Is this the end for Homo floresiensis?

The paper by Teuku Jacob and colleagues is being published in PNAS today. Today's papers haven't appeared yet, but the press release is available online at Science Blog. Here's a quote:

To study LB1's traits, 94 cranial features and 46 features of its mandible were compared to values for modern humans. All fell within the normal range of variation for Australomelanesians. Two anatomical details, particular grooves in the cranial base singled out as "not seen in modern humans," in the 2004 new species announcement are, according to Alan Thorne, archaeology and natural history, Research School of Pacific and Asian Studies, Australian National University, Canberra, commonly found in Australian and Tasmanian crania.

I will post a link to the paper when it appears online. (UPDATE (8/23/2006): Now available online, open access.) In the meantime, I have a copy and I'll post some comments about what it means.

John Noble Wilford has an article with extensive quotes from both Bob Eckhardt and Peter Brown.

Asymmetry

To me, the most important point raised by the new paper is the asymmetry of the skeleton. Here's a picture:

Liang Bua 1, in normal (l), right mirrored (c) and left mirrored (r) views. Photo courtesy of D. W. Frayer.

The authors argue that this asymmetry is a sign of developmental abnormality. This is from the paper (Jacob et al. 2006:3-4, citations omitted):

After allowance for left orbital region damage, asymmetry affects frontal breadth, position and contour of the lower orbital order, angle of inferior nasal margins, location of the (broken, off-center) nasal spine, and expression of the canine juga. Asymmetry also extends to other areas on the cranium. Detectable in the original figure 1 in ref. 1 but not mentioned, and more clearly here, the palate midline is rotated 4-5 [degrees] from the midsagittal plane, so that a line extending the midpalatine suture does not bisect the foramen magnum. In addition, right parietal and left occipital bones are flattened compared with their more rounded antimeres. The nuchal torus is markedly asymmetrical, and the antimeric mastoid regions differ greatly. This pervasive asymmetry suggests growth anomalies producing a vault and face with substantial lateral contrasts.
Patterns of asymmetry in the human skull are well documented, with typically greater neurocranial asymmetries than in the face, where asymmetries are slight overall .... These differences represent small (typically <1%), fluctuating asymmetries. Reanalysis of these data (citation) supported the earlier statement that "[t]he face is the most symmetrical region of the skull." ... Studies of facial asymmetry are medically important and have diagnostic applications, with low single-digit lateral deviations in percentages or millimeters marking thresholds for clinical intervention.
We quantified craniofacial asymmetry for LB1 to the extent possible by dividing digital photographs of the cranium into right and left halves along the midsagittal plane by using Photoshop, then making composite images by mirroring the left and right sides. On a digital image of the face, we also measured left and right deviations from the midline. Six of seven measures were larger on the right, by amounts ranging up to nearly 40% (distnace from mental foramen to midline). The only measurement larger on the left was the 6% for distance from the orbit lateral rim to midline (see Supporting Text). Gauged by anthropometric and clinical standards, LB1 asymmetry exceeds clinnical norms where determinable, providing evidence for rejecting any contention that the LB1 cranium is developmentally normal.

I quoted this extensively because the details are important as a response to an obvious criticism. The criticism is raised by Peter Brown in the Times article:

Dr. Brown said the critics' claim of "the asymmetry of the skull being the result of abnormal growth is fiction." The skeleton was buried deep in sediment, he said, and this brought on "some slight distortion."

Yes, it is true that any archaeological specimen is likely to be distorted to some extent by reconstruction or postdepositional deformation. That might be true of this skull also. But in this case, the asymmetry clearly extends to morphological characters that should be relatively unaffected by such distortion. For example, the position of the mental foramen shouldn't change by 40 percent from being buried in wet ground. And while distortion might affect the curvature or flatness of the occiput, it shouldn't affect the relative development of the nuchal torus or mastoid regions on each side, which must be a consequence of different muscle development or configuration.

That muscular problem appears to be reflected in other areas of the skeleton also. For instance, the authors argue that the humeral torsion is a correlate of "extremely weak muscle development indicated by muscle insertions", and may reflect a "response to the dynamic forces exerted by shoulder rotators on the growing bone" (p. 5).

Likewise, the femora are asymmetrical and show odd muscle features:

Proximally, the intertrochanteric crests on the femora are highly asymmetrical (larger on right), with sizes and positions of lesser trochanters differing substantially between sides. On the right, the spiral line, adductor insertions, and lateral gluteal lines converge toward the midshaft, where they run parallel for only ˜30 mm before dividing again into borders of the popliteal surface. These lines are barely visible, even in the middle of the shaft, unlike a normally robust and prominent linea aspera. The arrangement of muscle attachments is similar on the shaft of the left femur, but lines are not visible distally. Such atypical features imply severe muscle hypotonia (paresis) during life, associated with complications of abnormal growth (ibid.:5).

