open access

There are three skulls from putative "hominins" that date to 3.5 million years or earlier. Every one of these skulls is known now from extensive reconstruction or correction for distortion in the original.

By itself, the extensive reconstruction might not be a problem. But as Tim White has repeatedly shown, the specialists on these crania actively and vociferously disagree about the basic anatomy due to problems reconstructing them. White's ongoing dispute about the skull of KNM-WT 40000 is a matter of public record, both in his initial 2003 article on the skull, and in Michael Balter's description of the recent Royal Society meeting*:

When the talk was thrown open for discussion, White took the microphone and began firing questions at Spoor about the degree of variation of the cheekbone position among specimens of A. afarensis and other hominin species. “We took that into account,” Spoor responded, “and I just showed you a graph” about it. “I didn’t ask you whether you took it into account; I asked you what it was,” White said. Spoor, clearly frustrated, told the audience that he had no vested interest in this debate. At that point, the session chair interrupted and invited everyone to break for coffee, but Spoor and White continued to debate between themselves for the next half-hour.

If KNM-WT 40000 were the worst case, that would be bad enough. But Ardi's skull has required reconstruction even more extensive than would be required for the Kenyanthropus holotype.

In their description of the Ardipithecus skull, Suwa and colleagues (2009) mainly present metrics taken from the CT reconstruction. The publication strikes me as remarkable in that it includes few photographs of the original fossil, and only one or two of the photos are in standard anatomical orientation. A substantial part of the CT reconstruction is based on a second individual (ARA-VP 1/500), of which no photographs are provided. For this, readers may refer back to the single rather poor photo in the 1994 description. Anatomical comparisons in the present paper are limited to visualizations of the CT reconstruction.

As I've written elsewhere, I think that Suwa and colleagues did a remarkable piece of reconstruction. But it is non-replicable. The CT-reconstruction is a composite of two specimens that includes mirror-imaged parts. A tremendous amount of work went into it, but without access to the component parts, it isn't possible to test or verify the assumptions underlying the present model.

Another striking thing about the Ardipithecus skull description is the lack of anatomical comparisons with relevant samples. I mentioned above that most of the figures involve metric comparisons -- many of them scaled to the cube root of endocranial volume -- which of course can only be taken on a small fraction of early hominin crania. That leaves out the most relevant specimens in the Hadar sample, including all the cranial specimens from AL 333. It leaves out most of the Sterkfontein collection.

And it brings us back -- again! -- to Kenyanthropus. Reading back through the paper, it's hard for me to believe that reviewers allowed Suwa and colleagues to publish on Ardi's skull without including any comparisons with KNM-WT 40000. It's the earliest complete skull of an undoubted hominin.

They're entitled to their opinion that the skull is distorted. I agree. But you can still compare most of its nonmetric features and put some reasonable bounds on its metrics. I mean, they included OH 5, for goodness sake -- which has nothing whatever to do with hominin origins. Including the comparisons wouldn't have changed much about the paper, although I'll point out that there's at least one derived feature of later hominins that KNM-WT 40000 and ARA-VP 1/500 both lack, and which isn't noted either by Suwa and colleagues (2009) or in the table presented by White and colleagues (2009).

So what should we do? We can't see the scans, no independent reconstructions are possible, and the people who can see the scans refuse to present comparisons of these three skulls that together represent the supposed origin of the hominin lineage.

AAAARRRRGGHHHH!

We need to set up multiple sets of independent reconstructors having a replicable go at these skulls. These are the three earliest hominin skulls. Every one of them is crushed in some weird way. It would be a credit to the science to document their reconstruction in nauseating monograph-level detail. They're scans, for goodness' sake -- there's absolutely no argument that access should be limited for any reason.

If I were running this, I would set up a graduate seminar devoted to putting them together, split among four universities, with results to be reported in a session at the meetings and monographically by e-publication. The issue is not whether we can obtain an exact representation of the original anatomy. The issue is whether we can reject hypotheses about that anatomy. Testing hypotheses requires us to survey the range of possible reconstructions and how they relate to the range of anatomical variation in living and extinct analogs. The more reconstructions, the better the testability.

At the moment, that testability isn't there. I trust the anatomical expertise of the people who made the models, but they're just single models with no assessment of the range of error. I've written about the importance of open access for these reconstructions already ("Open access and fossil reconstruction"). The points here just amplify that theme.

* As an aside, I wonder if the title of the Royal Society meeting ("The First 4 Million Years of Human Evolution") contemplated the possibility that there may have been only 4 million years of human evolution in total?

I would love to be able to say that the Ardipithecus pelvic and cranial reconstructions were open access.

The reconstruction of fragmentary fossils has in the past been more of an art than a science. An anatomical expert can eliminate some possible morphological configurations based on the remains themselves. But for many, she has only her knowledge of variation in extant species as a reference. The bones she has studied might or might not be representative of anatomical variations; variations within extant taxa might or might not be relevant to ancient species. Working with casts of a fragile specimen is fraught with problems. Missing parts or uncertain joins in the fossil material can be shored up with plasticine, but to other scientists these This process poses obvious drawbacks: the resulting reconstruction may present the appearance of features that are in fact completely sculpted out of clay.

I describe this as an “art” for one important reason: there are many barriers to replicating a reconstruction. The reconstruction ends up including the quirks of other specimens used as reference material, may have fragments misplaced due to uncertain identifications.

The pelvic reconstruction of Sts 14 is one example of how the implicit assumptions of a reconstruction can affect the interpretation of a fossil. After its discovery in 1947, John Robinson reconstructed the distorted ox coxae and partial sacrum with a rounded pelvic inlet, more or less like humans. After the discovery of Lucy’s pelvis (AL 288-1), it was clear that A. afarensis had a very broad pelvis, flattened from front to back — different from Sts 14’s apparently rounded pelvic shape. The original reconstruction of Sts 14 was revisited in the 1990’s, when it was found that a flatter, more Lucy-like shape is consistent with the specimen’s anatomy (Abitbol, 1995). The point is not that Robinson’s reconstruction was wrong — any reconstruction will be wrong in some details. Nor is the point that the reconstruction was not replicable in principle – at any time, anybody could have sawn apart a few casts of Sts 14 into the component bones, and then built them back in a different shape. The point is that replicating the reconstruction would have been expensive and difficult, so for forty years nobody made the effort.

With digital scanning, all the expense and difficulty go into producing the initial scans. After that, the only limit on testing a model reconstruction is the time that someone is willing to spend studying the anatomy.

In principle, this is wonderful. A whole team of researchers can easily share digital models, working on the specimen with an explicitly shared referent. The digital model can be instantly transported anywhere in the world, allowing direct comparisons with original material housed in museums. The existence of such virtual 3-d images of fossil and model allows independent scholars to apply their own models, testing a model’s assumptions without needing to handle and possibly damage the original specimens.

But many of these benefits of the technology depend on scholars being able to access the scans. Today they can’t.

I’m hopeful that in the future we’ll be able to make full use of the technology — not only enabling a single reconstruction, but multiple reconstructions and widespread comparison of digital models.