Lost in the Mail

By Bernard Wood
December 11, 2013

Thank you for asking me to comment on the attached manuscript by Lorkipanidze et al. entitled ‘A complete skull from Dmanisi, Georgia, and the evolutionary biology of early Homo.’

It reports an adult hominin cranium, D4500 (aka skull 5), recovered in 2005 from layer B1y in Block 2 at Dmanisi. There are sound grounds for assuming the ‘new’ cranium and the previously described mandible, D2600, belong to the same individual. There is also some circumstantial evidence that skull 5 and the adult postcranial skeleton (there are too many ‘D’ numbered elements to list here) reported by Lorkipanidze et al. (2007) belong to the same individual.

I go through the manuscript in more detail below, but my “executive summary” of the points made by the authors is as follows: A) the Dmanisi crania and skulls sample a single hominin taxon; B) according to the authors the Dmanisi crania and skulls collectively record a range of cranial and mandibular morphology that either equals, or exceeds, the sum of the morphological variation observed within and among Homo habilis sensu stricto, Homo rudolfensis, Homo ergaster and Homo erectus sensu stricto; C) ergo any taxonomic distinctions that have been made over the years have been effectively falsified by the Dmanisi paleodeme; D) thus all the material listed in B) should be subsumed into the senior taxon, Homo erectus; E) presumably as (although this is not stated explicitly) Homo erectus habilis (ex-Homo habilis sensu stricto), Homo erectus rudolfensis (ex-Homo rudolfensis), Homo erectus ergaster (ex-Homo ergaster) (N.B., this is explicit) and Homo erectus erectus (ex-non-early African Homo erectus). The Dmanisi sample of the expanded H. erectus hypodigm is placed in Homo erectus ergaster georgicus (N.B., The ICZN disavows responsibility for nomina below the subspecies level. This is an implicit warning NOT to use them, as there would be no regulation whatsoever over a proliferation of taxon names for every variant in the fossil record perceived by someone to have importance).

The manuscript begins with a careful description of the new cranium. Notable among the many points about D4500 are its small endocranial volume (c. 546 cm3), the healed fracture in the right zygomatic arch and the striations on the non-occlusal surface of the upper and lower incisors. If the postcranial skeleton does belong to the same individual then the low value of the estimated EQ (c.2.4) is noteworthy because it is in the range of australopiths (i.e., archaic hominins).

The second, more analytical, section of the manuscript begins with qualitative comparisons of the morphology of the Dmanisi crania with well-preserved early Homo crania from eastern and southern Africa. These are followed by what is in many ways the crux of the manuscript, namely the results of a geometric morphometric (GM) analysis of the 48 landmarks listed in Table S6, plus a description and qualitative analysis of a set of qualitative data. Thirty-five of the landmarks (15 midline and 20 bilateral) are on the cranium, and 13 (4 midline and 9 bilateral) are on the mandible. The results of this analysis as they apply to cranial and calvarial shape are visualized in Figs. 4 and S7B, respectively (N.B., Fig. 4 and S7A are effectively plots of the same data). These results convinced the authors that the combined variation in shape seen in Homo habilis sensu stricto, Homo rudolfensis, Homo ergaster and Homo erectus sensu stricto is less than the variation observed in the Dmanisi sample (i.e., skulls 2-5).

I have comments about the data, the analyses (efficacy, results and presentation), and the logic that underlies the taxonomic inferences that are made on the basis of the results of the analyses. But before I offer my critique and suggestions let me first sincerely congratulate the field team at Dmanisi on this new evidence. I have had the pleasure and privilege of spending time at the site and I am conscious that this new evidence is the result of years and years of effort to keep the research at Dmanisi going under sometimes very difficult circumstances. The information content of the new fossil evidence is also much richer because of the painstaking research by Reid Ferring and colleagues that has provided the temporal and other contextual information about the hominins recovered from Dmanisi. It is this context that makes these fossils especially valuable.


Data

The significance of the results of the GM study depends crucially on how reliably and precisely the 15 midline and 20 bilateral cranial landmarks, and the 4 midline and 9 bilateral mandibular landmarks, can be located. It also depends on how reliably the qualitative characters were scored. The comments below focus on the landmark data, but many of my general concerns apply equally well to the qualitative character analysis.

First, with respect to the GM study I would need to be convinced by validation studies that it is appropriate to assume that A) landmark data collected on CT scans and casts (see S1) are as reliable as landmark data collected on the original specimens, and that B) landmark data collected on CT scans and casts are equivalent enough to be used in the same analysis. This could be investigated by collecting, for a handful of specimens, the same landmarks on the original specimen, on a CT rendering and on a cast. As far as I can see from S1 none of the observations used in the GM analysis were collected on the original specimens.

