2). Data have been jleft7rightj 0.4 -transformed. Data were corrected for size e ects when FA was signi cantly dependent on trait size. Tests for FA dependence on density: male eyestalk, F 1,350ˆ0.04 (L), F 1,154ˆ0.00 (G); male wing length, F 1,416ˆ0.38 (L), F 1,186ˆ0.34 (G); male wing width, F 1,428ˆ0.63 (L), F 1,192ˆ0.40 (G); female eyestalk, F 1,347ˆ0.10 (L), F 1,208ˆ0.21 (G); female wing length, F 1,348ˆ0.71 (L), F 1,247ˆ0.02 (G); female wing width, F 1,366ˆ2.44 (L), F 1,254ˆ0.22 (G); all p Tests without correction for size e ects: male eyestalk FA, F 1,350ˆ1.93 (L), F 1,154ˆ3.30 (G) (both non-signi cant); male wing length, F 1,416ˆ1.07 (L) (nonsigni cant), F 1,186ˆ4.94 (G), p ˆ DISCUSSION Our results support the hypothesis that exaggerated male sexual ornaments (eyestalks) are more sensitive to condition (larval density) than homologous female traits or other non-sexual traits. Strong condition dependence was evident in both the absolute and relative size of male eyestalks. We also showed that male eyestalk length provides additional information to that given by nonsexual traits (wing dimensions). In contrast, the homologous female trait provides no consistent additional information. Several lines of evidence suggest that female C. dalmanni stalk-eyed ies use the absolute and relative size of male eyestalks in their mate choice. Males with larger absolute and larger relative eyespan attract more females in the eld (Wilkinson & Reillo 1994), dummy males with arti- cially exaggerated eyespan are similarly more attractive (Burkhardt & de la Motte 1988), and males with larger absolute and larger relative eyespan get more copulations in controlled laboratory experiments (Wilkinson & Reillo 1994). Although we do not know the exact mechanisms of mate choice, the sex di erence in condition dependence ts with handicap theories of mate choice (Andersson 1986; Pomiankowski 1987; Grafen 1990; Iwasa & Pomiankowski 1994), which predict that females use exaggerated male sexual traits as indicators of male condition. These theoretical studies predict that female preference for exaggerated male sexual traits is favoured when condition has a genetic basis. Our experiments only manipulated phenotypic condition, but it is probable that genetic variation in condition is present in natural populations (Rice 1988; Burt 1995; Rowe & Houle 1996) and has similar consequences for trait expression. This view is supported by arti cial selection experiments which show that relative eyestalk length is heritable in natural populations of C. dalmanni (Wilkinson 1993). We note that condition dependence may not be the only quality that eyestalk length indicates in this species. Male eyespan correlates with the presence of a Y chromosome that is resistant to X-linked meiotic drive (Wilkinson et al. 1998). Therefore, male eyestalk length may synthesize two di erent qualities: condition and the ability to restore unbiased sex ratios. We can reject a second hypothesis that di erences in condition are better re ected by trait asymmetry. In contrast to trait size, no measures of trait FA showed evidence of condition dependence. The only signi cant

5 Male ornament size and asymmetry in stalk-eyed ies P. David and others 2215 regression of FA on larval density (wing length FA in G males) was negative (contrary to the expectation) and could have arisen by chance or indirect e ects of trait size (as it disappears when we corrected for size dependence). This lack of sensitivity was even apparent when the hypothesis was restricted to exaggerated sexual ornaments; FA in male eyestalks showed no response to larval density. Why was there no sensitivity in trait FA to larval density? The lack of response cannot be explained by measurement errors as our estimates of FA showed high repeatability (see ½ 2). Neither did it re ect an insu cient range of conditions because we were able to demonstrate clear di erences in trait size over the densities examined. In natural habitats, C. dalmanni larvae feed on decaying vegetation and it is possible that this has a lower nutritional value and di erent e ect on trait development than the mashed corn used as food in this experiment. However, this seems unlikely as the range of eyespans seen in our experiments is equivalent to that seen under natural conditions (Shillito 1971). Di erential survival could have biased our results if asymmetrical males tend to die before collection. However, there is no indication that mortality increases with larval density in our data set. Moreover, females choose between live males, so if detectable FA di erences are eliminated by early mortality, FA cannot be involved in sexual selection processes. Therefore, technical reasons do not explain why our results contradict the common belief that FA of sexual ornaments increases with stress. There are two lines of evidence for the belief that FA of sexual ornaments is highly sensitive to stress (MÖller & Swaddle 1998). First, comparative studies document higher mean relative asymmetry in sexual ornaments than in non-sexual traits in birds (MÖller & Ho«glund 1991; Dufour & Weatherhead 1996). However, this relationship is not supported by other studies (Balmford et al. 1993; Tomkins & Simmons 1995; Leung & Forbes 1996). There appear to be a number of reasons for variation in mean relative FA among traits, especially the degree of selective constraint on asymmetry (Balmford et al. 1993; Clarke 1998). The second line of evidence comes from comparisons of the e ect of stress on sexual and non-sexual trait FA. In addition to the present study on stalk-eyed ies, two other experimental studies have been performed. In