IT WAS prehistoric rock art that first drew archaeologists to the remote and arid caatinga (鈥渢horn forest鈥�) of northeastern Brazil. The paintings are rich, colourful, highly expressive and, in places, downright raunchy. But it was not the painted walls that lured dozens of archaeologists from the Americas and Europe, myself among them, to the caatinga in December 1993. Rather, it was the opportunity to examine some tantalising clues of even more ancient human occupation in one of the painted shallow caves or rock shelters. Stone tools, stone hearths and charcoal found at one of the shelters, Pedra Furada (鈥渉ole in the wall鈥�), suggest that humans were living there nearly 50 000 years ago 鈥� long before people were first thought to have set foot in the Americas.
This has made Pedra Furada controversial. Some archaeologists hail the site as offering the strongest evidence yet for a human presence in the Americas before the end of the Pleistocene. Others, however, are more cautious about accepting the antiquity of Pedra Furada. Within the archaeological community, the generally accepted theory 鈥� that has staunchly held its ground for nearly six decades 鈥� is that the first Americans were the Clovis people who arrived in the New World little more than 11 000 years ago.
The foundations for the 鈥淐lovis-first鈥� theory were laid in the early 1930s, when archaeologist Edgar Howard of the University of Pennsylvania discovered distinctive stone spear points associated with mammoth bones at a site near Clovis in New Mexico. Subsequent radiocarbon dating put an age of 11 200 years on these artefacts. Since then, Clovis spear points and other projectiles of comparable antiquity have been found across North America and as far south as Tierra del Fuego in South America. There鈥檚 no mistaking that people were living in America by 11 000 years ago, but were they here earlier?
Advertisement
Over the past 60 years, and increasingly in the past 30, there have been scores of claims for pre-Clovis sites in the Americas. But virtually all have failed. In some, the apparent pre-Clovis artefacts proved to have been made by natural forces. In others, the dating techniques used were experimental and turned out to be unreliable. In still others, deposits were so badly mixed that alongside the alleged pre-Clovis artefacts were far younger artefacts, making it impossible to be certain just how old anything at the site was. Pre-Clovis claims spoil almost as quickly as fish and house guests.
After decades of fruitless searching for acceptable evidence of a pre-Clovis presence in the Americas, most archaeologists have grown sceptical 鈥� even cynical 鈥� about the prospects of ever finding such a site. Occasionally, however, pre-Clovis contenders appear that are not so easily dismissed. The Meadowcroft rock shelter in Pennsylvania is one. Excavated by James Adovasio in the 1970s, the site has yielded a mass of artefacts, conservatively estimated to be 14 250 years old. Another contender is Tom Dillehay鈥檚 Monte Verde site in southern Chile, where a boggy setting has preserved artefacts and food remains from 13 000 years ago. And Pedra Furada, a site in Brazil鈥檚 northeastern state of Piau矛 excavated by Ni猫de Guidon, is a third.
These sites have not been spared the sharp criticisms that greet all pre-Clovis claims, but so far none has been refuted. But then, none has been accepted either, largely because these are complicated sites and their status is not easily resolved. Questions arise, not easily answered at a distance, about the deposits in which the artefacts were found, about possible contaminants that might inflate the site鈥檚 age and about the origin of the artefacts. No one ever promised breaking the Clovis barrier was going to be easy.
Art of attraction
Take Pedra Furada, one of the biggest and most profusely decorated of some 200 painted rock shelters found at the foot of a sandstone escarpment. Initially it was the spectacular art that attracted the attention of Guidon, who is based at the Funda莽茫o Museu do Homem Americano (FUMDHAM) in S茫o Raimundo Nonato. In an attempt to date the paintings, she started excavating in 1978. The work continued for over a decade, down through nearly 5 metres of deposits in which she identified two distinct periods of prehistoric human occupation. One, named Pedra Furada after the site, was radiocarbon dated to the late Pleistocene and lasted from 48 000 to 14 300 years ago. The other, the Serra Talhada phase, started 10 400 years ago and extended to about 6000 years ago (see Diagram).
The later, Serra Talhada phase is characterised by distinctive, well-made stone knives and tools fashioned, in many cases, from chert or flint imported from outcrops considerable distances away. Extensive and complex rock art panels were painted on the shelter walls during this time. On the shelter floor are well-defined hearths. The surfaces upon which these people lived are not only easy to pick out, but are also littered with artefacts, animal bones and plant remains. In contrast, the older Pedra Furada phase lacks bone, wood or other organic remains, save for bits of charcoal. Nor does it have the rock art, well-defined hearths or obvious living floors of the Serra Talhada phase. Instead, its proof of a human presence rests largely on what appear to be 595 choppers, scrapers and cutting tools crudely fashioned from cobbles of quartz and quartzite. The source of the cobbles is just 100 metres away 鈥� on the cliff top that looms directly above the site. This has led to the suspicion among sceptics that the 鈥渢ools鈥� and 鈥渉earths鈥� of this older phase are no more than fallen rocks flaked and arranged by nature (see Diagram).