Second, some of the specimens have been subjected to virtual reconstruction, but we are given no details about some of those reconstructions. Were they just corrected for displacement (as the authors suggest is the case for KNM-ER 1805, OH 24 and SK 847), or were better-preserved specimens used as templates for the reconstructions? If so, which specimens informed these reconstructions? How much uncertainty is there in these reconstructions, and how was that uncertainty reflected in the analyses? How do the unreconstructed and reconstructed versions compare in the analyses? Does reconstruction make any significant difference (e.g., is the location of KNM-ER 1805 influenced by its reconstruction)?

Third, the cranial and mandibular landmarks were designed to discriminate among modern human populations, and some of them (e.g., frontal eminence) are seen only in modern humans. I realize that semi-landmarks were used in specimens where the frontal bone is evenly curved, but it is not clear how equivalent/homologous such ‘landmarks’ are among modern humans, pre-modern Homo and archaic hominins? Can the authors provide reassurances about equivalency/homology?

Fourth, how was the GM method (i.e., the landmark location) validated? Did the same observer collect all of the landmark data for all of the specimens? If not, was an inter-observer study of reliability undertaken? If it was the same observer, did that observer undertake an intra-observer study of reliability? I ask this because my experience with the East African material is that some of the landmarks used in the study are either almost impossible to locate reliably (e.g., pterion) or their location is complicated by extra ossicles (e.g., asterion) so that any comparison among specimens is unreliable. How many times were these landmark locations repeated? My experience with locating landmarks on the East African fossil record is that if three attempts resulted in similar locations then that would count as ‘reliable’ enough to be considered as ‘data’ that usefully characterized that specimen. What ‘criterion for inclusion’ was used in the present analysis?


Analysis

First, in general I am puzzled why, when the Dmanisi crania are so well preserved, the crucial analysis uses only landmark data and GM methods (the inferences about the non-metrical data flow from the GM analysis). GM has many strengths, but it also has a number of weaknesses. My sense is that if one were to take an equivalent number of landmarks on a Ford Crown Victoria and a top-end Mercedes one would conclude that they are essentially the ‘same’ car. But such an analysis would ignore a host of crucial evidence that would show conclusively that they are not the ‘same’ car. Why did the authors not include variables (these are a well-covered in the appropriate literature) that have been shown to distinguish early Homo taxa?

Second, the results and their presentation puzzle me. Although in the SI the authors suggest that all of the 15 midline and 20 bilateral cranial landmarks were used in the analyses illustrated in Figs. 4 and S7B, I find this difficult to believe. The reason for this is that I know from familiarity with the African material that several of the landmarks are missing from the fossils they include in the analysis. Was the number used the maximum number that could be located on the fossil specimens plotted in Figs. 4 and S7B? If not, how were the missing landmarks located? It is difficult to assess the significance of these plots without this information.

Third, how much of the overall variation is captured by SC1 and SC2 in Fig. 4, which is so crucial for the authors’ thesis? In 5.2 in the SI, with respect to Fig. S7, the authors write “A. africanus is separated from the fossil Homo along higher-order shape components”. Might that also be the case among fossil Homo in Fig. 4? In other words there might be more distance between the early fossil Homo specimens than is captured in SC1 and SC2. All this could be taken beyond conjecture if the authors provided the proportion of shape variation in the entire data set captured by SC1 and SC2 as is typically done when reporting the results of principal components analyses. If they also provided the multivariate distances between individual specimens this would help readers interpret the partial view of the relationships presented in Fig. 4.

Fourth, when the authors considered the significance of the results illustrated in Fig. 4 and S7B, what conclusions did they draw from the fact that in Fig. 4, according to their logic, Dmanisi skull 4 and an unidentified Neanderthal cranium would not only belong to the same deme but to the same sex? Or that Sts 71 and KNM-WT 15000, or Sts 5 and KNM-ER 3733 are equally close? Did it not strike them that the juxtaposition of these very different crania in this plot suggests that the method they used may not be effective at sorting early Homo fossil crania? Admittedly their method is able to sort modern humans from chimpanzees, but to return to my motor vehicle metaphor, being able to sort cars from pick-up trucks does not mean the same method would necessarily be effective at sorting makes of SUVs.

Fifth, the authors write that the gnathic morphology of skull 4 (i.e., D3444/D3900) “was strongly modified by alveolar bone resorption after tooth loss” (p. 327) and they suggest that it suffers from “massive maxilla-alveolar in-vivo modification” (SI: S.6 p. 5) yet they include it in some analyses (Fig. 4), but not in others (S.6). Why?


Logic

Even if we assume for the sake of argument that the claims made in the manuscript on the basis of the GM analysis are correct, I have concerns about the logic used in connection with the taxonomic suggestions.

First, it is evident that if you exclude the edentulous, and by the author’s own admission (see above) effectively pathological, skull 4 from Fig. 4 then the amount of skull shape variation in the Dmanisi sample is effectively halved. It is this latter variation that should be used as the yardstick for variation within a regional sample, not the former. This is all the more curious because I have recently reviewed another ms. from some of the same authors that stresses the importance of taking age-related changes in the mandible into account when considering the range of mandibular morphology attributable to a single species.

