The correlation between human immigration into the Americas, the invention of Clovis spear points and the extinctions of megafauna (large mammals exceeding 44 kg) at the end of the Pleistocene has caused many to implicate human hunting in the event. Correlation, however, does not necessarily imply causation. The close of the Pleistocene also marked a prolonged period of global warming and climate changes. In the 1990s the “overkill” hypothesis fell out of favor both because of its original ties to the “Clovis First” hypothesis (1) and paleontological evidence regarding the pattern of megafauna extinctions (2). At the turn of the millennium however, mathematical models (3,4), indirect evidence for human and megafauna occupations (5), and evidence for the reality of Clovis specialization as hunters of large game (6) made human hunting a viable hypothesis again. Recently Barnosky et al.  argued that humans caused extinctions through multiple synergistic effects, and were likely responsible for extinctions in North America and Australia (although both the timing of human immigration and the extinctions in Australia have yet to be firmly determined), whereas extinctions in Europe were likely due primarily to climate change (7). Barnosky et al. also argue that megafauna extinctions in South America have yet to be properly studied, but are coincident with both human immigration and climate change (7). The debate however is far from closed.

Before delving into the arguments for and against human responsibility for megafauna extinctions, it is important to distinguish four potential causes of the extinctions. The first is that climate change was responsible for the extinctions. The remaining three are all in some form anthropogenic extinction models. The first of these is the blitzkrieg (rapid overkill) hypothesis, in which human hunters armed with Clovis spear points overhunted naïve American megafauna. Clovis spear point technology is named for a site in Clovis, New Mexico, where fluted spear points were found with mammoth remains (1). Clovis sites date generally between 11,500 and 11,000 B.P., around when a corridor opened in the glaciation between Alaska and the rest of North America (1). A second possibility is the overkill hypothesis, which does not necessarily involve Clovis technology or immediate extinction of megafauna falling before an expanding lens of human immigration. A third is a sitzkrieg model in which human immigration resulted in extinctions through a combination of hunting, fire, habitat fragmentation, introduction of exotic species, diseases, and modifications to food webs (7).

Large mammals such as the mammoth used to roam North America, but whether hunting played a role in their extinction is a subject of controversy.

Large mammals such as the mammoth used to roam North America, but whether hunting played a role in their extinction is a subject of controversy.

Meltzer largely argues that the presence of humans in the Americas prior to the invention of Clovis spear point technology refutes the blitzkrieg hypothesis (1). At the time, pre-Clovis archaeological sites, including Monte Verde in Chile, were still suspect (1), but Monte Verde is now generally accepted as pre-Clovis (7). Meltzer argues that the “apparent chronological correlation” between extinction of megafauna and the appearance of Clovis may not exist. He further further argues that, as defined by Martin, the hypothesis requires Clovis to be the first human culture in North America, and that mtDNA molecular clock, linguistic, and archaeological evidence make this unlikely (1). The confirmation of pre-Clovis cultures in the Americas would therefore appear to refute the overkill hypothesis. I would argue that while this may refute the hypothesis as presented at the time, this does not actually refute that humans or even human hunting using Clovis technology were responsible for megafauna extinctions. 

Beck takes a paleontological approach by looking at the geographic patterns of megafauna extinctions in North America (2). Beck does support the timing of megafauna extinctions as coincident with Clovis technology  (12-10 thousand years ago), but attempts to test the blitzkrieg models by looking at the distribution of the most recent remains of megafauna. If the blitzkrieg model is correct and humans advanced from an ice-free corridor in southwestern Canada, eliminating megafauna as they expanded into the United States, then megafauna would have been more likely to survive to later dates further from the initial point of intrusion. This implies that even though the fossil record is unlikely to preserve the last individual or even population of a particular species, statistically more terminal sites should be found in the Southern and Eastern portions of the ranges of megafauna. Only three of 11 megafauna taxa sampled by Beck have terminal sites in the southeastern halves of their ranges (2). Again this seems to refute the Clovis blitzkrieg model quite nicely, but rejecting this model of overkill does not mean that humans or even human hunting were not implicated in megafauna extinctions. Interestingly, Beck notes that the killing front he presumes is necessary for an overkill hypothesis is not needed by a mathematical model constructed by Whittington and Dyke (8), but that this model requires longer coexistence between humans and megafauna, which is precisely what Meltzer (1) uses in an attempt to refute the overkill hypothesis in general.   

