S08Summary: Primary sex ratios: Variation, causes and consequences
C. (Kate) M. Lessells1 & James S. Quinn2
1Netherlands Institute of Ecology, Heteren, The Netherlands, e-mail lessells@cto.nioo.knaw.nl; 2Department of Biology, McMaster University, Hamilton, Canada, e-mail quinn@mcmaster.ca
Lessells, C.M. & Quinn, J.S. 1999. Primary sex ratios: Variation, causes and consequences. In: Adams, N.J. & Slotow, R.H. (eds) Proc. 22 Int. Ornithol. Congr., Durban: 422-424. Johannesburg: BirdLife South Africa.
Attempts at adaptive explanations of sex ratios began with Darwin, and progressed with Fisher's insights in 1930 on frequency dependence. Nonetheless, Williams was still able to write in 1979 (Proc. R. Soc. Lond. B 205: 567-580) that he could find no support for adaptive sex ratio variation in vertebrates. In the twenty years since then we have seen an explosive increase in the number of examples of sex ratio variation in birds, some of them providing convincing support for adaptive theories (see Bensch 1999. In: Adams, N.J. & Slotow, R.H. (eds) Proc. 22 Int. Ornithol. Congr., Durban: 451-466. Johannesburg: BirdLife South Africa; Yezerinac 1999. In: Adams, N.J. & Slotow, R.H. (eds) Proc. 22 Int. Ornithol. Congr., Durban: 467-482. Johannesburg: BirdLife South Africa). Two developments have provided the impetus for this: first, Trivers and Willard's (1973. Science 179: 90-92) theory of maternal condition-dependent sex ratio adjustment has been generalised as an explanation for modification of sex ratios in relation to many phenotypic and environmental variables. Essentially, parents are selected to modify the sex of offspring produced in relation to any variable that predictably alters the relative fitness gain through sons and daughters. Second, the development of molecular techniques has allowed efficient sexing of individuals early in life (Quinn 1999. In: Adams, N.J. & Slotow, R.H. (eds) Proc. 22 Int. Ornithol. Congr., Durban: 434-449). Johannesburg: BirdLife South Africa.), so that variation in the primary sex ratio can be distinguished from subsequent differential mortality. Early techniques that were developed piecemeal allowing the sexing of single species or groups of closely related species have been replaced by near-universal techniques that can be adapted with little effort to most non-ratite bird species (Griffiths et al. 1998. Mol. Ecol. 7: 1071-1075; Kahn et al. 1998. Auk 115: 1074-1078).
Statistical evidence for sex ratio modification can be sought in three main ways: first, by looking for departure from a binomial distribution of sex ratios (Svensson & Nilsson 1996. Proc. R. Soc. Lond. B 263: 357-361; Westerdahl et al. 1997. Molec. Ecol. 6: 543-548). If all chicks in the population have the same probability of being male, brood sex ratios should follow binomial distributions. Greater (or less) than binomial variance implies that the probability of being male differs between broods (or offspring) (Williams 1979. op.cit.). Second, the repeatability of the sex ratio of broods of particular individuals can be investigated. For instance, the sex ratio of first broods of individual female Great Tits Parus major is weakly, but significantly, repeatable (C.M.Lessells unpubl.). This implies that the offspring of different females do not have the same probability of being male. Third, sex ratio may be shown directly to be correlated with some environmental or phenotypic variable (see Bensch op.cit.; Yezerinac op.cit. for examples). These three statistical approaches may be useful in different circumstances. In general, testing for a correlation with the variable in relation to which sex ratio is modified is most informative in understanding the selection pressures involved, and will also be most statistically powerful because the same amount of variance will be explained with fewer degrees of freedom. However, this approach depends on being able to identify, and measure, the relevant environmental or phenotypic variables. The converse applies to comparing the observed variation in sex ratio with that of a binomial distribution: the test is less statistically powerful, but can potentially detect sex ratio modifications even when the relevant variables cannot be identified. Measuring the repeatability of sex ratios has similar virtues. One example, where the lack of need to identify the relevant phenotypic variable might be exploited, concerns the hypothesis that females modify their sex ratio in relation to the quality of their mate. This might be tested by correlating sex ratio with some characteristic of the male parent (e.g. Ellegren et al. 1996. Proc. Natl Acad. Sci. USA 93: 11723-11728), but failure to find a correlation could always be put down to the relevant phenotypic variable not having been identified. On the other hand, a lack of repeatability of sex ratio across the broods of different females mated to the same male would provide evidence that females did not modify the sex ratio of their broods in relation to any permanent character, including genetic quality, of their mates.
