It has been argued that endurance running ability may have been important in hominin evolution, giving hominins an enhanced ability to scavenge by allowing them to reach carcasses before other terrestrial vertebrate scavengers. This would have allowed them to exploit the carcass before eventually surrendering it on the arrival of potentially dangerous large terrestrial scavengers. Here, we use a simple spatial model to evaluate the ability of competitors to hominin scavengers to find carcasses. We argue that both hominin and nonhominin terrestrial scavengers would often first have been alerted to available carcasses by overflying aerial scavengers. Our model estimates that nonhominin scavengers will generally be able to reach the carcass within 30 min of detecting a plume of vultures above a nearby carcass. We argue that endurance running over periods greater than 30 min would not have provided a selective advantage to early hominins through increased scavenging opportunities. However, shorter distance running may have been selected, particularly if hominins could defend or usurp carcasses from other mammalian scavengers.

The importance of meat eating on hominin evolution has been widely discussed (see Stanford and Bunn 2001 for a review). However, there has been much debate about the relative importance of hunting versus scavenging as methods of obtaining meat (e.g., Turner 1992; Domininguez-Rodrigo 2001; Bunn and Pickering 2010). Scavenging could have taken a number of forms. It could involve discovery of a carcass prior to its exploitation by other vertebrate meat eaters, or discovery of the abandoned remains of a predator kill; these methods are collectively called passive scavenging. Alternatively, scavenging could have involved usurping predators at their kills, driving them off and exploiting the remains of the carcass; a method generally called active scavenging.

The importance of active scavenging has been questioned by those who argue that it would require sophisticated teamwork and weaponry for hominins to displace predators such as lions (Panthera leo) or spotted hyenas (Crocuta crocuta) from a kill, because the predators—having invested in successfully capturing a prey item—would be reluctant to relinquish the carcass (Dominguez-Rodrigo and Pickering 2003). It has been further argued that it would be easier and safer for hominins to hunt prey themselves than to aggressively usurp other large predators from their kill (Blumenschine 1991).

As regards passive scavenging on the remains of kills that predators voluntarily leave after they have consumed all that they want or are able to eat, it has been argued that this would provide a poor source of meat because the primary predators can effectively consume much of the valuable parts of the flesh prior to abandonment, and small agile carnivores (such as the black-backed jackal Canis mesomelas) may also opportunistically take some material before (and especially after) the abandonment of the carcass by larger carnivores (Blumenschine 1991; Tappen 2001). Turner (1988) argued that sabretooth cats that were a part of some Pliocene carnivore guilds may have been less effective than extant large cats at defleshing carcasses, but this view remains contentious (see Van Valkenburgh 2001). It has also been argued that both the condition and types of bones that appear to have been exploited by hominins are not consistent with such a secondary type of carcass access (Bunn and Pickering 2010). This suggests that the discovery of carcasses by hominins prior to full exploitation by other large vertebrate carnivores may have been particularly important in any scavenging by early hominins, and it is this method that we will explore first in this article.

Recently, Bramble and Leiberman (2004) and Lieberman et al. (2007), developing ideas from Carrier (1984), have argued that endurance running ability may have played an important role in hominin evolution. One prominent strand to this argument is that endurance running may have given hominins an enhanced ability to scavenge by allowing them to reach carcasses before other large terrestrial vertebrate scavengers. They suggest that vultures would likely be the first vertebrate scavengers to find a carcass. Vultures have exceptional carcass-finding abilities: Houston (1986) found that vultures located 80% of chicken carcasses set out by him in dense tropical rainforest within 12 h. He argued that visual detection in more open environments such as savannah grasslands would be much quicker (Houston 1974). However, vultures characteristically circle in the air for long periods prior to landing at the carcass, and can be an effective long-distance visual indicator of the approximate location of a carcass to terrestrial scavengers—particularly as numbers build. Circling vultures are used by hunter-gather tribes, rangers, stockmen, law-enforcement agencies, wildlife photographers, and scientists as a means of finding dead animals (e.g., O’Connell et al. 1988). However, although hominins and other terrestrial scavengers might be equally able to detect the vultures, Lieberman et al. (2007) argue that the endurance running capability of hominins would have allowed them to reach the carcass first, especially in hot conditions. They argue that other scavengers would have had to move more slowly over such long distances or risk overheating.

