Hominin Stature, Body Mass, and Walking Speed Estimates Based on 1.5 Million-Year-Old Fossil Footprints at Ileret, Kenya

Home , Footprint, Homo ergaster, Ileret, Turkana Boy

Journal of Human Evolution 64 (2013) 556e568

Contents lists available at SciVerse ScienceDirect

journal homepage: www.elsevier.com/locate/jhevol

Hominin stature, body mass, and walking speed estimates based on 1.5 million-year-old fossil footprints at Ileret, Kenya

Heather L. Dingwall a,*,1, Kevin G. Hatala a,b, Roshna E. Wunderlich c, Brian G. Richmond a,d,* a Center for the Advanced Study of Hominid Paleobiology, Department of Anthropology, The George Washington University, 2110 G St. NW, Washington, DC 20052, USA b Hominid Paleobiology Doctoral Program, The George Washington University, 2110 G St. NW, Washington, DC 20052, USA c Department of Biology, James Madison University, MSC 7801 Harrisonburg, VA 22807, USA d Human Origins Program, National Museum of Natural History, Smithsonian Institution, Washington, DC 20560, USA article info abstract

Article history: The early Pleistocene marks a period of major transition in hominin body form, including increases in Received 30 April 2012 body mass and stature relative to earlier hominins. However, because complete postcranial fossils with Accepted 11 February 2013 reliable taxonomic attributions are rare, efforts to estimate hominin mass and stature are complicated by Available online 22 March 2013 the frequent albeit necessary use of isolated, and often fragmentary, skeletal elements. The recent discovery of 1.52 million year old hominin footprints from multiple horizons in Ileret, Kenya, provides new Keywords: data on the complete foot size of early Pleistocene hominins as well as stride lengths and other char- Bipedalism acteristics of their gaits. This study reports the results of controlled experiments with habitually unshod Body size Homo Daasanach adults from Ileret to examine the relationships between stride length and speed, and also ﬁ Human evolution those between footprint size, body mass, and stature. Based on signi cant relationships among Pleistocene these variables, we estimate travel speeds ranging between 0.45 m/s and 2.2 m/s from the fossil hominin footprint trails at Ileret. The fossil footprints of seven individuals show evidence of heavy (mean ¼ 50.0 kg; range: 41.5e60.3 kg) and tall individuals (mean ¼ 169.5 cm; range: 152.6e185.8 cm), suggesting that these prints were most likely made by Homo erectus and/or male Paranthropus boisei. The large sizes of these footprints provide strong evidence that hominin body size increased during the early Pleistocene. Ó 2013 Elsevier Ltd. All rights reserved.

Introduction Australopiths are estimated to have been relatively small compared with early Homo erectus or Homo ergaster (hereafter The late Pliocene and early Pleistocene mark a major transitional ‘H. erectus’). Based on hindlimb joint size, which is arguably one of stage in hominin evolution, with derived anatomical changes the best means of predicting body mass (e.g., Jungers, 1988a; Ruff, within the genus Homo including increased brain and body size, 2003; Gordon, 2004), Australopithecus afarensis specimens have potentially with decreased sexual dimorphism, reduced tooth size been estimated at 45 kg for inferred males and 29 kg for inferred suggesting dietary shifts, and elongated lower limbs that likely females (McHenry, 1992). Average stature for A. afarensis is also improved speed and energetic efficiency (Wood and Collard, 1999; small, with male and female estimates averaging about 151 cm and McHenry and Coffing, 2000; Wood and Richmond, 2000). The 105 cm, respectively (McHenry, 1991; McHenry and Coffing, 2000). increased body size and relative lower limb length (but see Pontzer, From these mass and stature predictions, as well as studies of size 2012) that distinguished some early Homo from Australopithecus and shape variation, it is clear that A. afarensis exhibited substantial resulted in a more derived hominin body shape that falls within the sexual dimorphism with regard to body size (Richmond and range of variation exhibited by modern humans (Richmond et al., Jungers, 1995; Lockwood et al., 1996; Plavcan et al., 2005; Scott 2002; Ruff, 2002). and Stroik, 2006; Gordon et al., 2008, 2010; contra Reno et al., 2003). Homo habilis estimates are also small, with male mean body mass and stature estimated at 37 kg and 131 cm, respectively. * Corresponding authors. Average female mass is estimated at 32 kg and stature at 100 cm E-mail addresses: [email protected] (H.L. Dingwall), kevin.g.hatala@ (McHenry, 1991; McHenry, 1992; re-analyzed in McHenry and gmail.com (K.G. Hatala), [email protected] (R.E. Wunderlich), [email protected] Coffing, 2000 excluding KNM-ER 1472 and 1481). It should be (B.G. Richmond). fl 1 Present address: Department of Human Evolutionary Biology, Harvard Uni- noted that although these body size calculations are in uenced by versity, 11 Divinity Ave., Cambridge, MA 02138, USA. the dearth of postcrania that can be reliably attributed to H. habilis,