Revival of anthropogenic models of megafauna extinctions consisted of further confirmations of the timing of megafauna extinctions and new mathematical models, which made overkill appear to be a reasonable mechanism for the extinctions. John Alroy argues that the coincidence of evidence for large human populations in the Americas by at least 13,400 years before present and confirmation of human hunting of at least some megafauna makes the overkill hypothesis plausible (3). Alroy therefore constructed a mathematical model of 41 prey species, some of which went extinct and others of which survived the end of the Pleistocene.  He incorporated differences in prey body mass, geographic ranges, population densities, population growth rates (as predicted by body mass), rates of primary production and caloric values of plants and small game food resources, human nutritional needs, maximal rates of increase for human populations, and the first appearance of significant human populations in the United States (3). Additional, less constrained components of the model included the number of humans entering the region, hunting ability, and hunting effort (it was assumed to be related to handling time and nutritional requirements and so “per capita kill rates never exceeded a low ceiling”). The single best fit scenario in the original paper correctly predicts the survival or extinction of 32 out of 41 species (78 percent). Extinction times are also accurately predicted. It also takes 410 years for human populations to exceed 10,000, which may mean that early human occupation was at levels unlikely to appear in the archaeological record. Alroy further states that if the model is modified so that humans were already present at low densities before the extinctions, rather than arriving as a small founding population, extinction rates increase (3).Grayson criticizes the model for assuming that no megafauna went extinct before Clovis times and states that only 15 of 35 extinct genera can be confirmed to have survived into Clovis times and that only mammoth kill sites are confirmed (9). Alroy counters that radiocarbon dates are scarce for all but six genera, all of which overlapped with Clovis, and (without directly referencing it) cites the Signor-Lipps effect which shows that extinctions which take place in rapid periods of time would appear to be staggered in the fossil record. Alroy also notes that kill sites are likely for mastodons and giant tortoises as well and that “kill sites for smaller species are not expected because smaller bones are fragile.” In addition, the model is conservative in predicting how the deaths of some animals came as the result of human hunting.  For example, it predicts that only nine percent of M. columbi deaths were due to hunting, while in reality nine out of 61 fossil sites are associated with kills (15 percent) (3). Slaughter and Skulan also point out an unrealistically low rm (species specific growth constant) used for some species, and argue that this would make megafauna extinctions too likely in the model (10). Alroy accepts their correction, noting that it was a programming error, and thanks them as the corrected model actually increases the predicted number of extinctions from 27 to 29 (3).

Johnson (2002) sought to identify characteristics of species that made them more likely to go extinct when compared with close relatives. Low reproductive rates were associated with both likelihood of extinction, and relatively large body size, but this was not a uniform threshold (10 kg in lemurs compared with 350 kg in bovid) (11). Low reproductive rates could have predisposed some species to extinction, even under low hunting pressures. Johnson goes so far as to state that the number of deaths due to human hunting necessary to cause the extinction of some species may have been so low that “archaeological evidence of killing would be very sparse and in many cases could well be effectively veiled by its rarity” (11). The percentages of survivors from arboreal or closed habitats when compared with extinct species from those habitats (68 percent versus 22.8 percent, p< 0.001 and 75 percent versus 42 percent p< 0.01, respectively) also support anthropogenic models (11). Barnosky et al. also acknowledge that since almost all the slow-breeding survivors in Australia, the Americas and Madagascar are nocturnal, arboreal, alpine, and/or deep forest dwellers, climate change alone seems less able to explain these extinctions than an overkill hypothesis (7).

