Shopping for rocks in the Olduvai Basin

The invention and proliferation of stone tool technology was one of the most significant events in human evolution--the ability to use stones as tools and, eventually, the wherewithal to modify them into sharp-edged knives and other implements enabled our early ancestors to access foods that would have been difficult or impossible to obtain and consume with their relatively small, unspecialized teeth. If you spend some time working with stones, it eventually becomes apparent that not all of them are created equal: some break easily, others are tough to fracture; some produce razor-sharp edges, others generate dull ones; some are close by and/or easy to get a hold of, others are far away and/or difficult to access; some are durable and last a long time, others are brittle and must be discarded after a single use.

Now, we know that a modern human can learn to recognize these attributes and, what is more, they can (not to say that they necessarily do) plan their days with them in mind ("well, let's see...there are two ways to get to the pond for fresh water, Path #1 and Path #2, but only Path #1 has an outcrop of durable rocks on the way, so I'll kill two birds with one stone and take Path #1"). The question, then, is this: to what degree did our early human ancestors appreciate the sometimes subtle differences among rocks, and what can this tell us about their cognitive capacities?

Before we can even answer this very interesting question, however, we need to figure out a way to (1) rank rocks in terms of their usefulness, and (2) determine where on the landscape early humans were getting their rocks in the first place. There is a long history of research on these topics in Paleolithic archaeology, and my colleagues and I added some data to the debate in a recently published paper in Quaternary International. My interest in the topic goes back to the late 2000s, when David Braun wrote a couple of really interesting papers on the stone tools from Kanjera South, a two-million-year-old site in Kenya. Most studies on rock "usefulness" are based on rather subjective and imprecise categories. These categories, and the studies that utilize them, have provided key insights, including the fact that rock selection by early humans was not random. Braun, however, explored the possibility that the material sciences might provide some useful tools to help archaeologists objectively describe the characteristics of rocks.

As I mentioned above, there are a host of features that one might consider when selecting a rock. We chose to concentrate on fracture predictability, largely because the creation of many types of stone tools involves breaking a rock into smaller (and hopefully useful) pieces. If a rock breaks differently every time you hit it, there is no way to predict what you're going to end up with. Sure, it might be useful, but, then again, it might not. With a rock that fractures predictably, though, you can be reasonably sure that the time and energy you've expended will pay off with the production of a useful tool. Flint knappers have known for a long time that homogenous rocks break more predictably than do heterogeneous rocks because they are stronger (they can resist strain) and more elastic (they can resist deformation) when impacted by an outside force. Thankfully, a rock's strength and elasticity are highly correlated with its hardness, something that can be quickly assessed with a rebound hammer. These nifty handheld devices, which were originally designed for use on concrete, fire a spring-loaded plunger onto the surface of a stone. The plunger then bounces back, or "rebounds," after impact. The distance of that rebound reflects the hardness of the stone. Braun and others have used this technique to estimate fracture predictability for the rocks available to early humans at several Pleistocene archaeological sites in Kenya.   

In 2014, we set out to produce comparable data in our neck of the woods, the Olduvai Basin of northern Tanzania. What is today a deep gorge surrounded by open grasslands was, about two million years ago, a stream-fed soda lake surrounded by lush vegetation. Largely unchanged, however, are the volcanic highlands that border the basin to the south and east and the numerous hills--remnants of Archean-aged metamorphosed bedrock--that rise above the plains. Importantly, both the volcanos and the hills are made up of rocks from which stone tools can (and, in the past, could) be made.

A view of Olduvai Gorge in the foreground and, in the background, Naibor Soit, a granulite outcrop from which quartz could be procured (photo: Amy Schnell).

With the help of students from the UNCG Olduvai Gorge Paleoanthropology Field School and Earlham College's Summer Collaborative Research Program, I and my good friends and colleagues Cynthia Fadem and Ryan Byerly have been traipsing around the Olduvai Basin hammering as many rocks as we can get a hold of. Since 2014, we've accumulated a database of 110 specimens, and some interesting patterns have emerged. It turns out that the volcanic rocks that occur as rounded cobbles within the seasonal streams that drain the volcanic highlands have high rebound values and, thus, high fracture predictability, while the metamorphic rocks from the hills show either intermediate or low rebound values. Now, if early humans were selecting their rocks based on fracture predictability, we might expect that most of the artifacts from the archaeological sites would be made from volcanic sources. It turns out, however, that among Olduvai's artifact assemblages, volcanic rocks tend to be very rare, while metamorphic rocks, especially those made largely of quartz, are very common, which implies that fracture predictability was not a major concern. But why not? We might interpret this pattern to mean that early humans in the Olduvai Basin were not clever enough to recognize the value of predictably fractured volcanic rocks. We're skeptical of this hypothesis, though, because experimental work indicates that there are good reasons not to select volcanic rocks, since they:

• usually occur as rounded cobbles, which are tough to flake because they don't have very many of the acute angles that make flake removal possible;
• require more raw muscle power to flake; and
• may not be as durable as other rock types.