0047-2484/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jhevol.2013.02.004 H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568 557 cranial size variation is consistent with small overall body size shorter and earlier period of growth relative to modern humans, in H. habilis s.s. On the other hand, mean estimates of H. erectus Graves et al. (2010) argue that KNM-WT 15000 was 154 cm tall at stature (male: 180 cm, female: 160 cm; Ruff and Walker, 1993) and his time of death and would have attained an adult stature of only mass (male: 66 kg, female: 56 kg; Ruff et al., 1997) suggest that the 163 cm. This estimate is lower than that based on Ruff’s (2007) substantial increase in East African hominin body size that took analysis, closer to Ruff and Walker’s (1993) average stature esti- place during the early Pleistocene occurred with the appearance of mates for female H. erectus and not much larger than McHenry’s the earliest H. erectus around 1.9 Ma. These estimates suggest an (1991) male stature estimates of A. afarensis. Ohman et al. (2002) especially significant increase in female hominin body size and a similarly claim that Ruff and Walker (1993) overestimated KNM- concomitant decrease in the level of sexual dimorphism in WT 15000’s stature at death. They propose a new estimate of H. erectus relative to earlier hominins. 147 cm, which they argue accounts for axial/appendicular pro- However, several recent discoveries of small cranial remains portions in H. erectus that differ from those of modern reference belonging to H. erectus, for example from Olorgesailie (Potts et al., populations. Based on newly identified rib and vertebral fragments, 2004) and Ileret, Kenya (Spoor et al., 2007), and postcranial re- Haeusler et al. (2011) conclude that the rib cage of KNM-WT 15000 mains from Dmanisi, Georgia (Lordkipanidze et al., 2007) and Gona, was symmetrical and question previous interpretations that scoli- Ethiopia (Simpson et al., 2008) suggest that H. erectus may have osis or other pathologies affected the skeleton. This, in turn, sug- shown considerably more size variation than previously thought. gests that disease may not have affected KNM-WT 15000’s Furthermore, the presence of these small H. erectus specimens proportions and supports Ruff and Walker’s (1993) original size raises questions about whether body size dimorphism in H. erectus estimates. It is clear from these and other studies that stature and was appreciably different than that of earlier hominins (Antón, mass in H. erectus are the subject of ongoing debate. Body size es- 2012). The recent discoveries at Dmanisi include postcranial ma- timates in other early Pleistocene hominins are even more uncer- terial from three adults and one adolescent attributed to H. erectus. tain given the scarcity of well-preserved long bones with confident Stature and body mass have been estimated for two of the adults, attributions to H. habilis or Paranthropus boisei (McHenry and which imply smaller adult body sizes than those predicted from Coffing, 2000). other H. erectus material. Stature estimates based on humeral, Conclusions about temporal trends in hominin body size have femoral, and tibial measurements for a ‘large’ adult individual from implications for the formulation of hypotheses about other aspects Dmanisi averaged 149.3 cm, while estimates of stature derived of hominin evolution, many of which are concerned with the shift to from the first metatarsal of a smaller adult individual yielded an more xeric climatic conditions in East Africa throughout the early estimate of 143 cm (Lordkipanidze et al., 2007). Body mass esti- Pleistocene (McHenry and Coffing, 2000; Antón, 2003; Bobe, 2011). mates for the larger individual based on joint dimensions of the Evidence of increased mass and stature lies at the heart of hypoth- humerus, femur, and tibia, averaged 48.8 kg. The smaller in- eses regarding behavioral and physiological changes in early Homo dividual’s mass was estimated at 40.2 kg based on first metatarsal (McHenry, 1994; McHenry and Coffing, 2000; Aiello and Key, 2002; joint surface dimensions (Lordkipanidze et al., 2007). A pelvis from Aiello and Wells, 2002; Lieberman et al., 2009; Pontzer, 2012). The Gona, attributed to an adult female H. erectus individual, has pro- early Pleistocene falls temporally between the Pliocene, character- duced stature estimates between 123 cm and 146 cm based on an ized by smaller-bodied hominins (australopiths and H. habilis s.s.), estimated femur length using the three major articular surfaces and the later Pleistocene, with species of Homo that certainly had preserved in the pelvis (Simpson et al., 2008). These estimates from larger body sizes (Grine et al.,1995; Carretero et al., 2012). Because of Dmanisi and Gona are all smaller than the male and female aver- the limited sample of hominin fossils known from the early Pleis- ages predicted for H. erectus prior to these discoveries (Ruff and tocene (between 2.0 and 1.0 Ma), and the further paucity of fossils Walker, 1993). from which it is appropriate to derive estimates of body size, new Some researchers have questioned the estimated body size of information from this time period is critical to clarify the tempo and male H. erectus, notably those estimates derived from the 1.53 Ma mode of hominin body size evolution. (millions of years ago) associated juvenile skeleton KNM-WT 15000 The recent discovery of three stratigraphically distinct hominin (the ‘Turkana Boy’). Ruff and Walker (1993) estimated for KNM-WT footprint assemblages dating to c. 1.52 Ma at the site of FwJj14E in 15000 a stature at death of 160 cm and mass of 48 kg. The same Ileret, Kenya, in the Koobi Fora Formation (Bennett et al., 2009; authors predicted adult long bone lengths for KNM-WT 15000, Richmond et al., 2010) not only increases the previously-known which they used to estimate adult stature (185 cm) and mass sample of hominin footprints from the Plio-Pleistocene, it also (68 kg) for the juvenile specimen assuming a human-like life his- provides opportunities to test hypotheses about hominin body size tory for H. erectus. More recently, Ruff (2007) reexamined KNM-WT during the early Pleistocene, a time in hominin evolution about 15000 and proposed new at-death body size estimates based on a which we know very little. Further, footprint trackways provide the juvenile comparative sample. Stature at death was estimated at only means for directly observing locomotor behaviors in the fossil about 157 cm and body mass at 50e53 kg (Ruff, 2007). Although record. these estimates are slightly different, they are similar to Ruff and In this study, we conducted controlled experiments with Walker’s (1993) earlier calculations of KNM-WT 15000’s body habitually unshod Daasanach subjects from Ileret, Kenya, in order size and would still result in similarly large adult stature estimates to establish relationships between stride length and speed, as well given a relatively human-like growth trajectory. If correct, these as between footprint size, stature, and body mass. Habitually un- estimates provide evidence that the shift to a larger body size and shod subjects are critical to such an analysis given that early stature, comparable with the sizes of modern humans (at least in Pleistocene hominins almost certainly lacked shoes (Trinkaus and males), occurred with the emergence of H. erectus. However, Shang, 2008), and the 1.52 Ma prints are distinctly barefoot skepticism regarding the accuracy of these size estimates has arisen prints. Furthermore, there exists substantial evidence that habitual based on questions about KNM-WT 15000’s age at death, growth, shoe use influences foot development and possibly biomechanics development, life history (Graves et al., 2010), and the possibility of (Hoffman, 1905; Wells, 1931; Sim-Fook and Hodgson, 1958; Barnett, spinal pathologies such as scoliosis (Lovejoy, 2005) and vertebral 1962; Ashizawa et al., 1997; D’Août et al., 2009; Lieberman et al., dysplasia resulting in disproportionate axial and appendicular el- 2010; Hatala et al., 2013). ements (Ohman et al., 2002; but see; Haeusler et al., 2011). Based More prints continue to be discovered within previously on several assumptions, including a younger age at death and a described trackways at FwJj14E that are preserved in at least three 558 H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568 distinct horizons (Richmond et al., 2010) representing different speed of travel for each experimental trial. Stride length was instances in time within a few thousand years of each other at c. measured as the distance from the touchdown of one step to the 1.52 Ma (Bennett et al., 2009; Behrensmeyer, 2011). Here, we pre- touchdown of the next step made by the same foot. Each stride sent new measurements of stride lengths and use experimentally length was measured using the printmaking foot when possible. To derived relationships to estimate walking speeds of the hominin account for any slight acceleration or deceleration from stride to printmakers. We also use our experimentally produced footprints stride, we measured velocity over the course of two gait cycles as the basis for inferring the statures and body masses of the completed within the same trial. The 2011 experiments were hominin individuals who made the fossil footprints at Ileret in or- completed using the same experiment design, but they were filmed der to evaluate the hypothesis that, relative to earlier taxa, body using a high-speed (210-Hz) digital video recorder (Casio EX-FH20) size increased in early Pleistocene hominins. and ImageJ (Rasband, 1997e2012) was used to measure stride lengths and speed in lieu of the Peak Motus system. A subset Materials and methods (n ¼ 48) of the video trials from 2010 were analyzed twice, once using each program, to ensure that there were no discrepancies Controlled field experiments were conducted in 2010 and 2011 between the results obtained from using the different software. We with 38 adult Daasanach subjects (19 males, 19 females) who live found no significant differences between the measurements taken near the northeast shore of Lake Turkana, Kenya. Subjects were using Peak Motus and those taken in ImageJ (p ¼ 0.327, Wilcoxon recruited and their informed consent was obtained in accordance signed rank test). with the policies of The George Washington University’s Institu- To address questions about how speed, body mass, and stature tional Review Board (#031030). The Daasanach are habitually un- could be inferred from footprints, parametric regression statistics shod throughout ontogeny and into adulthood, with only some of were used to assess the relationships between: 1) speed and stride the males wearing minimal footwear inconsistently beginning at length, 2) footprint length and stature, and 3) footprint area and the time of adolescence (personal observation and personal body mass. The relationship between speed and stride length was communication, A.K. Behrensmeyer, J.W.K. Harris). Relevant bio- evaluated using the speed of the experimental subjects for each metric data, including mass, stature, and foot length, were