Guidon has dismissed these doubts, and for seemingly good reasons. In particular, she points to work by one of her co-excavators, Fabio Parenti. In a painstaking analysis of the Pleistocene specimens, Parenti showed that they possessed characteristics that are quite unlike those produced by the natural flaking of rock.
Even so, the long dispute over whether people lived in the Americas in pre-Clovis times is littered with sites that testify to nature鈥檚 uncanny ability to flake and fracture stone (bone too, for that matter) in ways that mimic primitive human artefacts. These mimics are known as geofacts, and they don鈥檛 just trap the unwary: in the 1960s, Louis Leakey was snared at the now infamous Calico site in California, a hillside of water-laid boulders and chert blocks thought to contain artefacts dating back to between 50 000 and 80 000 years ago. It didn鈥檛.
Convincing a wary archaeological audience that a site鈥檚 鈥減re Clovis鈥� claim is genuine is no easy task. Archaeologists cannot independently reproduce a crucial site in a laboratory, nor fully recreate it in words and pictures. An on-site visit to examine the material first-hand is essential to cut through the misunderstandings and debates that surround a knotty claim.
So, when Guidon and the FUMDHAM generously extended invitations to proponents and critics alike to visit Pedra Furada, more than two dozen of us accepted. It was a landmark meeting 鈥� stimulating, intense, enlightening and sometimes downright uncomfortable. The venue was S茫o Raimundo Nonato, a dot on the map in Piau矛 and gateway to the Serra da Capivara National Park, a beautiful ecological and archaeological preserve of which Pedra Furada is the centrepiece.
Sheltered housing.
The shelter floor, much of which is excavated, covers a roughly semicircular area 70 metres wide and nearly 18 metres deep, in the shadow of the overhang of a massive, south-facing, sandstone cliff 100 metres high. The upper part of the cliff consists of a layer of densely packed pebbles and cobbles of quartz and quartzite several tens of metres thick.
At both ends of the rock shelter are water-stained chutes down which pebbles and cobbles, dislodged by weathering and erosion, plummet to the shelter floor, sometimes swept along by what Guidon鈥檚 describes as 鈥渧iolent torrents of water鈥�. Violent, indeed. At the base of the eastern chute there is a large pile of freshly shattered pebbles and cobbles, flaked by the impact of their 100 metre free-fall. Beneath the western chute are several potholes, one more than 3 metres in diameter, scoured out of the bedrock by the erosive pounding of cascading water and debris.
The deposits that filled the shelter to a depth of 5 metres (preserved in two 鈥渨itness sections鈥� left unexcavated) are mainly coarse sediment, pebbles and cobbles (hand to head sized), and are mostly from overhead sources: the cobble layer, the uplands surface beyond the cliffs, and the sandstone roof and walls of the shelter. Not surprisingly, given their source, the pebbles and cobbles are flaked to varying degrees.
Clearly, nature has been active over tens of thousands of years in the formation of the deposits at Pedra Furada. But were there people at the site in pre-Clovis times? The conference hall grew warm debating this issue, even before an electrical failure knocked out the ceiling fans. A deep division emerged on the question of a Pleistocene human presence at Pedra Furada. The divide was partly along national lines, but for reasons that have little to do with politics. The question of antiquity looks very different depending on the intellectual tradition from which one emerges. North Americans tend to be more sceptical than most, but then we brought Calico into the debate, and have much to atone.
Tool riddle
Adovasio, Dillehay and I had travelled together for much of the trip to Pedra Furada and, realising we were of like minds, decided that on our return we would write a joint technical assessment of what we鈥檇 seen (this appeared in Antiquity in December 1994). Of greatest concern to us were the 595 tools from the Pedra Furada phase.
In deciding whether humans rather than nature could have made them, Parenti used several criteria including the removal of more than three flakes, edge angles of less than 90掳, a 鈥渓ogic鈥� to the flake removal pattern and where they had been found 鈥� the theory being any specimen found in the rear of the shelter had to have been carried there by people. We lost faith in the final criterion when we saw the rearward slope of the natural strata: rocks in the back of the shelter would have tumbled there naturally.
All the same, we were readily convinced that the specimens carefully laid out on tables at the conference could be stone tools made by humans. So much so we were scolded by members of the French contingent for paying too little attention to the specimens. And yet, we also saw that many of these pieces bore a disturbing resemblance to the naturally flaked quartzite cobbles that today litter the surface beneath the chutes at the site. There鈥檚 the rub: it isn鈥檛 enough to prove a specimen could be an artefact; one must also prove that it could not be a geofact. But this is easier said than done particularly with specimens as crude as these, and where nature has had such splendid opportunities for mischief.
Just how guilty nature might be became clear on our visit to the site itself. Standing on the pile of debris beneath the eastern chute, I craned my neck upward to catch a glimpse of the cobble layer high overhead. How long would a dislodged cobble take to make the Pedra Furada plunge, I wondered, and how fast would it be travelling when it smashed into the rock-strewn surface where I stood? Deciding the question wasn鈥檛 entirely academic, I thought it wise to retreat several steps before pondering the physics of free-falling rocks.