What is more, the quartz-rich rocks in the Olduvai Basin:

• are readily available from conspicuous landscape features that are very close to most of the archaeological sites; and 
• are very friable, which, although reducing their fracture predictability, makes them relatively easy to smash into lots of small chunks, among which are typically a handful of useable tools.  

Finally, it's not like volcanic rocks were not utilized at all. In fact, early humans appear in some cases to have selected them over metamorphic rocks when they wanted to create handaxes rather than simple flakes. This makes sense given that the more complex production sequence of a handaxe probably requires a more predictably fractured rock. Unfortunately, you can't subject fragile artifacts to the impact of a rebound hammer. However, if you can tell where the artifact originally came from, we can correlate the hardness of our geological specimens with their archaeological counterparts without subjecting the latter to any damage. Well, in addition to hammering rocks, we also subjected them to X-ray fluorescence, which can help identify their chemical composition. The volcanic rocks are easily distinguished from the metamorphic rocks just by looking at them, but the metamorphic rocks themselves, even those from different hills, can look very similar to each other. Luckily, the chemical signatures of each metamorphic hill are distinct enough for statistical algorithms to correctly match chunks of rock to the correct hill with 75-80% accuracy.

In the future, we should be able to match Olduvai's metamorphic artifacts to the hills from which they were being collected, which in turn will give us an idea of how far early humans travelled when shopping for their rocks.

References:

Egeland, CP, Fadem, CM, Byerly, RM, Henderson, C, Fitzgerald, C, Mabulla, AZP, Baquedano, E, Gidna, A (2019). Geochemical and physical characterization of lithic raw materials in the Olduvai Basin, Tanzania. Quaternary International. doi.org/10.1016/j.quaint.2019.09.036

Society for American Archaeology meetings 2017

This year's Society for American Archaeology meetings were held in Vancouver, British Columbia from March 29th to April 2nd. I flew into town on Thursday night (the 30th) and had to run out on Saturday night. It was a whirlwind trip for me, as I presented a paper, co-chaired a special session, and supervised two undergraduate researchers who also presented at the meetings. As always, there was too much going on to see everything that was of interest, but I'll present my highlights:

• Way back in June 2016 I was invited by Karen Ruebens to contribute a paper to a symposium entitled "Connecting Middle Paleolithic Datasets: the Interplay of Zooarchaeological and Lithic Data for Unraveling Neanderthal Behavior," which was co-organized by Geoff Smith, Theresa Steele, and Tamara Dogandzic. Around that time I was reading an interesting paper by Marie-Cécile Soulier and Eugène Morin about the use of cutmark orientation to reconstruct planning depth in the Paleolithic, and this sparked my interest because, in 2014, we published a paper suggesting that cutmark orientation might be used to identify the skill level of prehistoric butchers. So, I thought the symposium would be a great venue to explore this issue further and I happily agreed to participate. The paper I presented with my colleagues Chris Nicholson, Kevin Covell, Robert Sanderford, Kristen Welch, and Metin Eren "Cutmarks, Lithics, and the Identification of Skill and Tool Utility in Experimental and Middle Paleolithic Contexts," turned out to be slightly different than what I had originally submitted in our abstract (a not unusual situation given that abstracts are due a full seven months, yes SEVEN MONTHS, before the meetings), but our major findings were (1) that inexperienced butchers tend to create more cutmarks and more variation in cutmark orientation than do more experienced butchers and (2) that Levallois flakes (which were produced by Metin, who is a world-class knapper and lithic technologist) are ideally suited to butchery because of their versatility. 

The preparation of cores for the removal of Levallois flakes.
From Eren and Lycett (2012: Figure 1).