recorded trial and the value obtained from dividing the subject’s stride for each subject. Subjects were asked to walk and run along a length by footprint length. We used this value rather than the cleared, flat, open-ended 15 m-long natural surface trackway con- subject’s stride length alone in order to adjust stride length relative taining a 3 m-long calibrated space. In order to obtain a wide range to a measure of ‘size’ that is preserved in the fossil print surface, of speeds, we asked the subjects to travel at a comfortable walk, fast following previous work (Charteris et al., 1981). An alternative walk, slow run, and fast run (each speed was later measured, see method for calculating speed is based on the principle of dynamic below). A pit measuring 150 cm long, 50 cm wide, and 15 cm deep similarity, which hypothesizes that two individuals move in a was dug midway along the experimental trackway and was filled dynamically similar manner when they travel at equal Froude with sediment taken directly from a fossil footprint layer at the values. Froude is calculated as F ¼ v2/gl, where v is velocity, g is FwJj14E site (A1 layer sediment was used in 2010, sediment from gravity, and l is ‘characteristic length’, resulting in a dimensionless the lower level was used in 2011; Bennett et al., 2009; Richmond value (Alexander, 1984a). Alexander (1984b) has demonstrated that et al., 2010). The sediment was rehydrated to approximate the dynamic similarity models are more appropriate in cases that conditions under which the 1.5 Ma fossil hominin footprints were involve individuals of different sizes and proportions. For example, made by adding water to the sediment until test footprint depths dynamic similarity models have been used to make predictions were comparable with those of the fossil footprints. based on modern human data about walking speeds of hominin Each subject performed as many trials as necessary until at least taxa with significantly different proportions and smaller size, two usable trials were completed for each gait category. Trials in including A. afarensis (Alexander, 1984b) and Homo floresiensis which the subjects visibly adjusted their gait or targeted the sedi- (Vaughan and Blaszcyzk, 2008). Because the most appropriate ment patch were discarded from the dataset and repeated. Subjects methods for the current study depend on the extent to which the were instructed to focus on a distant point on the landscape and fossil hominin print-makers at Ileret were geometrically similar to most subjects showed no evidence of adjusting their gait. Many modern humans, we report the results using both approaches. Here subjects missed the sediment patch during their running trials, we use effective limb length (i.e., height of the greater trochanter) suggesting that they were not paying close attention to its location. as the characteristic length for Froude calculation. Froude is then During each trial, the subjects created at least one footprint in the regressed against dimensionless stride length in log space to rehydrated sediment patch, which was photographed in high res- determine the relationship between these values (after Alexander, olution with a Canon 5D Mark II (21 megapixel) camera, using a 1984b; Raichlen et al., 2008). standard 50 mm lens to minimize radial distortion. The length and Stature estimates were based on regressions of stature by breadth of each experimental print was measured from these footprint length. Body mass was regressed against both footprint photographs using ImageJ (Rasband, 1997e2012). Footprint length length and footprint area, calculated as the product of footprint was measured as the linear distance from the posterior-most point length and maximum forefoot breadth. The resulting equations of the heel impression to the distal extent of the impression left by from the least squares regression for each set of relationships (see the longest digit (usually the hallux, but occasionally the second Results, below) were subsequently used to predict: 1) speed, 2) digit). A Wilcoxon signed rank test revealed no significant inter- stature, and 3) mass for the 1.52 Ma Ileret hominins based on stride observer error for these footprint measurements (p ¼ 0.299; length, print length, and print area measured from the fossil foot- n ¼ 20). prints at FwJj14E. Body mass data were restricted to those collected Recording and digitization differed slightly for the data collected during the 2011 field season due to equipment malfunction in 2010. in 2010 and 2011. The trials completed during the 2010 field season These experiments demonstrated that there is variation in were filmed using a 60-Hz digital video recorder (Canon ZR50MC) footprint size for the same individual between different experi- from a lateral view in order to capture 2-dimensional kinematic mental trials, which should be taken into account in predictions for data for each trial. These video data were later imported, calibrated, the hominin prints. Experimental footprints were found to range in and digitized using the Peak Motus motion analysis software lengths that were both longer (potentially due to foot movements) (7.2.10). Using this program, we measured the stride length and and shorter (e.g., in shallow prints when the foot may not sink H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568 559 completely into the sediment) than the actual foot length for a p < 0.0001) when speed categories are evaluated separately, but all given subject. On average, footprint length measurements differed relationships are highly statistically significant (see Table 1). from the measured foot length by 0.4 cm (mean ¼ 1.7% error; Based on the principle of dynamic similarity, we also tested this maximum difference ¼ 4 cm, or 14.5% error). Furthermore, since relationship using Froude numbers. The regression of Froude on gait dynamics change between walking and running, we completed dimensionless stride length is also both strong and significant the regression analysis for each relationship noted above in two (r2 ¼ 0.93, p < 0.0001), slightly stronger than the relationship for ways: 1) with all data pooled, and 2) with data divided by gait the raw velocities. When evaluated by gait category, this relation- category. For our predictions, we used the equation (i.e., walk, run, ship is equally strong for walking and running speeds (r2 ¼ 0.85, or combined) most appropriate for each individual trackway based p < 0.0001), although slightly weaker than the regression that in- on the speed estimates for each trackway. cludes the entire dataset. Stature of Daasanach sample Stature measurements for all subjects (n ¼ 38) ranged from 154 cm to 184 cm. Male stature was Results found to range from 163 cm to 184 cm while female stature ranged from 154 cm to 177 cm. The stature measurement for the tallest Daasanach experiment results female is an outlier (see below and Fig. 7a). Evaluation of the mean stature for each sex (male ¼ 174.2 cm, female ¼ 162.5 cm) shows Speed in Daasanach sample Digital measurement from video of that Daasanach males are significantly taller than females the travel speed for all subjects across all trials yielded a range of (Student’s t-test; p < 0.0001). Based on these data, the Daasanach e ‘ ’ speeds from 0.73 m/s to 6.35 m/s (Froude 0.05 4.23). The walk show significant sexual dimorphism in stature. ¼ trials (n 148) were characterized by speeds ranging from Regression statistics were applied to these biometric data to e ‘ ’ 0.73 m/s to 2.54 m/s (Froude 0.05 0.61), while the run trials determine the relationship between the lengths of the experi- ¼ (n 124) yielded speeds ranging from 1.79 m/s to 6.35 m/s mental footprints and the statures of the print-makers. For all data e (Froude 0.31 4.23). The overlap of walking and running speeds (i.e., combined data from both walking and running trials), a sta- between 1.79 and 2.54 m/s may imply a wide range of variation tistically significant relationship was found between stature and (0.8 m/s) in the transitional speeds between walking and footprint length (r2 ¼ 0.60, p < 0.0001; Fig. 2). Because they running. However, it should be noted that 99.97% of the running represent biomechanically distinct gaits, data from running and trials measured were at speeds greater than Froude 0.5, the walking were evaluated separately. Results show a statistically e walk run transition theoretically determined by the dynamic significant relationship between footprint length and stature for similarity hypothesis (Alexander, 1984a), and 99.95% of the both the walk-only (r2 ¼ 0.61, p < 0.0001; Fig. 2) and run-only trials walking trials measured were at speeds less than Froude 0.5. (r2 ¼ 0.58, p < 0.0001; see Table 2). Thus, for the majority of the subjects in this study, the transition Mass of Daasanach sample Mass for Daasanach subjects from the from a walk to a run occurred around speeds of 2.2 m/s (Froude 2011 experiments (n ¼ 19) ranged from 43.0 kg to 62.0 kg, with a 0.5), which corresponds with that predicted by the dynamic mean of 51.1 kg. Just as for stature, females (n ¼ 9; mean ¼ 49.4 kg) similarity hypothesis. were found to be lighter than males (n ¼ 10; mean ¼ 52.6 kg), Speeds of each experimental trial were plotted against the ratio although this difference is not statistically significant (two-sample t of stride length to footprint length for each corresponding subject. test, p ¼ 0.2). There is considerable overlap in the range of mass Regression analysis of the experimental speed data shows a sig- measurements for both groups (43.0e60.0 kg for females and fi 2 ¼ ni cant relationship between stride length and speed (r 0.91, 45.0e62.0 kg for males) as well as in the 95% confidence intervals < p 0.0001; Fig. 1). This relationship is slightly stronger for running of the means for the sex specific means (females: 45.9e53.0 kg, 2 ¼ < 2 ¼ speeds (r 0.80, p 0.0001) than for walking speeds (r 0.73, males: 48.5e56.7 kg; see below and Fig. 8a). Therefore, although the Daasanach are sexually dimorphic with regard to stature, the limited data here do not show significant dimorphism in mass. The relationships between footprint dimensions and body mass were also investigated via regression analysis. We tested the ability of both footprint length and footprint area to predict mass and, for each case, we tested regression models for the combined range of speeds as well as for the walk-only and run-only data. All of the linear regression models were statistically significant (p < 0.0001) but overall, body mass was correlated more closely with footprint area than with footprint length (see Table 3) and thus area was used to estimate body mass from the fossil footprints. While the regression based on the combined dataset of footprint area measured from both walking and running prints is somewhat more tightly correlated with mass (r2 ¼ 0.58) than are the footprint areas measured from only walking prints (r2 ¼ 0.52), for our predictions, we chose to use the walking print model for the fossil prints characterized by stride lengths and inferred speeds that fall within the experimentally determined walking speed range (Fig. 3).

