Monday, February 05, 2007

Fluctuating Asymmetry: honest truth and ballyhoo, part 1   posted by agnostic @ 2/05/2007 02:39:00 AM

To kick off GNXP's contribution to the Week of Science, during which participating bloggers will write only about science rather than fend off creationists and the like, the present entry will discuss the reality and hype surrounding the concept of Fluctuating Asymmetry (FA). Basically, it is the random discrepancy from perfect symmetry in a trait that is expected to be symmetrical, e.g. the length of your left ear compared to your right. FA excludes Directional Asymmetry, whereby individuals tend to depart from symmetry in a certain direction; for example, the right testicle tends to be larger than the left one in humans. FA also excludes Antisymmetry, whereby a population approaches bi-modality, with one sub-pop departing from symmetry in one direction and another sub-pop departing in the other direction: for example, handedness in humans.

I'll restrict my discussion here to FA, the random minute variance in values of a trait between the left and right sides (though there is also radial symmetry), which some researchers believe reflects Developmental Instability (DI), or random perturbations in development that prevent the developmental system from reaching its target phenotype in the individual. As I see it, researchers interested in this FA-DI relationship fall primarily into three camps: 1) ecologists studying environmental stress, who investigate whether greater levels of stress are reflected in greater levels of FA in the population; 2) quantitative geneticists who investigate the genetic architecture of FA; and 3) evolutionary biologists studying sexual selection, who investigate whether individuals use a prospective mate's FA to infer the extent of said mate's DI, typically on the assumption that individuals desire those who have "good genes" and who can thus better resist DI. Others are interested in the topic, to be sure, but I believe these three groups constitute the majority.

You could spend the rest of your career studying just one aspect of this phenomenon, but part of the value of a blog like ours is -- I hope -- that we can cover the main points of an interesting issue, with pointers to the literature, without requiring weekly attendance at a seminar. And unlike literature or book reviews in journals, you don't have to pay us $30 just to read it! (Although you're certainly welcome to.) This clearly is no replacement for studying the material in detail, but it's a better way of letting a target audience know what important other ideas there are outside of their narrow specialty, in the interests of cross-fertilization of ideas. In this spirit, I've summarized 4 important take-home lessons from my reading of the FA-DI literature. In brief, they are: 1) the link between FA and DI is far from clear; 2) the heritability (h^2) of FA is very low, though epistatic effects may be non-trivial; 3) FA is likely of minor importance at best in matters related to sexual selection; and 4) it can reveal environmental stress. I'll write about the first two today and the second two at a later date in the Week of Science.

Before moving on, it's worth providing links to two great websites on the FA and DI literature by leading participants in such research: Palmer's website on FA is pretty diverse and has a great bibliography section, which includes many free PDFs of the author's work (including a nice satirical review -- Palmer & Hammond 2000); and Leamy's website also has several free PDFs of his work. Reading the review PDFs there will give you a good hint of what the consensus is, though if you're really interested in this topic, Polak (2003) is required reading since its chapters cover the topic from a variety of angles (measurement of FA, the quantitative genetics of FA and DI, and so on).

Point 1: The FA-DI relationship, if it exists, is very poorly understood

If readers take away only one point, it must be that claims that FA is a reasonably precise measure of DI, or that the relationship between FA and DI is suitably understood, or even that DI exists, should be taken with a large grain of salt. We know FA exists, and as long as we're very careful -- i.e., take repeated measures, since FA is typically within the range of measurement error -- we can actually measure it. But as for accurately measuring DI, here is one phrasing of the consensus, from Leamy & Klingenberg (2005, p. 3; free PDF):

Because FA is calculated from only two sides (one degree of freedom) of a bilateral character in a given individual, it is a rather imprecise estimator of DI variability. Thus, sampling variation should result in many FA values at or near zero (the expected mean for FA) even in individuals with very high levels of DI. The correspondence between FA and DI has been parameterized by the hypothetical repeatability, [R], which expresses the degree to which differences in FA reflect differences in DI. Theoretical formulas for calculating [R] that are based on variation in FA have been derived by Whitlock (1996, 1998) and Van Dongen (1998), and estimates of [R] from a number of FA studies have averaged about 0.08 (Gangestad & Thornhill 2003). This very low value serves as an important reminder of the difficulty of obtaining good estimates of DI in populations from measurement of FA in various characters.

(NB: I've replaced the typographically awkward R with a normal R in brackets. Their list of references is contained in the free PDF linked to above.) So, when we read statements such as the following from Thoma et al. (2006, p. 1457), we should be aware that the qualitatively true statements are misleadingly omitting any mention that FA is a poor measure of DI, a phenomenon whose existence remains muddied:

A common marker of DI is obtained through the measurement of fluctuating asymmetry. . . As a marker of DI, fluctuating asymmetry reflects the interaction of stabilizing developmental processes with disruptive, random environmental (i.e., exposure to toxins) and genetic (i.e., nonadaptive mutations) processes. . .

