Dmanisi 'single species' claim draws critics

Last week Science printed an exchange of technical comments on the topic of the Dmanisi skull 5. The skull was described in a paper last fall (Lordkipanidze et al. 2013), which I blogged here: "The new skull from Dmanisi". The skull is beautiful and provides an almost-unprecedented look at undistorted and unreconstructed cranial form.

Now, Jeffrey Schwartz, Ian Tattersall and Zhang Chi (2014) have challenged the interpretation of the Dmanisi sample. The most provocative aspect of Lordkipanidze and colleagues' 2013 paper was its argument that the variation within Early Pleistocene Homo should all be collapsed into a single species, Homo erectus. Schwartz and colleagues think this is wrong. Instead, they think that Skull 5 represents a different species within the Dmanisi sample. Zollikofer and colleagues (2014) provide a response, pointing out that Schwartz and coworkers here and elsewhere have argued that the five Dmanisi skulls represent as many as four different species.

Morphometric shape and species

Lordkipanidze and colleagues (2013) rested their argument on a multivariate comparison of cranial shape. Their paper included the following figure, which presents a simplified view of the shape differences separating chimpanzees, modern humans, and some fossil hominins:

Figure 4, shape variation from Lordkipanidze et al 2013
Figure 4 from Lordkipanidze et al. 2013, showing the first and second "shape components" from their comparison of chimpanzee and hominin crania.

I discussed this figure in my previous post. The paper expresses two related arguments. First, the authors argued that the variation encompassed by the five Dmanisi crania is no greater than that encompassed by a large sample of modern humans, or of modern chimpanzees. The figure shows that the scatter of chimpanzee crania and the scatter of human crania are both fairly extensive, and the four plotted Dmanisi crania do not exceed that scatter.

Second, the authors made a more controversial claim. They observed that the variation of the Dmanisi crania in their shape comparisons actually encompasses the variation within all the samples of Homo erectus, Homo habilis and Homo rudolfensis. That led them to conclude that the null hypothesis that all these Early Pleistocene crania represent a single species. Homo habilis and Homo rudolfensis simply do not exist under this scenario. As I related last fall, Adam Van Arsdale and Milford Wolpoff (2013) presented a similar argument, testing it on the basis of change over time within the Early Pleistocene Homo sample.

There's no denying that skull 5, which includes the D4500 calvarium and D2600 mandible, reflects a combination of features that anthropologists did not expect to find. Its endocranial volume, at only 546 ml, is smaller than any other known Homo erectus adult. It is relatively small even compared to the sample of crania currently attributed to Homo habilis, and a full 200 ml smaller than KNM-ER 1470, the only complete cranial vault that anyone attributes to Homo rudolfensis. Yet despite its small size it shows marked cranial and mandibular robusticity, has very large teeth -- and especially enlarged second and third molars. If this skull had been found first, maybe we would be debating whether it would be a better fit in Homo habilis than Homo erectus. And yet, it shares many morphological features with Homo erectus skulls that have never been found in Homo habilis.

Facing the challenge

I'm a lumper by nature, as long-time readers know. Yet it is easy to see one problem with the analysis presented by Lordkipanidze and colleagues last fall. What is to stop us from lumping A. africanus in with early Homo as well?

As you can see from the figure above, if we define the range of shape variation in either chimpanzees or modern humans, Sts 5 and Sts 71 fall well within the range of variation that includes the early Homo specimens.

A group of the researchers on the Malapa project and I have gotten together to look at where the MH 1 skull would fit on this graph. It should be no surprise that it can be placed within the early Homo sample as well. Apparently it's not only Homo habilis that doesn't exist -- Australopithecus sediba should be Homo erectus as well.

We prepared a short paper to examine this curious result, focusing on how the gross cranial shape really does not recreate the taxonomic groups usually accepted by anthropologists. Let me point out that replicating this kind of analysis is very difficult, considering that the underlying CT data are not available for study, and the coordinate data from the morphometric analysis are not provided as supplements to the paper. In particular, when the dataset includes CT-based reconstructions, it is impossible to verify the measurements against the original specimens unless the CT reconstructions are available as datasets.

I have complained before about supplementary information packets from papers not providing measurements. A great example is table S6 in Lordkipanidze et al. 2014, which has a list of landmarks along with a "1" or a "0" depending on whether the landmark was included in the analysis.

This is reminiscent of the famous "data table without data" from Suwa and colleagues' 2009 study of Ardipithecus teeth (my post, "Whoa, who stole the data?"). I asked at the time, "What kind of rinky-dink journal is this?" Science obviously has not changed its review practices in the succeeding five years.

Schwartz and colleagues' argument

Fundamentally, Schwartz and coworkers present a very different argument than we have followed in our work. They present two pictures, and write beneath both, "Note the obvious morphological differences (not to scale)." Here's one of them:

figure from Schwartz et al illustrating ten skulls of Homo erectus
Figure 2 from Schwartz et al. 2014. Original caption: Array of specimens commonly deemed H. erectus. The type specimen is Trinil 2 (Java), and the only specimens similar to it in derived features are from Sangiran (Java). Otherwise, note the obvious morphological differences, not only between the Dmanisi specimens but also in the assemblage as a whole. ZKD, Zhoukoudian; ER, East Turkana; WT, West Turkana; r, reversed (not to scale).

