Archaeologists have begun trading verbal blows over a set of animal bones said to exhibit the earliest evidence of stone tool-assisted butchery on record. The skirmish raises questions about how archaeological analyses are conducted.
This past August, Shannon MacPherron of the Max Planck Institute for Evolutionary Anthropology in Leipzig, Germany, and his colleagues reported in the journal Nature that they had recovered bones of cow- and goat-size animals dating to around 3.4 million years ago that bear cutmarks inflicted with stone tools. The bones come from a site in Ethiopia called Dikika, just a few kilometers from the site where the famous Lucy fossil was found. Previously, paleoanthropologists working at Dikika found a spectacularly complete skeleton of an infant belonging to Lucy’s species, Australopithecus afarensis. Because the animal bones are believed to be roughly the same age as the A. afarensis remains from the site, and because A. afarensis is the only human species known to have lived in the area at this period in time, the team concluded that A. afarensis used stone tools to butcher large mammals some 3.4 million years ago (the researchers did not recover any stone tools, however).
Up until this point the earliest evidence for stone tool use and butchery dated to 2.6 million years ago, and although the identity of the butchers in this instance is unknown, archaeologists assume they were more advanced than Lucy’s kind. So pushing the origin of butchery back nearly 800,000 years and attributing the handiwork to a species as primitive as Lucy’s was a big deal. Furthermore, A. afarensis had been thought to subsist mainly on plant foods.
Now a team including archaeologists who described the 2.6-million-year-old evidence for butchery—found at another Ethiopian site, called Gona—has challenged the Dikika evidence. In a paper published yesterday in the Proceedings of the National Academy of Sciences early edition, Manuel Domínguez-Rodrigo of the University of Madrid and his colleagues charge that "the claims for unexpectedly early butchering at the site are not warranted." They contend that the alleged butchery marks are better explained as marks inflicted by animals trampling on the bones.
Comparing the Dikika marks to marks left on experimentally trampled bones, the skeptics found a number of similarities. For example, some of the marks that the Dikika team described as having V-shaped cross section consistent with stone tool use the rival team describes as being _/-shaped, which is characteristic of trampling. Other features that the Dikika team interpreted as percussion marks indicative of marrow extraction also look to Domínguez-Rodrigo and his collaborators like trampling marks.
The authors of the new paper take issue with the Dikika team’s suggestion that the hominins (ancient humans) who made the cutmarks might have used naturally sharp stones to cut the meat off the bone rather than flaking the stones to create sharp-edged implements for the task. They note that no one has published the results of butchery experiments using natural rocks rather than intentionally sharpened ones.
Curious about the Dikika team’s reaction to these criticisms, I emailed archaeologist Curtis Marean of Arizona State University, the team member who analyzed the surface modifications on the animal bones. He called their attempt at discrediting his team’s interpretation a "rather weak effort,” noting that Domínguez-Rodrigo and his collaborators “only worked from photos, and that is completely flawed." He says he offered them access to the fossils, but that "they rushed this [paper] forward without taking the time to examine the fragments." Certain details that are indicative of butchery—such as the V-shaped cross section of some of the marks—are visible only with firsthand inspection of the bones, Marean says.
In addition, he notes, of all the experimental trample marks the critics examined, none even came close to matching two of the marks shown on the cover of the Nature issue in which the discovery team’s paper appeared. “In some ways that makes our case better than we did—the marks fall outside the range of variability for trampling damage," he observes. Marean also had this to say about the approach he and his colleagues took to analyzing the marks:
"We used [the] gold standard in mark identification--a 3-person blind test of correspondence. In other words, after I did my [identifications], I then gave the fragments to 2 other highly experienced analysts, with them having no prior knowledge, and had them identify the marks….The correspondence was extremely high. This is [standard operating procedure] in science for controlling for bias. Domínguez-Rodrigo ignored this procedure, which is a crucial flaw in scientific method."
Marean stands by the suggestion that the Dikika hominins may have butchered the animals using naturally sharp stones, rather than intentionally flaked ones. In fact, he and his colleagues have recently conducted a detailed study of butchery marks created by naturally broken stone, and he says the resulting marks look a lot like the ones on the Dikika bones. Their paper detailing the results is currently under review. “This is why [Domíinguez-Rodrigo et al.’s] paper is so premature,” Marean remarks. “Science takes time and effort, and rushing something like this to print without even looking at the material makes no sense.”