The importance of endurance running to scavenging by hominins was questioned by Pickering and Bunn (2007). They argue that even if hominins arrived first at the carcass, in open savannah ecosystems they would have ultimately faced confrontation by later-arriving scavengers and so hominins would have avoided the dangers of such confrontation by searching for carcasses in riparian woodlands rather than open habitats. Such scavenging would not have required endurance running, because carcasses would only be discovered at short distances in such a complex environment. Bunn and Pickering (2010) also point out that spotted hyena, lion, and wild dog (Lycaon pictus) can be active even in hot midday conditions (see also Hunter et al. 2006), and therefore scavenging hominins would have faced competition for access to carcasses from other large carnivores. Our aim in this article is to quantify the strength of such competition to aid critical evaluation of the likely importance of scavenging through endurance running to hominin evolution.

A number of assumptions are essential to the argument that passive scavenging could have been an important driver of the evolution of endurance running ability in hominins. First, carcasses must be produced by processes other than predation, and we argue below that (based on evidence from extant large-mammal communities) this seems plausible. A second key assumption is that scavenging hominins would have been able to detect carcasses at sufficient distance that endurance running ability would provide a valuable improvement in speed of carcass access. We feel that detection from a distance by cueing on vultures seems a very plausible scenario, based on evidence from a range of modern humans, as discussed above. However, it is important to emphasize that while vultures would have provided a reliable cue of carcass location to hominins, that same cue would also have been available to competitor carnivores. All these carnivores have acute vision, and are known to respond to other cues of potential scavenging opportunities (such as the sight of animals giving birth or distant predation by cheetah [Acinonyx jubatus]: Hunter et al. 2006). It is difficult to imagine a situation where visual cues would be available to hominins but not to competitor carnivores in the same environment. Indeed, as Bunn and Pickering (2010) discuss, modern spotted hyena, lion, and wild dog have all been observed responding to cues from circling vultures to find scavenging opportunities (Schaller and Lowther 1969; Kruuk 1972; Schaller 1972). Our first aim in this article is to estimate how quickly one would expect nonhominin scavengers to reach a carcass, as this should allow evaluation of the potential temporal advantage that endurance running might have given hominins in terms of time at the carcass prior to the arrival of such competitors.

A key unknown in any early hominin scavenging is to what extent they could defend a carcass against a group of lion, spotted hyena or similar large carnivores, or usurp a carcass from such animals. Experimental approached have been tried (e.g., in the 1950s the Leakeys—nude and armed with only bone “tools”—tried defending carrion against large African predators [Cole 1975])—however modern carnivores presumably associate humans with firearms, so making such experiments of limited value. It seems plausible that a well-coordinated group of hominins armed with wooden thrusting spears might have been able to drive off lions or hyenas. However, the social structure and wooden tool use of such early hominins is entirely unknown. Hence, we will evaluate the link between scavenging and endurance running under the two alternative scenarios. First, we will consider the situation where hominins could not mount a credible threat to such large carnivores, such that they would not attempt to usurp them from a carcass and would have to surrender any carcass that they found first upon the arrival or other large mammalian carnivores (or risk potentially fatal attack). Then we will turn to the alternative where social sophistication and tool use was sufficiently developed to allow the defending and usurpation of carcasses. Note that we are only concerned with large group-living mammalian carnivores (such as modern-day spotted hyena and lion). Smaller or solitary carnivores such as modern cheetah (A. jubatus) and leopard (P. pardus) seem less likely to approach even a single unarmed adult hominin in daylight.

A model based on the assumption of vulnerable hominins

A key question associated with scavenging in this scenario is whether most large animals die as a direct result of attack by a carnivore, or whether they are killed by other causes (e.g., age, disease, parasitism, injury, or starvation) where a predator is not on hand to immediately consume them. Only those that die outside of a specific predator attack will be available intact to scavengers. The best available evidence from extant ecosystems seems to suggest that most large mammals (with body size larger than 150 kg) seem to die in other ways, whereas smaller animals are more likely to die in the grasp of a predator (Fritz et al. 2001; Sinclair et al. 2003). This view has been supported by calculations that compare the food requirements of carnivores to estimated new production of their prey (Houston 1979), which estimate that only 36% of herbivore production in the Serengeti is consumed by vertebrate carnivores. This view was, however, challenged by Owen-Smith and Mills (2008) on the basis of records of found carcasses over 46 years in the Kruger National Park, South Africa. They concluded that “almost all mortality was through the agency of a predator for ungulate species up to the size of a giraffe (800–1200 kg).” However, these records derive not from systematic searches of ground for carcasses but rather from ad hoc observations by ranger staff while patrolling. Rangers may have been more likely to detect carcasses associated with predators than those without, leading to overestimation of the importance of predators. Hence, on the basis of current information, it seems plausible that early hominins (alongside other vertebrate scavengers) may have had the potential to discover dead animals before they had been extensively consumed by predators.

