0 Birkhiiuser Verlag, Basel, 1997 J. evol. biol. IO (1997) 1010-061X/97/010051-05
51-55
1Journal
$l.50+0.20/0
of Evolutionary
Biology
Commentary
Genetic variation
in fluctuating
asymmetry
A. Pomiankowski The G&ton Laboratory, Depurtment of Biology, University College London, 4 Stephenson Way, London NW1 2HE, UK, e-mail:
[email protected] Key words:
Fluctuating asymmetry; coefficient of variation.
Introduction The study of fluctuating asymmetry has recently become a topical subject in the biological sciences(Markow, 1994). Fluctuating asymmetry (FA) is widely thought to reveal something deep about the relationship between genotype and environment. One of the major claims is that FA is a sensitive indicator of both genetic (e.g., inbreeding, hybridisation, spread of new mutants) and environmental (e.g., pH, temperature, pollutants) stress (review Parsons, 1990). Another is that the degree of fluctuating asymmetry closely reflects fitness (Clarke, 1995). These ideas have led to the promotion of FA as a useful measure of well-being at both the individual and population levels (Leary and Allendorf, 1989). But the link between FA and stress(or fitness) is not as consistent as many reviewers suggest.There is plenty of circumstantial evidence but few definitive experiments. For example, despite correlational evidence, uncertainty still surrounds the suggestion that FA increases with inbreeding and the degree of homozygosity. In this case, carfully controlled and well-replicated experiments have not shown an association of FA with inbreeding (Clarke et al., 1992; Fowler and Whitlock, 1994). There is also practically no unambiguous evidence that FA is a more sensitive measureof stressthan other measureslike trait size (an exception is M@ller, 1992). This experimental deficit does not mean that FA is not a useful index. For instance a substantial body of evidence has now accumulated which shows that FA in male sexual traits is indicative of condition and is used by females in their mate choice (Mnller, 1994; Swaddle and Cuthill, 1994a; 1994b). However not all sexual traits behave in the same manner (Mnller and Pomiankowski, 1993; Tomkins and Simmons, 1995). It appears as if the utility of FA, like other measures(e.g., size, variance), varies with the functional importance and history of selection on a trait. These qualifications need to be borne in mind when extrapolating from individual and population measuresof FA. 51
52 The heritability
Pomiankowski
of FA
Why is the heritability of fluctuating asymmetry interesting? We can expect FA to respond to selection if the heritability of FA is non-zero. The short-term response is predicted by the heritability which measures the proportion of phenotypic variation under genetic control. A non-zero heritability is also a necessary condition for some hypotheses linking the level of FA to genetic stress. Moller and Thornhill (1997) have done a good service by tabulating all attempts to measure the heritability of FA and showing that nearly all report a positive value. The reverse result would have been truly remarkable. Moller and Thornhill (1997) also attempt through meta-analysis to uncover the normal value of FA heritability. Their estimate is h2 = 0.27 (mean) or hZ = 0.21 (median) for 14 species. This estimate is far more problematic than the demonstration that heritability is on average non-zero. I have three brief comments to make. First meta-analysis is not a strictly suitable technique for estimating the value of FA heritability. Meta-analysis is a statistical technique designed to combine estimates made from the same underlying distribution (Arnqvist and Wooster, 1995). It has frequently been used to overcome the problem of small sample sizes and contradictory findings in individual studies. But it seems to be stretching the point to claim that the FA heritability data set are estimates made from the same underlying distribution. The data set contains different traits, from a variety of species analysed under various conditions. Moller and Thornhill’s (1997) point is not really to estimate the true value of FA heritability across all traits in all species but to compare the heritability of FA with the heritability of other traits. They claim that their mean estimate h2 = 0.27 is a small value, This conclusion would be more robust if it was compared to estimates of the heritability of other traits in the same species. In particular, comparison could easily be made with trait size in many of the reported studies. Such a comparison would have the added advantage that measurements of size and asymmetry were taken from the same individuals under the same conditions (thereby removing a number of sources of it2 variation). A second problem is that estimates of FA heritability are undoubtedly underestimates because of the inherent difficulty in measuring FA. This can be seen by comparing measurements of FA with those of trait size. The measurement error in FA compounds the measurement errors of left and right sides, This will increase the residual variance term in the estimate of FA heritability. In addition, differences between left and right usually contribute only a fraction (