Second, I am concerned by the implication that 3D cranial shape is taken to be the arbiter of early hominin taxonomy. It is perfectly possible for two taxa to share whatever aspect of cranial shape that is being measured in the plots in Figs. 4 and S7B (but remember that whatever ‘that’ is fails to discriminate between a morphologically very distinctive large-brained Neanderthal cranium and Dmanisi skull 4) and still be separate species. Some of the features that have been used to distinguish Homo habilis sensu stricto, Homo rudolfensis, Homo ergaster and Homo erectus sensu stricto might be captured in an analysis that focuses on aspects of cranial shape, but many (e.g., detailed basicranial morphology, bony labyrinth morphology, foot morphology, long bone strength, life history, relative tooth size, etc, etc.) would not be captured in such an analysis. So what justification is there for claiming to refute a taxonomic hypothesis when the grounds for doing so are so limited? It seems to me just as plausible to argue that Dmanisi samples a hominin taxon that exhibits a hitherto unknown combination of primitive (e.g., small brain, sagittally-oriented petrous axes, etc.) and derived (e.g., brow ridges) cranial and mandibular morphology.

Third, reading S.7 and looking at Fig. S8 I am puzzled. With respect to cranial comparisons (the right hand column in Fig. S8) the authors are using a criterion based on a metric that with respect to SC1 and SC2 cannot discriminate between skull 4 and a Neanderthal, or between Sts 71 and KNM-WT 15K. With respect to the calvarial comparisons (the left hand column in Fig. S8) the message can equally well be read as being compatible with conventional wisdom (i.e., early Homo from Africa is more variable than Homo erectus).

Fourth, according to the logic followed by the authors, why are Sts 5 and Sts 71 excluded from the single taxon that subsumes Homo habilis sensu stricto, Homo rudolfensis, Homo ergaster within Homo erectus? Homo erectus africanus?

Fifth, according to the authors, specimens in a paleodeme must come from a “spatially and temporally constrained stratigraphic setting.” Even if this is true, the converse is not necessarily true. Would the authors consider KNM-ER 406 and KNM-ER 3733 to be from the same paleodeme?


Lastly, some specific comments on the text: 

p. 327. “The skull 5 individual thus provides the first evidence that early Homo comprised adult individuals with small brains but body mass, stature, and limb proportions reaching the lower range limit of modern human variation.” This is incorrect. The same mismatch between brain size and body mass is seen in KNM-WT 15000.

p. 327. The authors refer to “estimated molar dimensions”. How were these estimated? Were the raw or the estimated dimensions used to draw the conclusions about molar size order set out in Table S4A?

p. 329. Fig. 3. The hypotheses in Fig. 3B are unrealistic and fallacious. I do not know of any early hominin taxonomic hypotheses that claim there is no point of overlap between species. Even the derived dental morphology seen in species such a Paranthropus boisei is shared with other taxa. The point about the morphology is that whereas it is found in 100% of P. boisei it is seen in a much smaller percentage of other taxa.

p. 329. Fig. 4. In the caption the authors write “SC2 captures shape change associated with grade shifts in neurocranial size between taxa”. Can this be correct if KNM-ER 1470, KNM-ER 3733 and Neanderthals have the same SC2 values?

p. 330. “The hypothesis of multiple independent lineages (paleospecies) (15, 31) appears less parsimonious, especially in the absence of empirical evidence for adaptation to separate ecological niches.” The assessment that there is no “empirical evidence for adaptation to separate ecological niches” is difficult to understand. First, a substantial part of the case Wood and Collard (1999) made for Homo habilis sensu lato and Homo erectus being in different genera was based on variables that reflected adaptive differences (e.g., diet, locomotion, etc.). Second, earlier this year Cerling et al. (2013) showed that in the Turkana Basin there was a significant shift in stable isotope values between Homo habilis sensu lato and Homo erectus specimens. The authors of the submitted manuscript should either refute the data, or the inferences.

I have not exhausted the comments that could be made, but I hope my referee’s report gives a sense of what needs to be done to strengthen this ms. This is extraordinary fossil evidence whose scientific value has been augmented because of its careful contextual underpinning. It deserves to be subjected to appropriate painstaking analyses that are interpreted in an open-minded way. In my judgment this ms. has some way to go before the analysis and interpretation match the significance of the fossil evidence.

(PS To avoid any misunderstanding, let me make it clear that my views were not sought by the editors who processed either of the two manuscripts mentioned in this blog. This is the referee’s report I would have written about one of the manuscripts)


References

Cerling, T.E., et al. (2013) ‘Stable isotope-based diet reconstructions of Turkana Basin hominins’. Proc. Nat. Acad. Sci. 110 (26): 10501–10506.

Lorkipanidze, David et al. (2007) ‘Postcranial evidence from early Homo from Dmanisi, Georgia’. Nature, 449: 305-310.

Wood, Bernard and Collard, Mark (1999) ‘The Human Genus’. Science, 284: 65-71.