barn swallows (Hirundo rustica), male tail FA increased with experimentally manipulated parasite loads, but female tail FA and non-sexual traits FA in both sexes did not (MÖller 1992). In contrast, in the horned beetle (Onthophagus taurus), male horn FA was not in uenced by food stress, whereas male elytra FA was largest in the fooddeprivation treatment (with marginal signi cance), but female elytra FA showed no response (Hunt & Simmons 1997). The di erences in the patterns of results may be explained by di erences in the constraints on ornament symmetry or in the kind of stress used. For instance, animals may have evolved mechanisms to cope with predictable stresses (e.g. starvation), but they may be less well adapted to novel or unpredictable stresses (e.g. coevolving parasites). In any case, we can reject the hypothesis that condition-dependent asymmetry is a general feature of sexual ornaments. We thank Martin Brookes, C. Doums, Jacques Gianino, Avis James, Paul Watson and Jerry Wilkinson (for y stocks). This research was supported by Research Fellowships (K.F. and A.P.) and a Summer Studentship (D.G.) from the Royal Society, and a Research Grant from NERC. This is publication 98^084 of Institut des Sciences de l'evolution. REFERENCES Andersson, M Evolution of condition-dependent sex ornaments and mating preferences: sexual selection based on viability di erences. Evolution 40, 804^816. Andersson, M Sexual selection. New Jersey: Princeton University Press. Balmford, A., Jones, I. L. & Thomas, A. L. R On avian asymmetry: evidence of natural selection for symmetrical tails and wings in birds. Proc. R. Soc. Lond. B 252, 245^251. Bjo«rklund, M. & Merila«, J Why some measures of uctuating asymmetry are so sensitive to measurement error. Ann. Zool. Fennici 34, 133^137. Burkhardt, D. & de la Motte, I Big antlers are favoured: female choice in stalk-eyed ies (Diptera, Insecta), eld collected harems and laboratory experiments. J. Comp. Physiol. A 162, 649^652. Burt, A Perspectiveöthe evolution of tness. Evolution 49, 1^8. Clarke, G. M The genetic basis of developmental stability. V. Inter- and intra-individual character variation. Heredity 80, 562^567. Crawley, M. J GLIM for ecologists. Oxford: Blackwell Scienti c Publications. Darwin, C The descent of man, and selection in relation to sex. London: John Murray. David, P., Hingle, A., Fowler, K. & Pomiankowski, A Measurement bias and uctuating asymmetry estimates. Anim. Behav. (In the press.) Dufour, K. & Weatherhead, P. J Estimation of organism-wide asymmetry in red-winged blackbirds and its relation to studies of mate selection. Proc. R. Soc. Lond. B 263, 769^775. Grafen, A Biological signals as handicaps. J. Theor. Biol. 144, 517^546. Hunt, J. & Simmons, L. W Patterns of uctuating asymmetry in beetle horns: an experimental examination of the honest signalling hypothesis. Behav. Ecol. Sociobiol. 41, 109^114. Iwasa, Y. & Pomiankowski, A The evolution of mate preferences for multiple handicaps. Evolution 48, 853^867. Iwasa, Y., Pomiankowski, A. & Nee, S The evolution of costly mate preferences. II. The `handicap' principle. Evolution 45, 1431^1442. Leung, B. & Forbes, M. R Fluctuating asymmetry in relation to stress and tness: e ects of trait type as revealed by meta-analysis. Ecoscience 3, 400^413. Merila«, J. & Bjo«rklund, M Fluctuating asymmetry and measurement error. Syst. Biol. 44, 97^101. MÖller, A. P Parasites di erentially increase uctuating asymmetry in secondary sexual characters. J. Evol. Biol. 5, 691^699. MÖller, A. P. & Ho«glund, J Patterns of uctuating asymmetry in avian feather ornaments: implications for models of sexual selection. Proc. R. Soc. Lond. B 245, 1^5. MÖller, A. P. & Pomiankowski, A Fluctuating asymmetry and sexual selection. Genetica 89, 267^279. MÖller, A. P. & Swaddle, J. P Asymmetry, developmental stability and evolution. Oxford University Press. Palmer, A. R Fluctuating asymmetry analyses: a primer. In Developmental instability: its origins and evolutionary implications

6 2216 P. David and others Male ornament size and asymmetry in stalk-eyed ies (ed. T. A. Markow), pp. 336^364. The Netherlands: Kluwer Academic Publishers. Palmer, A. R. & Strobeck, C Fluctuating asymmetry: measurement, analysis, patterns. A. Rev. Ecol. Syst. 17, 391^421. Pomiankowski, A The `handicap principle' does workö sometimes. Proc. R. Soc. Lond. B 127, 123^145. Rice, W. R Heritable variation in tness as a prerequisite for adaptive female choice: the e ect of mutation^selection balance. Evolution 42, 817^820. Rowe, L. & Houle, D The lek paradox and the capture of genetic variance by condition dependent traits. Proc. R. Soc. Lond. B 263, 1415^1421. Shillito, J. F Dimorphism in ies with stalked eyes. Zool. J. Linn. Soc. 50, 297^305. Swaddle, J. P., Witter, M. S. & Cuthill, I. C The analysis of uctuating asymmetry. Anim. Behav. 48, 986^989. Tomkins, J. L. & Simmons, L. W Dimorphisms and uctuating asymmetry in the forceps of male earwigs: no evidence for honest signalling. Proc. R. Soc. Lond. B 259, 89^96. Van Valen, L A study of uctuating asymmetry. Evolution 16, 125^142. Watson, P. J. & Thornhill, R Fluctuating asymmetry and sexual selection. Trends Ecol. Evol. 9, 21^25. Wilkinson, G. S Arti cial sexual selection alters allometry in the stalk-eyed y Cyrtodiopsis dalmanni (Diptera: Diopsidae). Genet. Res. 62, 213^222. Wilkinson, G. S. & Reillo, P. R Female choice response to arti cial selection on an exaggerated male trait in a stalkeyed y. Proc. R. Soc. Lond. B 255, 1^6. Wilkinson, G. S., Presgraves, D. C. & Crymes, L Male eye span in stalk-eyed ies indicates genetic quality by meiotic drive suppression. Nature 391, 276^279.