The authors consider that it seems incongruous to find such apparently weak musculature for a skeleton that has been described as morphologically robust. But so far, the only evidence for "robustness" has been the large diameters of the long bones in comparison with their short length. Of course, this may result from change in the denominator (shorter bones) as well as in the numerator (thicker bones).

So they scanned the bones. They discovered that the long bones have thin cortical bone:

Inflated circumferences, combined with very thin cortical bone showing very weak muscle markings, indicate not robusticity but long bone overtubulation indicative of disordered growth (5).

To me, all of these features make a very compelling case for pathology. The asymmetry is clear and widespread throughout the skeleton (this is not to mention the evidence for asymmetry in the endocast, which remains to be reported). The remaining traces of muscle attachments and bone thickness all are either asymmetrical or unusually weak. The humeral torsion is extremely unusual, and is apparently not explained by functional comparisons with any hominoids. From top to bottom, this skeleton has obvious problems.

The authors here argue that the pathology is likely some form of microcephaly, but they do not offer a specific diagnosis. However, they do note that many documented conditions involving microcephaly also involve small stature.

Comparisons with local pygmies

The other main element of the paper is the comparison of the Liang Bua morphological characters with those found among local people, as a result of the Rampasasa expedition:

From the beginning, the "H. floresiensis discovery" was treated as a matter of hominid taxonomy and phylogeny. Curiously, however, comparisons of LB1 were made mostly with H. sapiens from other geographic areas of the world, principally Europe. Yet it would have been logical even for a supposedly novel human species from the Australomelanesian region to have been compared with other human populations, present as well as past, from that region (2).

So they did the comparisons:

Aside from abnormalities discussed below, not one of the 94 descriptive features of the LB1 cranium or the 46 features observed on both mandibles lie outside the range for modern humans from the region. The form of the superciliary area, nasal floor, subnasal region, orbits, and occipital superstructures of LB1 all are encountered routinely among Australomelanesians.... Purportedly, the LB1 cranium displays two skeletal features "not seen in modern humans." In one, "a deep fissure separates the mastoid process from the petrous crest of the tympanic" bone. The other is "a recess between the tympanic plate and the entoglenoid pyramid" on the medial part of the mandibular fossa. Australian and Tasmanian crania commonly display both these features. The latter trait is also present in two Pleistocene Australians, Kow Swamp 5, and, in a reduced form, Keilor.

Personally, I have to observe that "not seen in modern humans" is like a hanging death sentence for any paper that uses the phrase. Somebody is always going to check that in museum collections, and they are always going to find humans that have the "never seen in humans" feature. I think there must be some cryptosadistic editors that let this phrase stay in papers because they secretly enjoy seeing the authors get their comeuppance.

They find many other "distinctive" features of the skeleton in local or regional samples. A strong majority (>90%) of the Rampasasa people sampled had "neutral or negative chins." Many of them had rotated teeth. There is a high regional frequency of third and fourth premolars with Tomes' roots. And the authors argue that a wide pattern of not-so-unusual morphological features of the teeth may be evidence of a link to local peoples:

Other dental traits linking LB1 to modern pygmies from the Liang Bua region include a tendency for the longitudinal fissure to shift away from the buccolingual axis on lower molars, tremata (spaces between teeth), squared lower molar outlines related to hypoconulid loss, and large buccolingual P3 diameters. Overall, the dentition of LB1 exhibits modern human traits, with bilateral rotation of the upper fourth premolars and tooth shape deviations in lower premolars, both of which seem to occur at elevated frequencies in the Rampasasa (4-5).

This echoes the Gary Richards' conclusion that the fossils share features with modern humans that would have to be explained by parallelism under the species interpretation, but with much more detail.

I can imagine a response for this argument is that these features are "plesiomorphic", shared from some ancestral species of Homo. But you can only push something so far before it starts to be pulled by something else. In this case, if these features of the skeleton that are shared with modern Australomelanesians are really primitive features that were inherited from some distant ancestor, then you have to wonder whether any features really mark modern humans in contrast with that surprisingly derived primitive ancestor. And just where is that ancestor supposed to have lived, hiding its derived morphologies from our prying eyes? The plesiomorphy explanation always seems to me like a stretch, because it depends so strongly on unobserved facts, but in this case it stretches the imagination. No, if Liang Bua represented a different species from modern humans, I think these features must certainly be explained as parallelism, which demands some account of how they could have evolved in common.

In any event, the authors in this case present their conclusions in terms of a null hypothesis that the Liang Bua fossils are the same species as us, and they conclude that the null hypothesis is not refuted by the features of the specimens. They don't measure or attempt to estimate whether the apparent parallelisms would be impossible if Liang Bua were a different species, but they conclude that they would be unlikely.

Island biogeography

They present two additional reasons to think that in situ evolution of a dwarf hominid species on Flores was unlikely. First, they argue that Flores just wasn't all that isolated over the relevant time period. They note that only two short water gaps separated Flores from mainland Asia during glacials, and there were at least two dispersals of elephants to the island during that time. They argue that this is evidence against isolation.