New fossil footprints

Several new footprints have been recognized (Richmond et al., 2010) since the initial announcement (Bennett et al., 2009) of fos- Figure 1. Linear ﬁt of stride length-to-footprint length ratio to speed for the full range of speeds measured demonstrating a signiﬁcant relationship between stride length sil hominin footprints at FwJj14E. Excavations of the print surface (adjusted for limb proportions) and speed. p < 0.0001, r2 ¼ 0.91, n ¼ 272. known as the ‘upper level’, conducted in July 2011, focused on 560 H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568

Table 1 Regression relationship between speed and stride length/average footprint length (SL/avgFPL).

Linear ﬁt for speed estimation

Gait category Linear ﬁt nR2 adj. S.E.E. Prob. > F Walk only Speed ¼0.38 þ 0.30*(SL/avgFPL) 148 0.73 0.20 <0.0001 Run only Speed ¼0.63 þ 0.41*(SL/avgFPL) 124 0.80 0.48 <0.0001 Walk and run Speed ¼1.39 þ 0.48*(SL/avgFPL) 272 0.91 0.41 <0.0001

collecting new data from the known print surface. These renewed breadths are reported in Table 4 for all measurable prints from the excavations resulted in the identiﬁcation of four previously un- upper, lower, and A levels. recognized hominin prints within the FUT1 trackway (Bennett et al., 2009). The FUT1 trackway was originally interpreted as a Estimates from fossil hominin footprints trail made by a single individual (Bennett et al., 2009). However, the position of the prints relative to each other, including the overlap of The stature, mass, and speed estimates derived directly from the two left footprints (FUT1-7A and B; Fig. 4), indicates that prints of at Daasanach regressions are based on the assumption that foot size least two individuals are preserved in the trail. Based on compa- and body size proportions of the Ileret printmakers were comparable distances between certain prints, similarities in print orien- rable with those of our modern human sample. However, three tation, and subtle differences in morphology (e.g., slightly more hominin species, H. erectus, H. habilis, and P. boisei, are known from abducted hallucal impression), we have developed a new hypoth- c. 1.5 Ma sediments at Koobi Fora, within a few kilometers of esis regarding which prints are most likely to belong in the same FwJj14E (Feibel et al., 1989; Spoor et al., 2007). Therefore, any of trackway and thus represent the same individual. Fig. 4 provides a these taxa could have made prints in sediments of this age and it is schematic representation of this new working hypothesis for the possible that multiple species were active at FwJj14E, especially FUT1 trackway. We use this hypothesis in our calculations of speed, given the fact that footprints are found in three temporally distinct stature, and mass predictions for the 1.52 Ma hominins. The accu- strata (Fig. 4). To assess the validity of assuming foot proportions racy of the speed estimates depends upon accurate stride lengths, like those of modern humans, we investigated the pedal pro- which in turn depend upon correct footprint association. Stature portions of H. erectus. This was not possible for H. habilis and and mass estimates, however, can be calculated for individual P. boisei since securely attributed postcranial fossils relevant to the prints, so they are not inﬂuenced by the accuracy of the trackway assessment of foot-body size proportions are not known for these hypothesis, particularly as the prints in this trail are comparable in species (but see below). The site of Dmanisi yielded an adult size with one another. H. erectus partial skeleton (D4166) that preserves an associated Two additional hominin prints (A2-I2 and A2-I3) have been femur and third metatarsal (MT3; Lordkipanidze et al., 2007). unearthed in a sedimentary level between the ’lower level’ and While foot length and stature cannot be directly measured on this, ’upper level’ (Richmond et al., 2010). This level is designated as the or any other, early H. erectus fossil, the lengths of the MT3 and fe- ‘A level’ (‘A’ was preferred over naming it ‘middle level’ to follow a mur provide the best available measure of foot length, leg length, naming scheme that allows us to name additional levels as we and stature as they are among the longest bones in their respective unearth them at this site). These prints are preserved in isolation; anatomical regions. The MT3:femur length proportions of D4166 they provide no data regarding speed. However, their lengths and fall securely in the range of modern humans (from the Terry breadths (Table 4) provide data regarding hominin foot size from collection, Smithsonian Institution’s National Museum of Natural which stature and mass can be estimated. Footprint lengths and

Figure 2. Linear fit of footprint length to stature for the ‘walk’ speeds only showing a Figure 3. Linear fit of footprint area to mass for the ‘walk’ speeds only showing a significant relationship. p < 0.0001, r2 ¼ 0.61, n ¼ 38. This relationship is used to significant relationship. p < 0.0001, r2 ¼ 0.52, n ¼ 19. This relationship is used to predict stature for the FwJj14E hominins. predict body mass for the FwJj14E hominins. H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568 561

Figure 4. a) Schematic representation of the stratigraphic orientation of the three footprint surfaces. The FLT (i.e., ‘Lower Trackway’) and FUT (i.e., ‘Upper Trackway’) surfaces have been published previously (Bennett et al., 2009). The A level contains two new isolated prints from different individuals. Stratigraphic section adapted from Behrensmeyer (2011). Sediment texture scale: C ¼ clay, Z ¼ silt, S ¼ sand, G ¼ gravel. b) Diagram showing our current working hypothesis for the FUT1 trackway, which is now believed to represent at least two individuals. Red footprints (dark grey in print version) represent the FUT1A trackway (individual 2) and blue footprints (light grey in print version) represent the FUT1B trackway (individual 1). (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

Figure 5. Comparison of third metatarsal length-to-femur length proportions Figure 6. Estimated speeds for each hominin trackway from FwJj14E based on the measured for eastern gorillas (G. beringei, n ¼ 4), western gorillas (G. gorilla, n ¼ 13), Daasanach regression using raw speed. Dashed box represents the range of speeds at the Dmanisi ‘large’ adult (H. erectus; Lordkipanidze et al., 2007), modern humans which our subjects were observed to transition from a walk to a run (2.0e2.3 m/s). (H. sapiens, n ¼ 20), and chimpanzees (P. troglodytes, n ¼ 20). The Dmanisi specimen Note that the majority of these estimated speeds are certainly walking speeds, with the falls squarely within the range of modern human proportions and has a distinct exception of the estimate for the FLT1 trackway, which may represent a slow jogging pattern from that of the African great apes. speed. Error bars represent 95% conﬁdence intervals. 562 H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568

Figure 7. a) Daasanach measured stature (separated by sex) compared with hominin stature estimates. b) Hominin stature estimates based on the Daasanach regression separated by individual trackway. Error bars represent 95% conﬁdence intervals.