From a height of 100 metres, it would take a cobble a scant 4.51 seconds to reach the bottom. When it hit, it would be travelling at a velocity of 44.27 metres a second (roughly 160 kilometres an hour). If the rock missile has a mass of, say, 1-2 kilograms, the force of the impact will be considerable 鈥� far more than a flintknapper (someone who fashions stone tools) could muster. Gravity alone, it appears, can provide the muscle to flake these cobbles. Repeatedly, too. The chances are that flakes would be driven off when the cobbles first impacted, when they bounced into other rocks after hitting and again when they were struck by later falling stones. And as if that weren鈥檛 enough, they would also be flaked as they were agitated by the torrents of water that had flushed through the shelter throughout prehistory.
But could such natural impacts have produced the apparently human flaking patterns seen on the Pedra Furada specimens? The FUMDHAM team was ready for this question. Parenti had gathered 2000 shattered cobbles from the modern debris piles beneath the chutes in and near the shelter. All were obviously flaked by nature, had fallen in the past decade or so and lay well above the archaeological levels at the site. Parenti was pleased to report that none of them exhibited the patterns of flaking observed among the sample of 595 artefacts from the Pleistocene levels in Pedra Furada.
Case closed? Not yet, for the issue is one of probabilities. Jacques Pelegrin, a French palaeolithic archaeologist who is also an expert flintknapper, admits that nature can flake stones in ways that mimic the Pedra Furada specimens. After all, the artefacts are not complex tools, but simple, little-modified cobbles and flakes. Nonetheless, he argued that it was extremely unlikely. He put the odds at less than 1 in 100. For the sake of discussion, let鈥檚 make the odds even smaller: 1 in 1000, say, or 1 in 10 000.
Of course, no matter how rare the chances, given sufficient time and raw material 鈥� Pedra Furada had plenty of both 鈥� nature can magnify even the slimmest odds to the point where geofacts occur in detectable frequencies. Consider this: the deposits in the Pedra Furada shelter span almost 50 000 years, during which time they have been receiving a steady bombardment of cobbles. The average rate at which they fell from the cobble layer is unknown, but the size of the modern debris piles, coupled with conservative estimates about the time they have been accumulating, would suggest that more than 10 million cobbles have taken the Pedra Furada plunge in the past 50 000 years.
That鈥檚 a potential yield, even at odds of 1 in 10 000, of roughly 1000 geofacts. Nature, it appears, has had ample opportunity to produce the artefacts that occur at the site, of which there were less than 600.
Still, why were there no geofacts in Parenti鈥檚 sample from the modern debris piles? The answer, again, is in the odds. If an event is expected to occur only once in 10 000 tries then the odds are very poor that it will occur in a sample of 2000. Increasing that sample by an order of magnitude would improve the chances that at least a couple of geofacts would appear. Increasing the sample even further, and 鈥� well, the Pedra Furada excavations may already provide the outcome of that experiment.
In the end, the issue comes down to context, as it so often does in archaeology. If the Pedra Furada specimens had been found in finegrained sediments, or even in other sites in the caatinga where a cobble layer did not loom high overhead, they would readily be accepted as human artefacts. But, found as they were amid countless flakes and flaked cobbles, their standing as artefacts becomes suspect.
Difficult choice
From the witness sections at the site, we could see that the excavation teams must have regularly faced the quandary of what was 鈥� or was not 鈥� an artefact. Which perhaps explains why, in picking through the 鈥渂ackdirt鈥� piles of material thrown away (they were extraordinarily tolerant of us), we found discards quite similar to specimens displayed as artefacts on the conference tables. Re-unite those discards with the apparent artefacts, and the rocks at Pedra Furada form an uninterrupted continuum from unbroken cobbles through ones slightly flaked to those relative few that look just like simple stone tools 鈥� even down to the seemingly deliberate design of flake scars.
There are other clues suggesting that humans may have occupied this site in the Pleistocene, though none so compelling or central as the purported artefacts. More troubling, the best of these clues 鈥� the stone hearths 鈥� also look suspiciously natural. That their frequency apparently rises and falls through the deposits in concert with changes in the number of artefacts and, it seems, with roof and debris fall rates, only heightens these suspicions.
In the end, our visit to Pedra Furada turned out to be a case of 鈥渨e saw, but we didn鈥檛 believe鈥�. As excavators of Meadowcroft and Monte Verde, Adovasio and Dillehay are hardly averse to the idea of a pre-Clovis human presence in the Americas. But more so than most, they are also keenly aware of the standards in this debate.
Of course standards of proof apply to sceptics too, and we haven鈥檛 proven the Pedra Furada specimens are geofacts. Guidon and her supporters insist there is subtle but sufficient evidence in the way the specimens are flaked to prove their human origins. Perhaps. But it鈥檚 clear where the main burden of proof lies. The specimens must be shown not to be geofacts before they can be accepted as artefacts.
Proving our suspicions are wrong can be done by increasing the sample from the modern piles of debris to see if geofacts lurk there, examining whether the purported artefacts (and a control sample of non-artefacts) display signs of human use, and clarifying the difference between the many discards and the few purported artefacts. There the case of Pedra Furada will stand or fall.