Second, they note that even though Flores is a large island, it wouldn't support a very large population at typical human hunter-gatherer population densities. With a probable census population size on the order of a few thousand (they estimate between 500 and 5000), the population would be continually in danger of extinction during the past 800,000 years.

These two arguments are sort of a pincer-effect on the possibility of endemic dwarfing on the island. If the island was really isolated enough for the evolution of an endemic species, then it couldn't have supported a large enough population to ensure survival. On the other hand, if it wasn't really isolated, then people could more easily have survived (with gene flow from outside), but they wouldn't be endemic.

I am less convinced by this element of the paper. For one thing, elephants swim, and their ability to colonize the island twice doesn't necessarily mean that humans could have done so. Of course, since the first humans got there by boat 800,000 years ago, we have to wonder why they subsequently lost that ability (otherwise, no isolation). But it would be more convincing if there were clear evidence of other Asian species appearing, and there isn't. It is, after all, Wallacea.

For another thing, the density estimates from modern human hunter-gatherers don't necessarily apply to super-small-bodied pygmy hunter-gatherers. If the reason that selection favors dwarfism on islands is limited resources, and if small body size really adapts species to those limited resources, then small-bodied hominids should have had much smaller areal requirements than large modern humans.

I don't think these criticisms refute the paper's arguments about biogeography, just that they aren't the strongest aspect.

Conclusion

So, is this the end for Homo floresiensis? I'm sure it won't be. But there are some things that I hope will come to an end.

First of all, it is now abundantly clear that some kind of microcephaly can explain the small size and small brain size of the LB1 specimen. Moreover, the specimen exhibits other very obvious signs of developmental pathology. It is a bad specimen on which to base the diagnosis of a new species; its most important features are quite plausibly caused by its manifest pathology.

The argument so far against pathology has been that it cannot explain other unique morphologies, like the lack of a chin, and Tomes' root, and so forth. But this paper shows that none of these other features are necessarily unusual for modern humans, in the local and regional context. So that argument is dead, unless someone can show that there is some unique character to the combination of traits in the specimen. Since most of the features that would differentiate it from Homo erectus -- purportedly due to endemic dwarfism -- are also shared with modern humans, that seems like a problem for the species idea.

So I completely accept the argument that LB1 is pathological. A corollary is that the skeleton cannot be a convincing type specimen for a new species.

But this isn't only about LB1: there are the other small specimens. This paper makes clear that none of the features of the LB6/1 mandible are outside the range of local peoples. This is not a case of two specimens that must share some rare pathology; the paper argues that they are two specimens that share a regionally-common suite of characteristics. They aren't, in other words, unusual.

The paper is silent on the other postcranial remains, except for the very short tibia LB8, about which there is this comment:

[T]he LB8 tibia, with an estimated length of 216 mm, is used to reconstruct a stature of 1.09 m, greater than that of LB1 at 1.06 m, although LB1 tibia length is 235 mm.

That says to me that there is still work to be done in establishing exactly what the body size of the other individuals was. Maybe they will all turn out to be very small, but it is hard to extrapolate because they are outside the range in which modern human regressions will be accurate -- even on modern-day pygmies. In short, the paper does not claim that the Liang Bua fossils do not sample an exceedingly small-bodied population. That population must itself be explained in evolutionary terms. Since we don't yet have a particularly good explanation for small-bodied human populations anywhere, this is certainly a worthwhile topic to consider further.

Another argument, about the tiny-brained humans making the modern human stone tools, was to me a nonstarter from the beginning. The thing is that all early humans were under selection for smaller brains -- for energetic reasons -- and all early humans were under selection for more complex culture. If there were a way to combine small brains and complex culture, some other population of humans would have found it, and that population wouldn't be a small population trapped on an island. So the Liang Bua tools should be sufficient to demonstrate the presence of modern humans.

I would be focused like a laser on three things. First, what happened to the early colonists of Flores? Do any 200,000-year-old archaeological sites survive? What was the ecology of the earliest Flores people?

Second, when did modern humans really show up in southeast Asia and Java? One of the only supports left for the idea of a Flores species is that the earliest levels at Liang Bua (>70,000 years) supposedly predate modern human arrival. Were modern humans there at that time? Are those levels just inaccurately dated?

And third, what happened on Timor? It seems to me that the game is over if humans were on Timor in the Middle Pleistocene. There is some indication that they were. If Homo erectus could manage sea crossings to that extent further west than Flores, then there is no way that Flores was a single, unique colonization. With an occupied Timor, we have to assume that regular contacts between Flores and mainland Asia were probable.

References:

T. Jacob, E. Indriati, R. P. Soejono, K. Hsü, D. W. Frayer, R. B. Eckhardt, A. J. Kuperavage, A. Thorne, and M. Henneberg. 2006. Pygmoid Australomelanesian Homo sapiens skeletal remains from Liang Bua, Flores: Population affinities and pathological abnormalities. Proc Nat Acad Sci USA. PNAS published August 23, 2006, DOI link