It does not seem possible for models alone to resolve the question of whether megafauna were hunted to extinction, due to the number of assumptions in the models (4). The number of assumptions and the apparently ethereal nature of mathematical models lead those who oppose anthropogenic or overkill models of megafauna extinctions to demand archaeological evidence that megafauna extinctions could have been caused by humans, and could not have been caused by climate change. Guy Robinson, David Burney, and Lida Pigott use indirect evidence to show that humans arrive very close to the megafauna collapse, and that climate change occurred much later than the collapse (5). This evidence comes in the form of spores of the fungus Sporormiella, which is partial to the dung of large animals and is used to determine the population levels of megafauna; tree pollen, which is used to track climate; and microscopic pieces of charcoal, which indicate human activity at four sites in southeastern New York. Microscopic bits of charcoal have been shown to be good indicators of human arrivals on islands around the world. Using the much more abundant charcoal and fungal spore information allowed better temporal resolution than the vertebrate fossil record and provided a means of estimating megafauna abundance that would not be affected by the Signor-Lipps effect. However, this method is unable to distinguish between individual species. Their analysis revealed that first the megafauna population collapsed, indicated by a 10-fold decrease in Sporormiella spores. Soon after this, charcoal abundances jumped 10-fold. According to Robinson, Burney and Burney, human hunting of megafauna would have led to an overabundance of fuel for both human and natural fires. Only around 1,000 years later did pollen data show the last major cooling in the Pleistocene. The last megafauna bones in the fossil record appear during this last cooling period, indicating that a combination of anthropogenic effects and climate may have led to the extinction of the megafauna, but only after humans had drastically reduced the megafauna population numbers. Similar patterns of extinctions indicated by fungal spores and charcoal in Madagascar, where extinctions occurred in the last 2,000 years without climate change, show that climate change may not even be necessary to provide a “knock out punch” (5). 

Now that there does appear to be some evidence for anthropogenic reductions in megafauna populations, it is worth reevaluating whether Clovis peoples were specialized as big game hunters. Waguespack and Surovell note that while nearly everyone acknowledges that Clovis peoples killed megafauna occasionally, the bulk of their diet was made up of small game and plant resources (6). Many critics argue that big-game hunting would not have been a reliable subsistence strategy for a colonizing population. According to Waguespack and Surovell, a number of archaeological sites indicate that Clovis hunter-gatherers did utilize plants, small mammals, birds, fish, and reptiles for food, but presence of small game does not provide an adequate indicator of generalized foraging (6). Other arguments against big-game specialization include: 1. ig-game hunting is ethnographically rare and only occurs in restricted, homogeneous, low biodiversity environments; 2. Big-game hunting is sustainable as a subsistence strategy only when targeted prey is abundant and has high renewal rates; 3. The diverse environments present during Clovis times would have favored a generalized resource acquisition strategy (6).

Proboscidians like this mastodon were found in 79 percent of Clovis sites.

Proboscidians like this mastodon were found in 79 percent of Clovis sites.

The argument that big game hunting would have been unreliable for Clovis peoples assumes that big game were rare and potentially dangerous, and this may not have been the case. Additionally, although hunters are occasionally wounded by elephants, this “does not appear to have deterred modern hunter-gatherers from pursuing them” (6). 

Waguespack and Surovell’s analysis of whether Clovis peoples were specialized or generalized foragers compares the number of individuals hunted with the predicted encounter rates for particular prey items (encounter rate assumes that small prey are far more common than large prey) (6). By this definition a big game specialist may have a relatively small proportion of the diet made up of large prey, so long as it was utilized above the predicted encounter rate. If high ranked, but relatively large game dominated Clovis faunal assemblages, Clovis peoples would qualify as big game specialists. Small game would have been utilized, but at lower levels than expected based on encounter rate. Estimated population densities were used as proxy measures of encounter rates, and only relative population densities are used (these are estimated using relationships between body size and population density and ecological principles). Proboscidians, bison, and ungulates are the most consistent members of Clovis sites (79 percent, 52 percent, and 45 percent respectively), with rodents, tortoises and birds next (39 percent, 30 percent, and 30 percent respectively). A strong negative correlation exists between body size and presence in a Clovis assemblage, “indicating that the largest, least diverse and least abundant taxa are the most consistent members of Clovis faunal assemblages,” which is indicative of specialization (6). Testing the actual minus expected number of sites for each size class (assuming a generalized subsistence strategy) using a two-tailed Spearman’s ρ yields a value of -1, which strongly rejects the null hypothesis (a generalized subsistence strategy). Waguespack and Surovell acknowledge that the archaeological record may be unduly biased towards megafaunal kill sites, but state that this is “the only direct source of information on Clovis hunting behaviors” (6).  I find the biases towards megafauna kill sites, including the increased likelihood of finding kill sites relative to gathering communities, biases towards the preservation of large bones and biases towards finding large bones even when some bones from small game are present, to be too numerous and influential to take this information at face value as no attempts are made to correct for these biases, but the strength of the pattern is intriguing.