Sinclair et al. (2003) estimate lion and hyena densities in the northeastern Serengeti as 0.22/km2 and 0.5/km2, respectively. Owen-Smith and Mills (2008) give equivalent densities of 0.09/km2 and 0.14/km2 for the Kruger National Park. Both these species often live in social groups, with four to eight adults being typical of lion prides (MacDonald 2009; Wilson and Mittermeier 2009). Hyena clan sizes are much more variable with 10–15 adults being typical (Kingdon 2003). Adult lions are approximately three times the body mass of spotted hyena (MacDonald 2009). Hence, if we take the Kruger densities to be conservative, the large carnivore density that hominins might have been competing with would be 0.41 hyena equivalents per km2 (0.14 + (3 × 0.09)), which assuming a large group size of 15, gives a density of groups of competitors to hominins of 0.027/km2 (0.41÷15). A similar calculation, using a large lion group size of eight gives 0.018/km2. Hence we will make the very conservative estimate that large mammalian competitors to hominins for carcasses existed at a density of 2 social groups per 100 km2 (equivalent to a density of 0.02/km2). This ignores singleton individuals not part of the social structure (which are common in lions; McDonald 2009) and other large carnivores in extant savannah ecosystems (such as leopard, cheetah, and wild dog).

The range at which one vulture can detect another in flight is often assumed to be 4 km in recent modeling studies (Jackson et al. 2008; Deygout et al. 2009, 2010; Dermody et al. 2011). This value can be traced back to Pennycuick (1971) who observed a vulture in flight detect a small party of eight vultures on the ground at this range. However, the numbers circling a carcass can easily approach 100 in the Serengeti (Houston 1974). Houston (1974) estimated that this plume might be detectable by other vultures at a range exceeding 35 km. If we conservatively assume that vultures can be spotted by terrestrial mammals from a range of 5 km, then we can use our density estimate to consider how likely it is that a scavenger group will be close enough to detect the vultures circling over a carcass. A circle of radius 5 km has an area of 25π km2. The smallest square that can accommodate this circle has edge length 10 km and thus an area of 100 km2. We have already estimated that on average, such a square area would contain two scavenger groups. The probability of randomly placing two groups in the square and neither of them being within the circle is simply given by


which equals 0.046. Thus, by this calculation, on over 95% of occasions we expect that a competitor group will be close enough to detect the circling vultures as soon as they form over a carcass, even using our highly conservative parameter values. The next question is how quickly we expect such carnivores to reach a carcass, having been alerted by vultures to its existence.

If we assume that the groups of mammalian scavengers are randomly located in the environment, and carcasses also appear at random, then the distribution of distances from carcasses to the nearest scavenger group can be seen as a Poisson process and the distribution of the area of the circle with the carcass at its center and the nearest scavenger on the circumference will be exponential with the mean being the inverse of the spatial density of the scavenger groups. If we again assume a density of 0.02/km2, this converts to a mean area of 50/km2, or a circle with radius 4 km. That is, the mean distance the scavenger group needs to travel is around 4 km. We must now consider how long it is likely to take the scavengers to cover this distance. Bramble and Leiberman (2004) cited the extensive study of Heglund and Taylor (1988) into the long-distance sustained speeds of quadrupedal mammals. That work provides the following allometric relation between preferred trotting speed (v, in m/s) and body mass (M in kg): v = 1.09M0.222. Again, to be conservative, we will use the lower bodyweight of spotted hyena, rather than a lion, and assume M= 60 kg, giving an estimated trotting speed of 2.7 m/s. This is perhaps half the endurance running speed of an athletic modern human (Bramble and Leiberman 2004), but would still allow such an animal to cover 4 km in 25 min. By the properties of the exponential function, 75% of the values are less than 1.4 times the mean. This means that we predict that 75% of carcasses will fall within a 4.7 km radius (at 4 m, a 17% increase in radius provides a 40% increase in area) of a nonhominin scavenger group and so be reached in 29 min of the discovery of the carcass by vultures.