While both conclusions are interesting, the latter was of particular note in the context of the symposium, since the creation of Levallois flakes is a labor intensive, not to mention difficult to learn, process (see figure above). Our results suggest that Levallois flakes' utility and durability as butchery implements may partially offset the energy expended to create them. In other words, it might make sense to spend an hour or two preparing cores (and wasting a lot of raw material in the process) if it results in the production of three or four Levallois flakes that can butcher a couple of carcasses without being replaced. I slipped out briefly after our talk to visit the poster session down the hall since my student, Liat Lebovich, was presenting on some landscape taphonomy we've conducted on the savanna just north of Olduvai Gorge, but returned for the symposium's first question break and the remainder of the session. Several questions were thrown my way, relating mainly to the application of our cutmark data to archaeofaunal assemblages (we have not yet taken that step). One interesting query, I believe from Eugène Morin, involved cutmark frequencies: we found that novice butchers produce more cutmarks within each "cluster" (defined by us as a group of cutmarks separated from other groups of cutmarks by at least one centimeter) than do experienced butchers. He asked if I thought that novice butchers would create more, or fewer, cutmark clusters. After thinking about it for second, I responded that while novice butchers create more cutmarks per cluster, they probably produce fewer clusters since the marks they create tend to be scattered across relatively expansive portions of the bone surface rather than the few discrete clusters (that make anatomical sense) created by more experienced butchers. I hadn't thought to examine this variable, and I certainly will do so as we move forward with our study. The rest of the papers were interesting, and the integration of faunal and lithic data is obviously critical to understanding Neandertal lifeways. Very honored to have been invited to participate in the symposium.


• On Saturday morning, I co-chaired a session with my friends and colleagues Tori Johnson and Ryan Byerly entitled "Ecological Perspectives on Hominin Landuse Use during the Early Stone Age of Africa." Included in the session were papers presented by Julio Mercader, Cynthia Fadem, David Braun, Briana Pobiner, and another of my students, Cory Henderson. The session was well-attended (it wasn't a massive room, but most of the seats we did have, perhaps 75 or so, were full). As I stated in my short introductory remarks, the impetus for the symposium stemmed from Rick Potts's "Variables Versus Models of Early Pleistocene Hominid Land Use" manuscript published in Journal of Human Evolution in 1994. In that paper, Potts argued (correctly, I believe) that paleoanthropologists should resist the urge to force early Pleistocene archaeological sites into specific "types," or, using Potts's terminology, models (e.g., Home Base vs. Refuge vs. Central Place). A more productive approach involves the identification of the ecological variables (e.g., the distribution of raw material outcrops, plant and animal foods, competitors, etc.) that influenced where and when early hominins accumulated stones and bones on the landscape. I also made it clear that our purpose in convening the symposium was not to chastise paleoanthropologists for not taking this approach--it is, in fact, relatively common among researchers investigating Early Stone Age archaeology--but, rather, to highlight the latest work being done from this perspective. UNCG undergrad (and recently accepted graduate student to the biology program at Penn State) Cory Henderson summarized the analyses we've been conducting on raw material outcrops in the Olduvai Basin over the past couple of years. Quartzite dominants a majority of the Bed I and Bed II sites at Olduvai, and it has always been assumed that the source of this raw material is Naibor Soit, a set of three closely spaced inselbergs just north of the gorge.

View looking south at one of the inselbergs (foreground) that makes up the
Naibor Soit quartzite outcrop. Photo by C.P. Egeland.

This is not an unreasonable assumption: Naibor Soit is very close to most of the archaeological sites and would certainly have been a conspicuous feature on the landscape, just as it is today. However, Naibor Soit may not have been continuously accessible, especially when a fully transgressed lake Olduvai likely transformed the inselbergs into islands. There are at least two additional sources of quartzite in the basin, Shifting Sand and Naisiusiu Hill, and we wanted to know if hominins were traveling to other parts of the basin to access this type of toolstone. This is not a straightforward exercise, though, because the quartzite from all the sources is macroscopically very similar. We decided to use X-ray fluorescence to determine if the outcrops could be distinguished based on their elemental composition. It turns out that there is considerable overlap, but samples of known origin can be correctly classified with between 36% and 76% accuracy, which is better than random chance (which would be 33%). When we threw some archaeological artifacts from Bed I into the analysis, most (64%) were inferred to have derived from Naibor Soit. However, a couple were a better match for one of the other quartzite outcrops. So, while hominins appear to have used Naibor Soit a majority of the time, they also visited other parts of the basin to acquire quartzite (from Shifting Sand, Naisiusiu or, perhaps, other sources that are currently buried or otherwise unobservable). The ability to tie artifacts to individual outcrops could potentially reveal very nuanced patterns of movement, which is required to rigorously test various models of hominin land-use at Olduvai. The most interesting paper, at least for me, was presented by Dave Braun on behalf of his co-authors Jonathan Reeves and Matthew Douglass and entitled "Behavioral Inferences from Early Stone Age Sites: A View from the Koobi Fora Formation." Every so often archaeologists, particularly those working in the Paleolithic, need to be reminded of the importance of scale: that is, the scale at which we would like to reconstruct behavior and ecological affordances (months, years or, ideally, days) and the scale at which our proxies of these variables actually permit (months, years, decades, and millennia--rarely, if ever, days). Dave provided exactly that reminder within the context of their on-going work at Koobi Fora, Kenya. They point out that archaeologists continue (wishfully, perhaps?) to apply behavioral models suited for temporal scales that are in most cases simply not reflected in the scatters of bones and stones that litter early Pleistocene paleolandscapes. Cynthia presented the latest geoarchaeological data from our ongoing excavations at DK East (Bed I) and BK East (Bed II). Perhaps most important is her identification of several unnamed tuffs within the DK East sequence. Dating these sediments should allow us to increase greatly the chronological precision in the area and, thus, derive narrower age ranges for the numerous fossiliferous deposits and more accurate estimates of sedimentation rates.    