Furthermore, Waguespack and Surovell studied 92 hunter-gatherer populations and compared the proportion of large game in their diets with population density (6). Groups which derive more than 46 percent of their diets from hunting had mean population densities less than 0.25 people per 100 km2 (6). They argue that since Clovis peoples were colonizing an uninhabited landscape, big-game hunting was a viable subsistence strategy. Although there is now evidence for pre-Clovis inhabitation of the Americas, big game hunting may still have been a viable subsistence strategy for the initial colonizers (6). 

In light of this, it is worth reexamining the assumptions of Alroy’s 2001 model with regard to specialization. Alroy’s best fit model required that only 0.111 percent of the human diet come from meat at 5.82 to 10.86 persons per 100 km2. In fact, various iterations have 8.8 to 13.2 percent of human dietary needs (i.e., 194 to 290 kcal/person/day or about 64 to 97 g meat/person/day) being met by hunting large game (3). For comparison, the model used by Barnosky et al. correctly predicted the fate of 34 of 41 megafauna species with human population densities ~28 people per 100 km2 obtaining 30 percent of their diet from meat (7). 

As the mathematical models now seem quite plausible and the patterns of survivors versus extinct species seem inexplicable by climate change and easily explicable by hunting (7,11), it is worth considering comparisons to other systems. Barnosky et al. note that on islands, humans cause extinctions through multiple synergistic effects, including predation and sitzkrieg, and “only rarely have island megafauna been demonstrated to go extinct because of environmental change without human involvement,” while acknowledging that the extrapolation from islands to continents is often disputed (7). The case for human contribution to extinction is now much better supported by chronology (both radiometric and based on trace fossils like fungal spores), mathematical simulations, paleoclimatology, paleontology, archaeology, and the traits of extinct species when compared with survivors than when Meltzer and Beck rejected it in the 1990s, although the blitzkrieg model which assumes Clovis-first can be thoroughly rejected by confirmation of pre-Clovis sites. Grayson and Meltzer (12) argue that the overkill hypothesis has become irrefutable, but the patterns by which organisms went extinct (7,11), the timing of megafauna population reductions and human arrival when compared with climate change (5), and the assumptions necessary to make paleoecologically informed mathematical models for the extinctions to make accurate predictions all provide opportunities to refute the overkill hypothesis, or at least make it appear unlikely. However, all of these indicate human involvement in megafauna extinctions as not only plausible, but likely.


1. D. Meltzer, Annual Review of Anthropology 24, 21-45 (1995).
2. M.W. Beck, Paleobiology 22, 91-103 (1996).
3. J. Alroy, Science 292, 1893-1896 (2001).
4. B.W. Brook, M. J. S. Bowman. Proceedings of the National Academy of Sciences of the United States of America 99, 14624-14627 (2002).
5. R.A. Kerr. Science 300, 885 (2003)
6. N.M. Waguespack, T.A. Surovell American Antiquity 68, 333-352, (2003).
7. A.D. Barnosky, P. L. Koch, R. S. Feranec, S. L. Wing, A.B. Shabel. Science 306, 70-75 (2004).
8. S.L. Whittington, B. Dyke. Martin and Klein 446-482 (1984).
9. D.K. Grayson, Science 294, 1459-1462 (2001).
10. R. Slaughter, J. Skulan. Science 294, 1459-1462 (2001).
11. C.N. Johnson, Proceedings: Biological Sciences 269, 2221-2227 (2002).
12. D.K. Grayson, Journal of Anthropology 31, 133-136 (2003).