History) and are distinct from those of gorillas and chimpanzees estimates would not differ substantially if the Ileret printmakers had (American Museum of Natural History collections; Fig. 5). Thus, if more primitive foot proportions. H. erectus individuals were indeed responsible for the FwJj14E In light of the general similarity in foot:body size proportions footprints, our regression models are based on reasonable as- of early hominins and modern humans, as well as the comparability sumptions about foot to body proportions. between the sizes of the Ileret prints and those made by our Daa- To assess how the results would differ if the Ileret printmakers had sanach subjects, we prefer the geometric similarity models (e.g., more primitive foot:body size proportions, we examined the only stride length/foot length) in our analyzes. Nonetheless, we also available estimates of foot and stature lengths for an earlier hominin, report speed predictions based on a dynamic similarity model for namely A. afarensis, which may approximate the primitive condition comparison. The dynamic similarity prediction equation was from which H. habilis and P. boisei evolved (until good evidence of derived by regressing Froude number on relative stride length in relative foot size is available for H. habilis and P. boisei,weworkwith log space using effective lower limb length rather than stature for the parsimonious assumption that they retain relatively primitive characteristic length, since the former yields a better relationship anatomy, such as that seen in A. afarensis). Fleshy foot length in AL and is theoretically more appropriate (Alexander, 1984b; Raichlen 288-1 has been estimated as 16.5 cm based on measurements of a et al., 2008). Below, we provide estimates derived directly from composite foot skeleton including the scaled length of the AL 333-115 the Daasanach regressions (assuming modern human proportions toe and metatarsal head region, and the scaled length of much of the for the Ileret printmakers) and an estimate of how those values remaining foot based on OH 8 (White and Suwa, 1987). A larger foot would differ if the proportions of the Ileret hominin printmakers length estimate of 17.26 cm was reached based on the relationship were more primitive (i.e., like those of A. afarensis). between talus length and subtalar length (Jungers,1988b). Combined Speed estimates from fossil footprints The experimentally-derived with AL 288-1’s femur length (28.1 cm), the relative fleshy foot length regression equation representing the relationship between speed (100 fleshy foot length/femur length) is 58.7e61.4, well below that and the ratio of stride length to footprint length was used to of bonobos (73.9), chimpanzees (82.5), and LB1 (70.0; the type estimate speeds from five partial trackways from two stratigraphic specimen of H. floresiensis), and slightly above the values observed for levels at the FwJj14E site. Stride lengths of fossil trackways were samples of modern humans (54.2, 54.5 for small-bodied pygmies; measured as the distance from the heel of one print to the heel of Jungers et al., 2009). When comparing fleshy foot length estimates the next print made by the same foot where possible. For prints with AL 288-1’s estimated stature (106.7 cm, or 30600; Jungers,1988b), with poorly preserved heel impressions, stride length was the foot:stature ratio for AL 288-1 is 0.155e0.162, which is signifi- measured as the distance from hallux to hallux. Where only the cantly smaller than that observed for bonobos (mean ¼ 0.188; range: length of a single step was preserved, the step length was 0.171e0.200; data from Coolidge and Shea,1982) and above the mean measured and doubled to approximate stride length. but within the range of our Daasanach sample (mean ¼ 0.150; range: The trackway of two prints from the oldest stratigraphic level 0.138e0.162). This suggests that the stature, mass, and speed (the lower level) has a stride length/foot length that overlaps those

Table 2 Regression relationship between stature and average footprint length (FPL).

Linear ﬁt for stature estimation

Gait category Linear ﬁt nR2 adj. S.E.E. Prob. > F Walk only Stature ¼ 74.47 þ 3.72*avgFPL 38 0.61 5.40 < 0.0001 Run only Stature ¼ 71.47 þ 3.87*avgFPL 37 0.58 5.70 < 0.0001 Walk and run Stature ¼ 73.29 þ 3.78*avgFPL 75 0.60 5.40 < 0.0001 H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568 563

Figure 8. a) Daasanach measured mass (separated by sex) compared with hominin mass estimates. b) Hominin body mass estimates based on the Daasanach regression separated by individual trackway. Error bars represent 95% conﬁdence intervals. observed for the walking and running trails in the Daasanach ex- are also slow walking trails. We have estimated that FUT2 was periments. Using the equation that incorporates the full range of made at a speed of 1.00 0.20 m/s and FUT3 was made at a speed of speeds yields an estimate of 2.69 m/s 0.41 m/s (95% prediction 0.45 0.20 m/s. interval), compared with 2.21 m/s 0.20 m/s for the walk-only Speed estimates derived from the regression of Froude speed on regression and 2.94 m/s 0.48 m/s for the run-only equation dimensionless stride length, based on the walking data, are similar (Fig. 6). These speeds are consistent with a slow run or fast walk; to those reported above. The speeds of the FUT1 trails were pre- the overlap between walking and running gaits established by our dicted to be 0.45 0.13 m/s (95% PI; Froude 0.02) for individual 1 experiments occurs around 2.2 m/s. Because the FLT1 trackway and 0.62 0.12 m/s (Froude 0.04) for individual 2. The dynamic only consists of two footprints, in other words a single step, more similarity speed estimate for the FUT3 trackway was also slightly work is necessary to determine whether the morphology of these lower than the estimate based on the raw speeds at 0.39 0.13 m/s prints can conclusively establish whether the prints were made by (Froude 0.02). The FUT2 and FLT1 trackways, however, produced a walking or running gait. slightly larger estimated speeds than those reported above. Using The other four trackways are in the youngest stratigraphic level this dimensionless model, we calculated speeds of 1.08 0.12 m/s (the upper level). The walk-only regression equation was used to (Froude 0.11) for the FUT2 trackway and 2.59 0.12 m/s (Froude estimate the speeds at which these fossil prints were created 0.74) for the FLT1 trackway (see Table 5), suggesting that the FLT1 because their stride length/footprint length values fell within the individual may have been traveling at a slow run. Nonetheless, the range of those observed during walking trials in the Daasanach differences between the two sets of estimates are only slight; in experiments. Two of these trackways (FUT1A and B) are unique fact, the conﬁdence intervals for the speeds calculated by the two within the FwJj14E assemblage in that each trackway contains different methods overlap with each other. multiple prints (Fig. 4). Whereas all of the other trackways consist If the makers of the FwJj14E footprints instead belonged to a of only one or two useable prints, each of the FUT1 trackways different taxon that had proportions different from those of contains six footprints. Estimates of both FUT1 trails show that H. erectus and modern humans (e.g., longer feet relative to leg these two individuals were traveling at comparable speeds (FUT1A: length and stature), our regression models would have under- 0.68 0.20 m/s; FUT1B: 0.49 0.20 m/s), both of which are estimated the hominins’ traveling speeds. To test how such differ- characteristic of a slow walking gait (Fig. 6). Based on our calcula- ences in limb proportions would affect these estimates, we have tions, the remaining trackways in the upper level (FUT2 and FUT3) calculated the FwJj14E speeds using the limb proportions known

Table 3 Regression relationships between mass and average footprint length (FPL) and area (FP area).