• The session I most enjoyed, and learned the most from, was chaired by Erick Robinson, Robert Kelly, and Nicolas Naudinot and was organized around "The Future of 'Big Data' in Archaeology." An archaeologist's idea of big data (thousands to hundreds of thousands of records) differs substantially from what is typical in computer science (millions and millions of records), but increased computing power and more sophisticated statistical techniques now permit the aggregation and analysis of huge archaeological databases. I recently received a Digital Partners Grant from the UNCG libraries to collaborate on a database for Paleolithic sites, so I was particularly interested to hear about the latest from folks that have been thinking deeply about data and databases. There are a number of open-access databases, some of which I am familiar with and others with which which I am not, including:
• Canadian Archaeological Radiocarbon Database (CARD): a global compilation of 14C dates.
• Ancient Cacao Map: an interactive database for samples of ancient cacao in the Americas.
• ARIADNE: this is a database of archaeological databases spanning many periods and regions.
• The Paleoindian Database of the Americas (PIDBA): a repository of artifact distribution maps and images. 
• The Digital Index of North American Archaeology (DINAA): this site creates interoperability among archaeological site databases in North America. A good friend with whom I attended graduate school, Josh Wells, is one of the senior investigators on this project.  


Some of the major lessons that I drew from this symposium are:
• The longevity of databases is a real issue, since the people who run them retire and the funding that supports them expires.
• There was a lot of discussion about 14C databases. Many people are using them to model demography, for example (the basic idea: more 14C dates for a particular time period = higher population sizes), but the quality of dates remains very difficult to assess. This same issue of data quality applies to many large databases.

• I recently developed an interest in the use of social network analysis in archaeology. In fact, I've begun a collaboration with a colleague of mine at UNCG, Jing Deng, who is a computer scientist that specializes in online networks. We've joined forces to use his technical know-how and my background in archaeology to apply network analysis to the Paleolithic. Our Undergraduate Research, Scholarship and Creativity Office here at UNCG just funded a proposal we put together to build a social network database for the Magdalenian of western Europe with the aid of Amanda Chase (an anthropology major) and Nathaniel Arnold (a computer science major). The Magdalenian is an ideal test case since there are hundreds of well-dated sites with thousands of artifacts suited to the construction of social networks: portable art. The idea is that stylistic similarities reflect some sort of shared social identity, so if we can identify sites with high frequencies of shared artifacts with shared stylistic variants, we may be able to reconstruct social networks. Jing is adept at using social network software to identify important nodes and relationships between nodes. I do not know much about Paleolithic art and how it is categorized, so I sat in on a "Not Just Good to See: Global Perspectives on Scenes in Rock Art" session. While the focus was on rock, and not portable, art, I gained some perspective on how we might organize art into an analysis. Elizabeth Culley, who is a PhD candidate at Arizona State University, outlined an ontology of parietal art from France between 40,000 and 11,000 years ago. I listened closely as she discussed her use of the number of distinct scenes, the number of narrative scenes, the theme of the scenes (e.g., fighting animals), and the number of animals as important features that distinguish various sites. The other paper I was able to listen to, from Melanie Chang of Portland State University and April Nowell of the University of Victoria, discussed the (mis)use of gender binaries in interpretations of Paleolithic art. They show convincingly that many images that are traditionally identified as either male or female are, in fact, quite ambiguous when it comes to biological sex. Much of this "forcing" of images into male or female reflects modern western gender biases. This was pertinent to me, too, because I need to be careful about past interpretations of some of these images, since biological sex would be an important trait to use in the classification of images for social network analysis. The last presentation that I visited related to social network analysis was a poster by William Ridge (University of Illinois at Chicago) on social networks in the Early Bronze Age of the Aegean. What I thought was interesting was his use of geographic distance between sites ("nodes" in social network terminology) as the null hypothesis for further analysis. In other words, all else being equal, one would expect sites that are closer to each other in an absolute geographic sense to have stronger social ties than sites further apart. If the network analysis does not show strong ties between closely spaced sites, then other hypotheses (trade routes, etc.) can be explored. I will certainly use this approach with our own work in the Magdalenian.  

Well, that's it from my notes, great meeting!