Thus, our calculations suggest that if early hominins did have superior endurance running performance to other terrestrial scavengers in hot environments then this would likely not have translated into a substantial opportunity to scavenge fallen animals before competitor scavengers arrived. This occurs because nonhominin carnivores would have to cover relatively short distances (less than 5 km) from first detection of a carcass site to arrival at that site. Any endurance running advantage of hominins may be further weakened by the need to avoid blundering into the path of other large carnivores attracted to the same carcass, causing more cautious travel by relatively vulnerable hominins. Our prediction is that those nonhominin competitors would arrive at the carcass within 30 min of both of them and any nearby hominins detecting it at a distance. Hence, hominins would often arrive after their competitors and when they did arrive first they would only have a few minutes to exploit the carcass before competitors arrived. Early hominins may have utilized exceptionally sharp flakes knocked from larger rock cores in exploiting carcasses. Given their small size if such tools were to be carried in anticipation of carcass discoveries, and the short time period available, it may have been possible for fleshy parts to be cut from carcasses, but substantial disjointing (such as removal of a leg) seems less likely in the time available. More importantly, only relatively small pieces of meat could be carried to a place of safety (such as into trees) without substantially slowing the running of hominins.

It is important to remember that the 29 min estimated above is the time taken from discovery of the carcass by vultures and hence by distant nonhominin terrestrial scavengers, and thus represents an upper limit to the time that humans could exploit the carcass free from competition with other large carnivores, unless hominins could discover carcasses before vultures do. Except in closed wooded environments (as discussed later), the aerial view of vultures will give them a considerable advantage in food finding. Because vultures are often almost-obligate scavengers but cannot out-compete large mammalian carnivores, their survival is possible only by fast and efficient carcass discovery (Houston 2001). Further, vultures are most active in hot, dry conditions when they can soar most efficiently on thermals (Pennycuick 1971; Houston 2001); these being exactly the conditions when hominin endurance running has been argued to offer a competitive advantage. Hence, we consider our assumption that vultures find the carcass before hominins (and other terrestrial mammalian scavengers) to be very appropriate for the current study.

An important conclusion of our model then is that endurance running over periods of 30 min would not have provided a selective advantage to early hominins through increased scavenging opportunities. However, running for a few tens of minutes may have enhanced the ability to exploit small carcasses that could be dismembered quickly or carried to a nearby tree. In 20 min, a modern human walking briskly (1.6 m/s) can cover 1.92 km, whereas running (5 m/s), achieves 6 km in the same time. Hence, the ability to run for 20 min unbroken may increase the range at which it would have been sensible for humans to exploit cues from vultures to explore small carcasses before the arrival of other large carnivores. However, there are some important caveats to this observation. First, whether the carcass was small enough to be dismembered rapidly by hominins would not be known until the hominins arrive at the carcass, so this strategy would sometimes cause them to invest energy reaching a carcass that could not be exploited in the time remaining before arrival of other carnivores. Further, small carcasses might well be exploited by smaller carnivorous mammals not considered in our model; such as the black-backed jackal (C. mesomelas) of sub-Saharan Africa and the golden jackal (C. aureus) of northern Africa, both of which are capable of killing and quickly consuming prey at least up to the size of adult Thomson's gazelle (Eudorcas thomsonii) (Kingdon 2003). Lastly, this scenario does not represent the type of endurance running discussed by, for example, Bramble and Lieberman (2004), who define endurance running as “running many kilometers over extended time periods using aerobic respiration.”

Linking scavenging and endurance running when assuming hominins could defend themselves against large mammalian carnivores

If early hominids were able to defend a carcass against large mammalian carnivores and to drive them from a kill, then we would expect them to respond to detection of vultures, and travel toward carcasses. Further, we would expect there to be strong selection pressure to reach those carcasses quickly, because this will increase the value of the carcass at the time of arrival, when hominins are assumed under this scenario to be able to take control of it. However, it seems likely that this selection pressure would have related to running short distances quickly, rather than endurance running. We have previously assumed that the maximum distance at which vultures can be detected is 5 km. Modern athletes can run sustainably at 5 m/s, allowing them to cover 5 km in 17 min. Running such relatively short distances quickly might have been energetically advantageous for such well-defended hominids, because this might allow them to claim a substantial energetic reward. That is, it may allow them to usurp carcasses generated by predation before all the valuable tissue has been removed. It would also (based on our model above) significantly enhance their ability to reach carcasses generated by other means before the arrival of other large mammalian carnivores.