Linear ﬁt for mass estimation

Gait category Measurement Linear ﬁt nR2 adj. S.E.E. Prob. > F Walk only Footprint length Mass ¼ 4.71 þ 1.82*avgFPL 19 0.51 3.80 0.0005 Run only Footprint length Mass ¼ 1.99 þ 1.97*avgFPL 19 0.51 3.70 0.0004 Walk and run Footprint length Mass ¼ 3.94 þ 1.87*avgFPL 38 0.51 3.70 <0.0001 Walk only Footprint area Mass ¼ 23.64 þ 0.11*avgFP area 19 0.52 3.70 0.0003 Run only Footprint area Mass ¼ 18.94 þ 0.13*avgFP area 19 0.62 3.30 <0.0001 Walk and run Footprint area Mass ¼ 21.57 þ 0.12*avgFP area 38 0.58 3.40 <0.0001 564 H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568

Table 4 161.9 5.4 cm, which is the smallest stature estimate from the * Hominin footprint lengths and breadths. upper level. The print lengths from the FUT1B trackway yield a Individual Footprint Length (cm) Breadth (cm) stature estimate of 174.3 5.4 cm for individual 1, while the FUT1A FUT1-3 25.5 10.2 estimate for individual 2 (FUT1A) is smaller at 169.0 5.4 cm. FUT1-4 e 8.7 We estimate a stature of 169.4 cm for the individual in the oldest FUT1-5 27.0 10.4 stratigraphic layer stature (Fig. 7b). In sum, the stature estimates FUT1-6 25.0 9.4 for the seven individuals, based on 14 fossil prints, range from FUT1-7A 26.0 10.0 FUT1-8 23.5 9.3 153.0 cm to 185.8 cm, which is similar to the Daasanach stature FUT1B FUT1-1 28.5 10.5 range (154e184 cm; Fig. 7). The error in footprint length due to FUT1-2 e 9.9 variation within an individual (see ‘Methods’) results in a 3% FUT1-4i 27.0 9.04 error in stature estimation. Thus, incorporating this error into our FUT1-4ii 25.0 9.2 estimates would expand the range of possible statures for the FUT1-7B e 9.62 FUT2 FUT2-1 30.5 e Ileret hominins to 148.4e191.4 cm. FUT2-2 29.3 11.4 If the Ileret printmakers had AL 288-1’s estimated foot:stature FUT3 FUT3-1 23.5 8.1 proportions (0.155e0.162), their stature estimates would range FLT1 FLT1-1 25.5 9.0 from 130.2 cm for the smallest footprint to 192 cm for the indi- FLT1-2 25.5 8.7 A2-I2 A2-I2 21.1 7.7 vidual with the largest prints (no prediction limits are available A2-I3 A2-I3 26.5 10.0 when using a simple ratio instead of a regression), indicating large statures for the Ileret printmakers regardless of assumptions about * More subtle prints that do not preserve reliable length and breadth measurements have been omitted. relative foot proportions (see Table 5). Mass estimates from fossil footprints The experimentally determined relationships between footprint area and body mass (see from AL 288-1 (Jungers, 1982; Pontzer et al., 2010), the results of Table 4 for equation) were used to estimate body mass from which are reported in Table 5. The reduced relative leg length es- measurements of the FwJj14E footprints. Only the equation derived timate (about 7.3% shorter) based on AL 288-1’s proportions from measurements of walking footprints was used because the resulted in slightly faster speed estimates, but do not change the fossil stride length/footprint lengths and inferred speeds were most conclusions. Thus any calculation method shows that these in- similar to those of the walking footprint experiments, with the dividuals were walking slowly, with the exception of the FLT1 trail exception of individual FLT1, for which combined walking and with a stride length reflecting an individual traveling at a slow running data were used. Estimated masses for the hominin running or fast walking pace. individuals range from 41.5 kg to 60.3 kg (3.8 kg, 95% PI). These Stature estimates from fossil footprints Living stature estimates for mass estimates fall within the range of Daasanach masses reported the Ileret hominins were calculated using the equation that de- above (see Fig. 8a). scribes the relationship between each Daasanach subject’s stature The largest individual is again attributed to the FUT2 trackway. and his or her experimental footprint lengths. For all fossil track- Likewise, the smallest mass estimate is associated with the A2-I2 ways made at walking speeds (see above), the regression data were individual, who also exhibited the smallest stature estimate. limited to the footprints produced during walking trials; pre- Overall, the stratigraphic levels containing tall individuals also dictions for individual FLT1 used combined walking and running produced heavy mass estimates-upper level: 44.5e60.3 kg; A level: data. Stature was estimated for a total of 14 hominin prints from 41.5e52.7 kg; lower level: 48.4 kg, all with estimate intervals of three different stratigraphic levels at FwJj14E: two prints from a 3.8 kg. A 5.7% error in mass estimation resulted from the variation single individual from the lower level, two isolated prints of two noted in experimental footprint size. With this source of error individuals from a middle level (A2), and 11 prints representing included, a conservative estimate of the possible mass for these four individuals from the upper level that, all together, likely hominins would widen to a range between 39.1 kg and 63.7 kg. represent seven individuals (Table 4). Stature estimates were based If the Ileret printmakers had AL 288-1’s estimated foot on at least two prints for each individual except in the case of the length:body mass proportions (ranging from 0.543e0.632, using two isolated prints, A2-I2 and A2-I3, which yield estimates of the fleshy foot lengths estimates above and mass estimates of 153.0 5.4 cm (95% PI) and 173.1 5.4 cm, respectively, and 27.3 kg and 30.4 kg from Jungers, 1982; Jungers, 1988c; McHenry, thus are not likely to have been made by the same individual. 1992), their mass estimates would range from 33.4 kg to 55.1 kg The youngest level (upper level) contains the largest stature (no prediction intervals are available when using simple ratios). estimate of the assemblage, 185.8 5.4 cm, which is attributed to These estimates overlap with the range previously reported for the FUT2 trackway. The FUT3 trackway produced an estimate of A. afarensis (McHenry, 1992; McHenry and Coffing, 2000), but body

Table 5 Comparison of estimates using proportions of H. sapiens versus A. afarensis.

Hominin individual Speed (m/s)* Stature (cm) Mass (kg)

H. sapiens A. afarensis H. sapiens A. afarensis H. sapiens A. afarensis estimate estimate estimate estimate estimate estimate FUT1A 0.62 0.12 (Froude 0.04) 0.68e0.71 (Froude 0.06) 169.0 5.4 156.8e163.9 50.6 3.8 40.2e46.8 FUT1B 0.45 0.13 (Froude 0.02) 0.48e0.51 (Froude 0.02) 174.3 5.4 165.6e173.1 52.1 3.8 42.5e49.4 FUT2 1.08 0.12 (Froude 0.11) 1.11e1.17 (Froude 0.13) 185.8 5.4 184.6e192.2 60.3 3.8 47.3e55.1 FUT3 0.39 0.13 (Froude 0.02) 0.44e0.46 (Froude 0.02) 161.9 5.4 145.1e151.6 44.5 3.8 37.2e43.3 y FLT1 2.59 0.12 (Froude 0.74) 2.83e2.98 (Froude 0.96) 169.7 5.4 157.4e164.5 48.0 3.5 40.3e47.0 A2-I2 ee153.0 5.4 130.2e136.1 41.5 3.8 33.4e38.9 A2-I3 ee173.1 5.4 163.6e171.0 52.7 3.8 41.9e48.8

