Sexual dimorphism in A. afarensis, again

The Journal of Human Evolution early access section has a paper by J. Michael Plavcan and colleagues that critically examines the case for low sexual dimorphism in A. afarensis.

To catch you up briefly with the story, here is a synopsis to date. Before 2004, the consensus about A. afarensis was that the samples from Laetoli, Maka, Hadar, and other smaller samples belonged to a single species with substantial sexual dimorphism (gorilla-like or orangutan-like in extent) and considerable temporal change from the early to late end of the sequence. A minority view was that there were actually multiple species in the sample. Some thought this because they thought that some of the important specimens (especially AL 288-1, Lucy) had been misidentified as to sex. Others were either unconvinced by the morphological similarities among the samples, or were rightly skeptical about the weakness of the test for sexual dimorphism within species. That is to say, the variation would have to exceed that found in gorillas (with males double the mass of females) before the single-species hypothesis would be rejected.

In 2004, Reno and colleagues added another perspective. They applied a resampling technique to estimate the sexual dimorphism in the A. afarensis sample, transforming the sizes of different skeletal elements to a single scale in order to increase the effective sample size. Their conclusion was that the variability in the A. afarensis sample was most consistent with a low level of sexual dimorphism, similar to humans. They used this observation to suggest that the social behavior of early hominids may have included a more humanlike mating system, consistent with Lovejoy's (1981) account of the origin of bipedalism.

Plavcan and colleagues (2005) present several arguments as to why the conclusions of Reno et al. (2004) may be flawed.

  1. They suggest that the AL 333 sample, upon which Reno and colleagues hinge some of their conclusions, is likely to be biased, In particular, they suggest that there are probably more males than females, and probably many of the elements included as separate individuals by Reno et al. (2004) actually belong to a smaller number of individuals. Plavcan and colleagues suggest that the MNI for the postcranial elements alone at this locality is three adults and one subadult. Their conclusion from this is that estimates based on AL 333 are likely underestimates. However, in my view, this probably does not detract substantially from the results of Reno and colleagues' analysis. Consider that the overall A. afarensis sample was statistically similar to the AL 333 sample alone, and Reno and colleagues attempted to assess the possible effects of sample bias by simulating samples in which one sex was highly overrepresented.
  2. Plavcan and colleagues argue that the non-AL-333 elements of the A. afarensis sample actually show high variation. This observation is not present in Reno and colleagues (2004), who instead present these remains together with the AL 333 locality as a "Combined Afar" sample. This is more of a problem. Reno and colleagues suggest that the "Combined Afar" sample should be more variable than AL 333 because the combined sample includes specimens across a broad time interval, but as Plavcan et al. note, there is actually little variation over time noted (as yet) for this time span. And Plavcan and colleagues provide an illuminating figure that shows that temporal variation in overall size (without differences in sexual dimorphism) does not result in higher variation in sexual dimorphism.
  3. The most critical point raised by Plavcan et al. is that skeletal dimorphism is not well related to body mass dimorphism. Presumably it is body mass dimorphism that has implications for social structure. They apply their own range of comparisons to examine the variation of skeletal dimorphism (in particular femoral head diameter variation) with that of body mass dimorphism. In the end, they conclude that the Afar sample is consistent with a body mass dimorphism greater than that in any human population that they examine, and between that of chimpanzees and gorillas.
  4. Finally, Plavcan and colleagues question the premise that social behavior can be inferred from the level of sexual dimorphism. This problem is well-known, and they give little detail, but the argument is well-taken.

I think the bottom line coming out of this argument is that there really isn't enough to infer much about the level of dimorphism of A. afarensis. For reasons of my own, I think the template method used by Reno and colleagues is problematic. Without this, there are only a handful of specimens for any single skeletal element that can be compared. This is probably a sufficient sample to determine whether humanlike skeletal dimorphism overall can be rejected, but not enough to examine the relationships of dimorphism for different skeletal elements, and probably not enough to infer mass dimorphism.

And the problems with inferring behavior from dimorphism come not from the fossils, but from the comparisons available among living species. Until a model can account for the dimorphism among living hominoids based on their social behavior, there is certainly no point in trying to infer the behavior of ancient hominids in this way. Social groups in all hominoids are flexible to some extent, and in humans they are extensively so. This bodes poorly for a resolution of this problem.