I mean, how does this figure get through peer review? The caption simply says, "note the obvious morphological differences". It presents a series of ten crania "not to scale", and does not denote any differences for us to observe. The other figure, figure 1, shows the three Dmanisi mandibles with teeth, and bears the caption, "Note the obvious morphological differences in bone and tooth morphology (not to scale)."

We could easily do the same with a series of ten human crania, or a series of ten chimpanzee crania. "Look for yourself how different they are!" is fundamentally unconvincing. Especially when the different specimens are shown at different scales.

Aside from the two figures, the comment focuses on two issues: the invalidity of using overall shape comparisons, and the specific morphological differences among the Dmanisi specimens. Schwartz and colleagues begin by criticizing Lordkipanidze and coworkers for their "assumption" that the Dmanisi sample represents a single species.

The Dmanisi fossils are assumed to sample a single population, primarily because they come from the same site and a relatively short time period. By a priori defining “difference” as intraspecies “variation,” this permits focusing uniquely on general shape and gross morphology.

They imply that if we instead assumed a priori that Dmanisi includes multiple species, there would be no basis to use the variation in shape at the site for any further comparisons. They go on to recite a list of differences among the mandibular and cranial specimens at the site, in each case claiming that the differences are "potentially species-distinguishing features".

The reply

Christoph Zollikofer and colleagues (including most of the authors from the paper last fall) present a reply to Schwartz and coworkers' comment. They begin with some sarcasm:

According to these authors, Dmanisi would now comprise at least four different hominid taxa and thus hold the world record in hominid paleospecies diversity documented at a single site that extends over a mere 40 m2, and probably over a mere couple of centuries.

They then point out that the small area and apparently short time represented by the Dmanisi fossil sample is not sufficient to consider the entire sample as representing a single species. To the time and space element, they add that the sample does not exceed the variation within demes of living humans or other primates.

Through several examples, Zollikofer and colleagues devastate the anatomical case pitched by Schwartz et al. For example:

  1. Schwartz et al.’s assertion that premolar root number has “taxonomic valence” ignores basic comparative evidence from extant populations. Modern human sub-Saharan populations exhibit the full range of root variation, from single and Tomes’ roots to mesiobuccal + distal and buccal + two lingual roots (9). Variation in root morphology is also observed in the third mandibular premolars [P3s in (2)] of Pan troglodytes verus (10).

A trait that is variable in humans, and variable within other species of primates, is probably a bad trait for distinguishing hominin species. That's not a sophisticated argument, it's just basic taxonomic practice. And consider:

  1. Quantifying buccolingual compression of the mandibular canines (11) shows that D2735 has buccolingual (BL)/mesiodistal (MD) ratios of 0.88 and 0.90 (left and right canines, respectively), and D211 has ratios of 0.96 for both canines. D2600 has ratios of 0.86 (left) and 1.12 (right). Evidently, the two sides of the latter mandible do not represent two different taxa.

There seems to have been a lot of this lately, taking two parts of a single specimen and saying that they represent two different species.

Obviously, Zollikofer and colleagues get the better end of this exchange. Schwartz, Tattersall and Zhang seem to have been highly motivated to write their comment because of previous assertions about the great taxonomic diversity of Dmanisi in particular and early Homo in general. But they are the most extreme splitters in the game. They ignore the variation within samples of living primates, and explicitly minimize the importance of within-species comparisons. Their position may be consistent, but in the face of the comparative data it is incoherent.

Concluding thoughts

That doesn't mean I think that Zollikofer and colleagues are entirely correct. Dmanisi is an instructive example about the variation within a sample of fossil hominins, but that doesn't necessarily mean we should discard the distinctions between East African fossil specimens.

As Van Arsdale and Wolpoff (2013) have shown, those distinctions have their own problems. I pointed out last fall ("The new skull from Dmanisi") that what we now know about the widespread intermixture of Middle and Late Pleistocene lineages should begin to change the way we think about the Early Pleistocene. That doesn't mean we should erase all distinctions, but the evolutionary pattern does not have to be a simple branching tree.

The evidence from Dmanisi will be stronger when the cranial and postcranial evidence can be integrated into a single picture. The Malapa hominins have shown how evidence from across the skeleton can make the comparisons of different samples much more complicated.

That means its time to re-evaluate the evidence with a combination of approaches and the full set of data.


Lordkipanidze D, Ponce de León MS, Margvelashvili A, Rak Y, Rightmire GP, Vekua A, Zollikofer CPE. 2013. A Complete Skull from Dmanisi, Georgia, and the Evolutionary Biology of Early Homo. Science 18 October 2013: 342 (6156), 326-331. doi:10.1126/science.1238484

Schwartz JH, Tattersall I, Zhang C. 2014. Comment on "A Complete Skull from Dmanisi, Georgia, and the Evolutionary Biology of Early Homo". Science 344:360. doi:10.1126/science.1250056

Van Arsdale AP, Wolpoff MH. 2013. A single lineage in Early Pleistocene Homo: Size variation continuity in Early Pleistocene Homo crania from East Africa and Georgia. Evolution 67 (3), 841-850. doi:10.1111/j.1558-5646.2012.01824.x

Zollikofer CPE, Ponce de León MS, Margvelashvili A, Rightmire GP, Lordkipanidze D. 2014. Response to Comment on "A Complete Skull from Dmanisi, Georgia, and the Evolutionary Biology of Early Homo". Science 344:360. doi:10.1126/science.1250081