* Calculated using Froude speed equations. y Estimates calculated using the combined regression equations that include all gait categories. H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568 565 mass estimates for most individuals are near or above the upper even the estimates that were calculated using A. afarensis pro- limit for A. afarensis, and are more comparable with the mass es- portions indicate larger body size. Indeed, they do overlap with timates of H. erectus and the upper end of the P. boisei range (see australopith body size estimates, but most are larger than sizes Tables 5 and 6). predicted for australopiths such as A. afarensis or Australopithecus africanus, which is even smaller-bodied (McHenry, 1992). Further- Discussion more, foot:leg proportions for A. afarensis are tenuous and should be treated with caution, as they are based on an amalgamation of Speed specimens from multiple hominin taxa (White and Suwa, 1987)or on very fragmentary remains (Jungers, 1998a). We also note that The speeds of the hominin individuals differ between the two stature estimates for A. afarensis (McHenry, 1991, Table 6) are likely stratigraphic horizons that preserve trails (the upper and lower to be overestimated because they are based on a femur:stature ratio layers; prints A2-I2 and A2-I3 in the A layer are isolated). The lower observed for modern humans (Feldesman et al., 1989, 1990), which level FLT1 prints show evidence of a large individual traveling at a does not take into account the relatively short femur length in slow run or fast walk (2.2e2.7 m/s, or 4.9 to 6.0 miles/hour). The A. afarensis (Richmond et al., 2002) or the positively allometric four trails in the upper layer all have short stride lengths, which relationship between femur length and stature in modern humans suggest that these hominins were traveling at very slow walking (Auerbach and Sylvester, 2011). Shorter stature estimates for aus- speeds (0.45e1.0 m/s, or 1.0 to 2.2 miles/hour). These speed esti- tralopithecines would further accentuate the stature differences of mates are robust, regardless of assuming modern human or prim- the Ileret printmakers. itive (A. afarensis-like) foot:hindlimb length proportions, or using Estimated body size for the Ileret hominins may also be direct regressions or dimensionless Froude numbers (Alexander, consistent with the body size of male P. boisei. Existing mass esti- 1984b; Raichlen et al., 2008; Vaughan and Blaszcyzk, 2008). In mates for individual P. boisei specimens range from 32 kg to as large this respect, the upper layer footprints show some similarity to the as 70 kg (Hartwig-Scherer, 1993; Aiello and Wood, 1994; only other hominin footprints known from the early Pleistocene, Kappelman, 1996). Based on specimens with uncertain taxonomic those at the site of GaJi10 in Area 103 at Koobi Fora, Kenya attribution, McHenry and Coffing (2000) estimate P. boisei male (Behrensmeyer and LaPorte, 1981), dating to c. 1.43 Ma (Bennett mean mass of 49 kg and female mean of 34 kg. Given the body mass et al., 2009). The single GaJi10 hominin trail preserves large (c. estimates from orbit sizes (Aiello and Wood, 1994; Kappelman, 26 cm length) footprints with a very short stride length (c. 80 cm). 1996) for likely male individuals of P. boisei, specimens KNM-ER Using our walking speed regression (Table 1), this hominin would 406 (60e70 kg) and OH 5 (40e58 kg), and the high degree of size have been walking at a speed of approximately 0.55 m/s (or 1.2 dimorphism in this taxon (Kappelman, 1996; Silverman et al., miles/hour), falling at the low end of the speed estimates for the 2001), the male mean of 49 kg suggested by McHenry and FwJj14E upper layer trails reported here (Fig. 6). Furthermore, the Coffing (2000) is probably an underestimate. Similarly, stature es- GaJi10 prints are oriented at an angle relative to the direction of timates for P. boisei (male mean ¼ 137 cm; female mean ¼ 124 cm; travel, suggesting to Behrensmeyer and LaPorte (1981:169) a McHenry and Coffing, 2000) are considerably smaller (and prob- “hesitant, somewhat sideways progression across a slippery ably overestimated without taking allometry into account; see surface.” above) than those for H. erectus and the estimates derived here from the Ileret footprints, but little confidence should be placed on Body size earlier estimates for P. boisei given the lack of securely attributed postcranial fossils. Some fossils that differ from those of H. erectus, The new stature and mass estimates for the FwJj14E hominins such as the very large humerus, KNM-ER 739, may belong to reported here provide evidence of large body size at 1.51e1.53 Ma. P. boisei (McHenry, 1978). If so, and if cranial-based body size esti- Because three hominin taxa e H. erectus, H. habilis, and P. boisei e mates are reasonable, then male P. boisei almost certainly reached are present in Okote Member deposits within a few kilometers of larger mass and stature than those typical of earlier gracile the FwJj14E site, it is unclear which species was responsible for australopiths. making the footprints at this site. Of the three known taxa, it is Under either assumption of modern humanlike or more primi- unlikely that the small-bodied H. habilis (McHenry and Coffing, tive foot and body proportions, the estimates based on the FwJj14E 2000) made such large footprints. The stature and mass estimates footprints (Figs. 7 and 8; Tables 5 and 6) are consistent with cranial- reported here are consistent with estimates based on H. erectus based mass estimates for male P. boisei (Aiello and Wood, 1994; fossils (Ruff and Walker, 1993; Aiello and Wood, 1994; Kappelman, Kappelman,1996), suggesting that we cannot rule out the possibility 1996; Ruff et al., 1997; McHenry and Coffing, 2000; Ruff, 2007); that larger P. boisei individuals, presumably males, were responsible

Table 6 Hominin body size comparison.

Species Dates (Ma) Stature (cm) Mass (kg)

Male Female Average Male Female Average Australopithecus afarensis 3.9e2.9 151.0 105.0 128.0 45.0 29.0 37.0 Paranthropus boisei* 2.3e1.2 ee e 63.7 32.0 53.1 Homo habilis 2.3e1.6 131.0 100.0 115.5 37.0 32.0 34.5 Homo erectus (Turkana) 1.9e1.4 180.0 160.0 170.0 66.0 56.0 61.0 (Dmanisi) 1.8 149.0 143.0 146.0 49.0 40.0 44.5 Homo heidelbergensis 0.7e0.2 ee 177.0 ee 67.0 Homo sapiens (Daasanach) Extant 174.0 162.0 168.0 53.0 49.0 51.0 FwJj14E hominins 1.51e1.53 ee169.5 (153.0e185.8) ee50.0 (41.5e60.3)

Stature estimates from: McHenry 1991, McHenry and Coffing, 2000 (A. afarensis, P. boisei, H. habilis); Ruff and Walker 1993 (H. erectus); Carretero et al. 2012 (H. heidelbergensis). Mass estimates from: Kappelman, 1996 (P. boisei); McHenry 1992, McHenry and Coffing 2000 (A. afarensis, H. habilis); Ruff and Walker 1993 (H. erectus); Carretero et al. 2004 (H. heidelbergensis). FwJj14E hominin estimates are based on modern human proportions. * Estimates are based only on cranial remains. 566 H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568 for some or all of these footprints. This is consistent with the current and Coffing (2000). While it should be noted that body size is known hypothesis that the prints were made by H. erectus and/or male to vary ecogeographically (Holliday and Ruff, 1997), estimates of P. boisei (Bennett et al., 2009), with the possible exception of the body size for H. heidelbergensis differ very little from estimates for smallest print, A2-I2. Furthermore, the footprint sizes and estimated H. erectus. There is not another notable change in hominin body size statures for each of the three stratigraphic levels, which are sepa- until the appearance of early modern humans, whose average stat- rated by several thousand years (Fig. 4), are consistently large, ure has been estimated at 177.4 cm (Carretero et al., 2012), followed demonstrating that a large-bodied hominin taxon (or taxa) was by a decrease in modern human body size in the past 50,000 years active at this location in deltaic, river-shore environments (Ruff et al., 1997). (Behrensmeyer, 2011) over an extended period of time. The results here support hypotheses that the shift to hominin occupation of more open, dry environments (Bobe, 2011)was Body size evolution associated with an increase in hominin body size during the early Pleistocene (McHenry and Coffing, 2000; Antón, 2003; Bobe, 2011). The sizes of the footprints at the FwJj14E site near Ileret, Kenya, These results also support the ’grade-level’ difference between provide strong evidence in support of the idea that hominin body early H. erectus compared with Australopithecus and even H. habilis size increased in the early Pleistocene. This site is situated at an (Wood and Collard, 1999), and indicate that hypotheses regarding important period of time in hominin evolution; the age of the site at behavioral and physiological changes in early Homo can reliably Ileret coincides with the hypothesized body size increase in the rest on assumptions of increased mass and stature (McHenry, 1994; genus Homo and, perhaps, Paranthropus. Furthermore, it is McHenry and Coffing, 2000; Aiello and Key, 2002; Aiello and Wells, contemporaneous with the age of KNM-WT 15000 (1.53 Ma). Our 2002; Lieberman et al., 2009; Pontzer, 2012). body size estimates for the Ileret hominins (and the individual at GaJi10) exceed those proposed by Ohman et al. (2002) and Graves Conclusion et al. (2010) for KNM-WT 15000. Regardless of the adult size of this particular individual, both our predicted averages (169.5 cm for This study provides strong support for the hypothesis that stature, 50.0 kg for mass) and ranges (153.0e185.8 cm for stature, hominin body size increased in the early Pleistocene. Footprints at 41.5e60.3 kg for mass) reflect a larger body size for these hominins. FwJj14E from multiple individuals in three different stratigraphic The small body size estimates for the Dmanisi and Gona specimens levels provide evidence of relatively large hominin body mass and indicate that the size range for H. erectus is likely broader than was stature at 1.51e1.53 Ma and, with the exception of one individual previously thought (Lordkipanidze et al., 2007; Simpson et al., 2008; who was running slowly or walking quickly, these hominins were Antón, 2012). The estimated size of the Gona and smaller Dmanisi traveling at slow walking speeds. Fourteen prints from these three individuals are smaller than even the smallest of the Ileret hominins. levels, which likely represent seven individuals, predict body sizes On average, the Ileret hominins are estimated to be only slightly that are comparable with the high end of existing stature and mass larger than the ‘large individual’ from Dmanisi, whose average es- estimates based on H. erectus fossils of this age (Ruff and Walker, timates for stature and mass are 149.3 cm and 48.8 kg, respectively. 1993; Aiello and Wood, 1994; Kappelman, 1996; Ruff et al., 1997; However, the largest individuals from FwJj14E are substantially McHenry and Coffing, 2000; Ruff, 2007). Alternatively, it is possible larger than the ‘large’ Dmanisi individual, suggesting that H. erectus that some or all of the FwJj14E footprints were made by male was characterized by a relatively high degree of size variation. P. boisei, given the large size variation of taxonomically diagnostic Compared with earlier hominins, the early Pleistocene foot- craniodental remains (Silverman et al., 2001) and large postcranial prints presented here, along with some H. erectus and probable fossils potentially belonging to this species. However, regardless of P. boisei fossils, show evidence of substantial size increase. Aus- what taxon or taxa created the footprints at FwJj14E, the stature tralopithecus afarensis and H. habilis have estimated statures of c. and body mass estimates reported here provide evidence of large 100e150 cm and body mass of c. 29e45 kg. In contrast, the seven hominin body size at 1.51e1.53 Ma, comparable with that observed Ileret hominin printmakers at FwJj14E have estimated statures of in a sample of modern humans, and substantially larger than the 170 cm (range: 153e186 cm) and body mass of 50 kg (range: 41.5e typical size of Pliocene hominins. 60.0 kg). Their average body size was well above the mean male sizes of A. afarensis and H. habilis, particularly with regard to stat- Acknowledgments ure. The largest Ileret footprint is comparable in size with the largest male footprint observed in the modern human Daasanach We thank David Green, David Braun, Purity Kiura, Emmanuel sample. The Ileret printmakers’ sizes are larger than the individuals Ndiema, Jack Harris, the Koobi Fora Field School, the National at Dmanisi, but consistent with mean estimates of stature (male: Museums of Kenya, the people of Ileret, and the Daasanach 180 cm, female: 160 cm; Ruff and Walker, 1993) and mass (male: experimental subjects for their contributions to this project. 66 kg, female: 56 kg; Ruff et al., 1997) based on various H. erectus Additionally, we thank David Hunt, the Smithsonian’s National fossils from the Turkana Basin. Museum of Natural History, Will Harcourt-Smith, Eileen Westwig, Later Homo size estimates do not reflect a significant size increase and the American Museum of Natural History for access to skeletal post-H. erectus if the results presented here, Ruff’s(2007)estimates collections. We also thank Bernard Wood and Herman Pontzer for for KNM-WT 15000, Aiello and Wood’s(1994)and Kappelman’s their advice and assistance during the preparation of the manu- (1996) cranial-based estimates, and McHenry and Coffing’s (2000) script, as well as Mark Teaford, the Associate Editor, and two average estimates for H. erectus are reasonably accurate. Wood and anonymous reviewers for their helpful comments. Finally, we thank Collard (1999) estimated Homo heidelbergensis body mass at 62 kg the following funding agencies for supporting this research: Na- based on both African and European specimens. Based on an analysis tional Science Foundation’s Biological Anthropology Program (BCS- of more recently recovered fossil femora, Carretero et al. (2004) have 1128170, -0924476) and Integrative Graduate Education and estimated a mean body mass of 67.1 kg (range: 58.7e81.4 kg), only Research Traineeship Program (DGE-0801634); George Washing- slightly larger than the H. erectus estimates. Stature estimates for ton University’s Selective Excellence Fund, University Facilitating H. heidelbergensis based on the Sima de los Huesos fossils report a Fund, George Gamow Undergraduate Research Fellowship, Of fice of mean stature of 163.6 cm (Carretero et al., 2012), slightly smaller the Vice President for Research Undergraduate Research than the mean stature (170.0 cm) for H. erectus reported by McHenry Fellowship. H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568 567

References Jungers, W.L., 1982. Lucy’s limbs: skeletal allometry and locomotion in Austral- opithecus afarensis (A.L. 288-1). Nature 297, 676e678. Jungers, W.L., 1988a. Relative joint size and hominoid locomotor adaptations with Aiello, L.C., Key, C., 2002. Energetic consequences of being a Homo erectus female. implications for the evolution of hominid bipedalism. J. Hum. Evol. 17, 247e265. Am. J. Hum. Biol. 14, 551e565. Jungers, W.L., 1988b. Lucy’s length: stature reconstruction in Australopithecus Aiello, L.C., Wells, J.C.K., 2002. Energetics and the evolution of the genus Homo. afarensis (AL 288-1) with implications for other small-bodied hominids. Am. J. A. Rev. Anthropol. 31, 323e338. Phys. Anthropol. 76, 227e231. Aiello, L.C., Wood, B.A., 1994. Cranial variables as predictors of hominine body mass. Jungers, W.L., 1988c. New estimates of body size in australopithecines. In: Grine, F.E. Am. J. Phys. Anthropol. 95, 409e426. (Ed.), Evolutionary History of the “Robust” Australopithecines. Aldine de Alexander, R.McN., 1984a. The gaits of bipedal and quadrupedal animals. Int. J. Gruyter, New York, pp. 115e121. Robot. Res. 3, 49e59. Jungers, W.L., Harcourt-Smith, W.E., Wunderlich, R.E., Tocheri, M.W., Larson, S.G., Alexander, R.McN., 1984b. Stride length and speed for adults, children, and fossil Sutikna, T., Due, R.A., Morwood, M.J., 2009. The foot of Homo floresiensis. Nature hominids. Am. J. Phys. Anthropol. 63, 23e27. 459, 81e84. Antón, S.C., 2003. Natural history of Homo erectus. Yearb. Phys. Anthropol. 46, Kappelman, J., 1996. The evolution of body mass and relative brain size in fossil 126e170. hominids. J. Hum. Evol. 30, 243e276. Antón, S.C., 2012. Early Homo: who, when, and where. