Frontiers of human origins

This month's Discover came in the mail today. In celebration of their 25th anniversary, their issue is devoted to "Frontiers of Science", with articles covering the (speculated) cutting edge in different fields. One spread is dedicated to "Human Origins", with a short piece by Carl Zimmer, an interview of Tim White, and a graphic.

Zimmer's article, "Digital ancestors walk again" covers the increasing use of CT imaging and reconstruction of hominid fossils. The subtext is that anything high tech must be better -- for example, the article labors under the misimpression that we cannot study endocranial contours without cutting a skull open.

Also, a read of the article gives the impression that every finding from this new advanced technology supports splitting hominids into several species (in particular, it mentions the Liang Bua endocast reconstruction and the virtual Neandertal growth series assembled by Ponce de Leon and colleagues). Probably this trend will continue as long as few people work with scans except the people who do them.

Here's the conclusion:

As the use of CT scans expands, paleoanthropologists are developing new avenues for uncovering clues to our past. They are discovering signs of healed wounds, of toothless old hominids who must have been cared for by others. Some researchers are even producing full-length virtual skeletons to which they can attach virtual muscles and make the ancient hominids walk again. Most significantly, CT scans can liberate hominid fossils from museum drawers. Once a research team makes a scan, they can post the data on a Web site for other researchers to analyze, bringing a precious hominid fossil to new sets of eyes and new sets of questions.

So utopian....

But Tim White has a different view of the kind of technology changing the field:

What technology advances are changing the way you study evolution?
W: The global positioning satellite system. WIth GPS, we no longer have to worry about the position of a fossil. Some of the biggest blunders in the history of paleoanthropology were made by people who lost the place where a fossil came from. There's no excuse for that anymore. The other big advance is in geochemical dating....

No CT scans there.

Personally, I think CT will have a limited set of impacts. The best thing is that it will allow any lab in the world to have as full a set of comparative data as have been released. Currently, it's useless for that purpose; there's just not enough access. But that is changing, and CT scans are as useful to a practiced eye as casts -- which are much less available today even as CT increases. In fact, high-resolution CT may essentially end casting of new fossils, since that is one of the major sources of damage. We'll be doing a lot of comparative work with imaging in the future.

On the other hand, I think CT will have a really limited impact on the study of new fossils. For one thing, those of us who are used to studying fossils are trained to deal with fragments. People do reconstruct fossils, but reconstruction is not essential to studying most morphology. Another thing we are trained to deal with is distortion. Especially plastic deformation can affect the very shape of the fossils we work with. We correct for it by examining which morphology is affected or unaffected, and by making conservative estimates. CT imaging is too tempting in this scenario -- it encourages people to think they have corrected problems, when instead it is merely adding geometric assumptions.

CT imaging and reconstruction is often proposed as a way to deal with distortion and fragmentation: Zimmer mentions the "sophisticated mathematical software to find the best way to assemble" reconstructions. But these can end up just as biased as any handmade reconstruction, even when the distortions are fairly apparent. There has been substantial disagreement about the CT reconstruction of Stw 505; we may find that the reconstruction of the Sahelanthropus skull faces similar problems.

There are three main benefits of CT reconstruction: it allows repeated trial and error assembly without actually grinding bone contacts against each other, it allows mirror-image substitution for missing parts, and it allows a skull to be geometrically fit to a model. The repeated trials are very valuable: they give an experienced anatomist a chance to try slightly different configurations to sense the range of variation resulting from the state of preservation. The mirror-image reconstruction is valuable for visualization, but potentially misleading for scientific comparison -- try taking a flat mirror and reflecting half your face: notice how hard it is to align the mirror properly and how odd it makes you look and you'll have an impression of the problem. The widespread use of geometric fitting is a potential disaster: by encouraging the use of a model, it reduces the range of biological variability expressed in fossils. When these mathematically-fitted reconstructions are then fed into mathematical comparisons, the structure of the data will be biased by the reconstruction technique in ways that may not be visually apparent.

Anyway, despite all the math, the computer is only as good as the scientist running it -- the principle of "garbage in, garbage out" is everlasting.