For more details (and barbed retorts), check out the prepared statement the Dikika team released to the press in response to the new paper:
"Domínguez-Rodrigo et al. have rushed into print a paper challenging our interpretation of the surface modifications on the DIK-55 bones without taking the time to look at the original specimens. Their arguments are based on a highly selective reading of our published photos and lack the rigor of multiple person blind testing. Their criticisms of the geological context are unwarranted, and their stated underlying assumption that stone tool use by A. afarensis is surprising is a reflection of their own biases and not based on theory or data. Thus, while we welcome debate, we find their paper unconvincing. Some specific responses follow:
First, we disagree with the line of reasoning that runs throughout the paper that our reported findings are surprising or unlikely and therefore suspect. Our find is the first to associate stone tool use and meat consumption with A. afarensis, but this point is irrelevant when it comes to assessing the marks themselves. Thus when they state that "In a less contentious context, the marks would likely be accepted as genuine cutmarks," the simple point of the matter is that it is only contentious if you think A. afarensis could not have eaten meat or used stone tools. There is, however, no a priori reason to think this was the case and they present no data to conclude otherwise.
Similarly, we question the paradigm that says that manufactured stone tools must be found prior to the first evidence of stone tool modified bones. It is possible, perhaps even likely, that the initial stone tools were not manufactured (ie. natural stones were used) or that they were infrequently made and distributed in very low densities across the landscape. If this was the case, it will be very hard to recognize archaeologically and in this regard stone tool modified bones might be a better line of evidence. At a minimum, the order of discovery (stone tools and then stone tool modified bones) or the immediate association of stone tools with stone tool modified bones are not valid criteria for assessing our finds.
Second, we disagree that because the bones were found on the surface their provenience is unknown. We acknowledged a very specific degree of uncertainty in the stratigraphic provenience, but this is not the same as "unknown". Conservatively, we know that the fossils come from the sediments exposed somewhere on the DIK-55 slope (from the location of the lowermost surface find to the top of the crest in our SI photo). That slope is a total ~16m of strata of the total 76m exposed in Andedo, and ~100m exposed in the Andedo/Simbil Dere area. So we can narrow the provenience to 16 of 76 meters or approximately 21% (ie. we know the provenience is within the lower ~20% of the interval between 3.42 and 3.24 Ma).
Third, however, Domínguez-Rodrigo et al. are more specifically concerned with the sedimentary matrix and our mention of the fossils as lack adhering matrix. They note that our fossils do have some sediment adhering to them and suggest that this is normal for fossils from the Sidi Hakoma member at Dikika. We agree, however, they have conflated two separate arguments we were making (one about the likely provenience of the fossils and one about the state of their preservation) and two separate localities. Previously reported fossils from Dikika, such as the DIK-1 nearly complete skeleton which is encased in a block of sands that have taken over ten years to remove, come from a 4m plus layer of cemented sands. By contrast, the DIK-55 fossils come from a thin (10-30cm - not thick as they state) layer of loose sands. As a result, the DIK-55 fossils are found on the surface in an excellent state of preservation, not requiring extensive cleaning, and with very little matrix still attached.
Fourth, they further question the provenience of the finds arguing that we should have excavated to find fossils in situ. We too would prefer to find fossils in situ, but this is not what happened in this case and in fact most of the stone tool modified bones prior to 2.0 ma come from surface collections (including those from Gona). Excavation could show that bones can be found in the sand layer we attribute our stone tool modified bones to, but it is logically impossible to demonstrate through excavation that our surface finds must have come from this sand layer.
Fifth, on the marks themselves, Domínguez-Rodrigo et al., having failed to examine the original specimens, compare our photographs to photographs of their experimental trampled sample and experiments with conventionally flaked stone tools. Out of a very large sample of experimentally trampled specimens, they succeed at finding a small sub-set of the trampled sample that superficially resembles a small sub-set of the Dikika marks. This of course means that they failed to find any trampling damage that resembled the many other stone tool inflicted marks on Dikika. Thus, the Dikika sample, as we originally argued, falls outside the range of variation of trampling damage. As they themselves note in their paper, their massive sample of trampling damage could not produce a set of marks that overlapped in morphology with some of the key Dikika marks - particularly DIK-55-2 marks A1 and A2. This is our point exactly.
Sixth, their approach, which they call the "configurational approach", fails to meet a basic premise of the scientific method - independent tests of diagnosis. They did not do a blind test of inter-analyst correspondence. In our study, three highly trained analysts with wide backgrounds in taphonomy independently identified marks of the original specimens as stone tool inflicted. In this PNAS paper, three analysts who regularly work together (two of which are holding their prior analysis up as the earliest stone took marks, and thus have a clear bias), reach a conclusion as a group - this is a well-known failed and biased approach. Without the blind test of correspondence, their conclusions become "group-think" applied to picture-matching.
Seventh, the authors fail to consider the item that could be causing the marks - unflaked stone - in their assessment. Rather, they base their assessment on standard flaked stone tools, which are not present at Dikika. As we suggested, the most likely tool was unflaked stone, which has broader less sharp edges than flaked stone.
Recent experiments reported in a paper now under review show that the Dikika marks are a tight fit to marks produced by unflaked stone. By rushing a response paper to press, the authors failed to both 1) examine the specimens first-hand, and 2) design proper comparative data-sets. That is unfortunate, and just muddies the water."
Image: Copyright Dikika Research Project