It is possible that our 5 km detection range for vultures is overly conservative. If we relax this and assume that hominids can detect vultures at such a distance that it takes them an hour's sustained running to reach the carcass, this might reasonably be described as endurance running. However, we expect that such a foraging tactic would not have been energetically attractive. Elite human athletes running aerobically for sustained periods require investment of approximately ten times their basal metabolic rate. At the end of this investment, scavenging hominins will almost always have to fight for the carcass and defend it (continuing their need for high energy activity), because our model predicts that a group of competitor large carnivores will almost always have found the carcass by this time (if they did not generate it themselves through predation). The carcass will normally have been open to consumption by the group of large carnivores for at least 30 min according to our model, and this will significantly reduce its value to late arriving humans. If vultures have been circling over the carcass for an hour, and if the large mammalian scavengers have exploited the carcass all they want within this period, then we might expect that as soon as they lose interest the remains would be picked over by large numbers of scavenging birds, as well as smaller mammals such as jackals. Thus, running for long periods of time would likely often offer scant energetic reward relative to investment for a group of armed hominids: especially relative to hunting game, utilizing the teamwork and weapons assumed under this scenario. Thus, we suggest that if hominins had the teamwork and weapons to allow defense against large mammalian predators, this would certainly have enhanced their ability to scavenge (allowing defense of carcasses discovered first and aggressive scavenging of carcasses). However, performance in these scavenging tactics would have been enhanced by the ability to cover short distances quickly and economically.


Like all models, ours is a great simplification of biological reality. In particular, we assume that both nonhominin large mammalian carnivores and potential opportunities to scavenge are randomly distributed in a uniform environment. Again, we have made a conservative assumption in this regard. In reality prey animal deaths will not be uniformly spread across the environment, but will be distributed nonrandomly (for example because of nonrandom distribution of forage quality) and we would expect scavengers to focus their attention in areas where their potential for finding food is highest. Thus our model will provide a conservative estimate of the time taken by nonhominin scavengers to access a carcass. However, our estimate is in good accordance with the observation of Cooper (1991) that hyena arrived at 79% of lion kills within 30 min. Indeed such is the speed with which spotted hyenas respond to sounds associated with a kill that playing audiotape of such sounds to attract them to vehicles is an established census technique (Mills 1996). Hunter et al. (2006) quotes the mean time of arrival of hyena at cheetah kills in their Serengeti study of over 400 kills as 24 min. This congruence between our quantitative model predictions and empirical observation could be interpreted either as strong support for our underlying model assumptions, or for the suggestion that the predicted arrival of nonhominin carnivores at carcasses is robust to violation of the exact assumptions underlying our model. In support of the second of these alternatives, both Cooper (1991) and Hunter el at. (2006) observe early arrival of hyenas at carcasses even at night or in the early morning when cues from vultures are not available. In both of these cases, carcasses were produced by big-cat predation and the noise of this may have provided sufficient cues to allow hyena to effectively find the location of the kills.

It might be argued that our assumption that large carnivore densities in the African early Pleistocene are broadly similar to those of current African Savannah systems overestimates this density, if these carnivores faced competition for food from hominins. This seems unlikely, indeed the problem is often to explain the apparent high species richness of carnivores in the fossil record (O’Regan and Reynolds 2009). Further, in at least the early part of the Pleistocene for at least some localities it has been suggested that the large carnivore guild on African Savannas was more diverse than now (Turner 1988; Van Valkenburgh 2001; O’Regan and Reynolds 2009). Van Valkenburgh (2001) also reviews the evidence that interaction with modern humans artificially decreases large carnivore densities even in places (like the Serengeti) without large-scale human settlement, and North American data suggestive of higher large carnivore biomass density in the early Pleistocene than in modern African plains. Therefore, the carnivore densities in our model are likely to be conservative.