The existence of variation in sex ratio at hatching in relation to environmental and phenotypic variables raises obvious questions as to how this is achieved (Emlen 1997. TREE 12: 291-292 ; Krackow 1999. In: Adams, N.J. & Slotow, R.H. (eds) Proc. 22 Int. Ornithol. Congr., Durban: 425-433. Johannesburg: BirdLife South Africa). The mechanism by which sex is modified in turn affects the sex ratio that is expected because the optimal sex ratio strategy differs when parents can only influence sex ratio by abandoning offspring of the 'wrong' sex after some investment has already been made (Maynard Smith 1980. Behav. Ecol. Sociobiol. 7: 247-251). Females are the heterogametic sex in birds, so modification of sex ratio is presumed to be under their control. Because of the relative timing of meiosis (that determines the sex chromosome carried by the gamete) and ovulation, there are essentially two ways in which females could influence hatching sex ratios. First, is some kind of segregation distortion in which the undesired sex chromosome is banished to the polar body during the first meiotic division. Krackow (op.cit.) argues that this is unlikely because it is incompatible with what is known of the mechanism of meiosis and because there is little or no selection on the sex chromosomes to participate in such segregation distortion. The second mechanism involves discontinuing investment, before or after ovulation, in gametes or zygotes of the 'wrong' sex once their sex is detectable by the mother. Because of the sequential production of follicles by avian mothers, such discontinued investment implies a gap in the laying sequence (or hatching sequence if achieved by killing the gamete or embryo but still laying the egg with the rest of the clutch), except if applied to the first egg of a clutch. Evidence for such a mechanism might therefore be sought by examining the sexes of eggs in relation to their position relative to any laying gaps. However, the predictions are not as straightforward as they might at first appear. For instance, if mothers never abandon more than one gamete or embryo consecutively, either the initial egg is of the correct sex, and laid, or of the incorrect sex, not laid, and replaced by an egg which has an unbiased probability of being each sex. As a result, it is the sex of eggs laid when gaps do not occur, rather than the sex of eggs after gaps, that differs from a random expectation. An alternative approach is to examine the ovaries of females after laying to determine whether the number of eggs laid is equal to or less than the number of matured ovarian follicles. Such an investigation requires a species in which sacrificing statistically adequate sample sizes of females is acceptable, and in which sex ratio biases can be reliably induced experimentally (e.g. Kilner 1998. Animal Behaviour 56: 155-164).
One pattern emerging from empirical studies of sex ratios is that the extent of sex ratio modification found is often rather modest (see studies reviewed by Bensch op.cit.; Yezerinac op.cit.), with only a few studies showing extreme sex ratio modification (Komdeur et al. 1997. Nature 385: 522-525 ; Heinsohn et al. 1997. Proc. R. Soc. Lond. B 264: 1325-1329). Yet Trivers & Willard's (1973. op.cit.) model, modifications of which are often used to explain the sex ratio modifications found, predicts extreme 'bang-bang' strategies. In other words, single-sex broods are expected whenever the relevant environmental or pheontypic variable applies, as it usually will, to the entire brood and not to individual chicks. (One variable that is an obvious exception is position in the hatching sequence in asynchronously hatched broods.) The limited extent of much reported sex ratio modification raises a number of issues. The first, disquieting, possibility is that many of the published cases of sex ratio modification are statistical type I errors, but are not recognised as such because publication bias results in negative results not being published (the so-called 'file drawer problem'; Csada et al. 1996. Oikos 76: 591-593). We urge referees and editors to recognise the importance of publishing negative results, but acknowledge that publication bias will always exist because studies with small sample sizes and statistically significant effects will be publishable, but similar-sized studies without statistically significant effects will be dismissed as having insufficient statistical power. Given this problem, confirmation of reported sex ratio trends in independent data sets is particularly valuable.
A second possible explanation for the lack of the extreme sex ratios predicted by Trivers & Willard's model is that the model omits important factors, and that a more sophisticated model would indeed predict intermediate sex ratios. For instance, inclusion of small costs of modifying sex ratio in models of the evolution of seasonal sex ratio trends in raptors can lead to gradual seasonal trends (Pen et al. in press. Amer. Nat.). Knowledge of the mechanism by which sex ratios are modified may thus have an important effect on the predictions of models of sex ratio evolution.
Lastly, as Hamilton (1967. Science 156: 477-488) has pointed out, mechanisms that allow sex ratio to be modified may be inherently evolutionarily unstable because of the ease with which they might be exploited by selfish genetic elements. The evolution of an efficient mechanism for modifying sex ratio might be followed first by selfish exploitation by some parts of the genome, followed by quashing of the ability to modify sex ratio by other parts of the genome disadvantaged by such exploitation. In this case, efficient sex ratio modification mechanisms may be a transient evolutionary phenomena.