In two separate studies, Furlow et al. (1997) showed fluctuating asymmetry to be negatively correlated with a measure of fluid intelligence, the Cattell Culture Fair Intelligence Test. Several recent studies have replicated and extended these results (Bates, 2004; Prokosch et al., 2005; Thoma et al., 2005), supporting a link between DI and performance on standard intelligence tests and g estimates. This perspective raises the important possibility that genetic variation in susceptibility to early (perhaps prenatal) environmental influences adversely affects the development of human intelligence.

Hold on there! Now, if this had appeared in an article geared toward those already somewhat familiar with FA and DI, the omission of qualifiers might be justifiable, but this article is geared toward psychologists (it appeared in a neuroscience journal), so all of the qualifiers should appear. Not only is FA a dicey measure of DI, but we don't know that DI is non-trivially heritable either, and our best guess for the heritability of FA is that it's close to 0 (see Point 2 below). The murkiness of the issues surrounding FA and DI should temper the exuberance typified in the quoted text above. One of the reasons I've selected this topic for the Week of Science is that those unfamiliar with the consensus in the FA and DI literature -- such as most professional psychologists and anthropologists, as well as close to all journalists -- could be easily seduced into believing that FA offers far greater explanatory power than it truly does for their object of inquiry (e.g., human intelligence). Given that the article quoted from above was published in 2006, it is clear that non-specialists must still be wary of confident claims that FA is important in explaining some property they're interested in.

Point 2: Additive genetic variance in FA is very low

If I, as a non-expert, had to guess what event set off the "spring cleaning" in the FA-DI literature that culminated in an edited volume (Polak 2003), I would say it was a controversial meta-analysis of the heritability of FA (Moeller & Thornhill 1997*), which elicited a large number of articles in reaction just within the issue in which it was published. The authors claimed to have found a modest h^2 of 0.19, although later meta-analyses and literature reviews pointed to an h^2 surely less than 0.1, and typically around 0. For example, Table 11.1 of Fuller & Houle (2003) is a review of the literature on the heritabilty of FA and (by adjusting that by R) the heritability of DI. They found a mean h^2 for FA of 0.026, and for DI of 0.255, although the interpretation of the latter result must reflect the fact that the estimates of R were pretty erratic, commonly resulting in an h^2 for DI that lay outside of [0, 1], which is meaningless in context. Thus, there is likely something inappropriate about calculating h^2 for DI in this way.

That is not to say that genetic variation plays no role -- indeed, QTL research by Leamy and colleagues has shown that, at least for FA in the jaw morphology of mice, statistical epistasis contributes substantially to FA differences (Leamy et al. 2005; free text). Nevertheless, our imperfect but best guess is that FA probably will probably not respond to selection due to such low additive variance, a point to which we return when discussing the use of FA as a "fitness indicator" in sexual selection theories. Since we know so little about how to fruitfully infer DI from FA, the former may end up showing an h^2 in the modest or strong range, which would facilitate response to selection. Again, the enthusiasm will have to wait until our understanding of DI rises above the level of head-scratching.

In Part 2, I'll address the uses that FA has in studying sexual selection (including hunting for "fitness indicators") and environmental stress on natural populations.

*I spell the surname as "Moeller" for typographical ease, although the first syllable's vowel is just the o with a slash through it.


Fuller, R & D Houle (2003). Inheritance of developmental instability. In M Polak (Ed.), Developmental instability: Causes and consequences. New York: Oxford University Press, pp. 157-183.

Leamy, LJ & CP Klingenberg (2005). The genetics and evolution of fluctuating asymmetry. Annual Review of Ecology, Evolution, and Systematics, 36, 1-21.

Leamy, LJ, MS Workman, EJ Routman, & JM Cheverud (2005). An epistatic genetic basis for fluctuating asymmetry of tooth size and shape in mice. Heredity, 94, 316–325.

Moeller, AP & R Thornhill (1997). A meta-analysis of the heritability of developmental stability. Journal of Evolutionar Biology, 10, 1-16.

Palmer, AR & LM Hammond (2000). The Emperor's codpiece: A post-modern perspective on biological asymmetries. International Society for Behavioral Ecology Newsletter, 12, 13-20.

Polak, M (2003). Developmental instability: Causes and consequences. New York: Oxford University Press.

Thoma, RJ, RA Yeo, S Gangestad, E Halgren, J Davis, KM Paulson, & JD Lewine (2006). Developmental instability and the neural dynamics of the speed-intelligence relationship. NeuroImage, 32, 1456-64.