In semi-arid regions (rather than grasslands), herbivore and thus carnivore densities would have been significantly lower than those considered here. Lower carnivore densities may provide more of an advantage to a putative endurance-running hominin that our modeling suggests. However, without vegetation to hold the substrate together, scavenging hominins might sometimes have had to cope with loose sandy substrates that are known both to increase the cost and dramatically reduce the speed of running in modern humans (Alexander 2006). Arid environments would certainly have provided a very low density of carcasses, with a higher fraction of these being consumed by highly mobile flying scavengers or spoiled microbially before any hominin or nonhominin terrestrial scavenger discovered them. In any case, many modern reconstructions of the environments relevant to early hominins are generally not arid (Pickering and Bunn 2007). The conclusion that we draw from our modeling is that endurance running is unlikely to have played an important role in early hominin exploitation of meat through scavenging, and thus that early access of carcasses for scavenging is unlikely to have been an important selection pressure on endurance running in hominins. Our model suggests that hominin endurance running (even at the levels of modern elite athletes) would not have allowed hominins long periods of access to carcasses prior to the arrival of other large scavenging mammals. For small carcasses, it is possible that if hominins arrived first then one or two individuals could lift the carcass and run with it to the safety of trees or a geographical feature such as a rock outcrop where the hominins could sequester their prize away from quadruped carnivores (cf. Bailey et al. 2011). However, given our model prediction that arrival of other large scavengers at the original site would likely only be minutes behind the hominins, that hominin running would be slowed by their burden, and that they would be trailed and harried by small scavengers (e.g., vultures and jackals); this would only be an effective strategy if the place of safety was very near (otherwise the large scavengers would catch up with the hominins). Hence, this scenario would again not cause selection for endurance running ability. Finally, we argue that if hominids had the social cohesion and tool use to allow them to defend kills against packs of lions then (although this would have certainly expanded the opportunity for scavenging), this would have selected for short-term running speed and not endurance running. Our modeling suggests that even if early hominins could detect carcasses from a much greater range than we have estimated, the ratio of energy invested in traveling for long periods to reach a carcass against likely energetic return from the carcass would not have been attractive compared to using such social cohesion and weaponry in hunting. Further, if carcass detection ranges are similar to those we estimate, then endurance running would not be required to reach those carcasses.

We do not interpret our results to mean that scavenging was not important to meat eating by early hominins, only that long-distance endurance running would not have been an important component of such scavenging. We agree with Bunn and Pickering (2010) that densely wooded areas might present a greater opportunity for hominin scavenging than the open environments where endurance running might be relevant. First, competition from other large scavengers might be reduced in woodland. Generally, forested ecosystems do not have pack-hunting large carnivores such as wolves, lion, and hyena, but solitary animals such as leopard, tiger (P. tigris), jaguar (P. onca), cougar (Puma concolor), and bears. It may be more feasible for a band of hominins to defend a discovered carcass from such solitary animals, although grizzly bears for example have demonstrated willingness to attack large groups of modern humans to obtain a meal. More importantly, trees would offer a place of refuge into which hominins could quickly retreat having dismembered a carcass (perhaps while simultaneously holding solitary nonhominin scavengers at bay). In this regard, characteristic carcass size is likely to be smaller, because physically complex environments such as forests generally select for smaller body size. Further, it seems unlikely that nonhominin scavengers would pursue hominins into the trees to compete for the carcass. Extant leopards rarely hunt in trees, perhaps because the dangers of falling during aggressive contests are too high (Bailey 1993). Hence, scavenging may have been important to hominin evolution, but we suggest that scavenging opportunities might have been more successful in wooded than in open environments, at least until hominin weaponry and social group structure became sufficiently complex to allow defense or usurpation of a carcass against groups of animals such as lion and hyena. Long-distance endurance running is unlikely to have been an important element of scavenging in hominin evolution. Modern humans do have a conspicuous endurance running ability, but we suggest that the selection pressure behind this, if it is linked at all to foraging, is more likely to lie in the active pursuit of live prey than in efficient exploitation of carrion. However, our arguments suggest that in some circumstances scavenging may have selected for running ability over distances of 5 km or less. It may be that the anatomical and physical changes associated with selection for improved short-distance running may also have enhanced performance over longer times and distances.

Associate Editor: D. Carrier


We thank H. O’Regan, D. Carrier, T. Pickering, H. Bunn, and another anonymous referee for very useful comments on previous versions of this article. We dedicate this paper to the memory of Alan Turner—late Professor of vertebrate paleontology at Liverpool John Moores University.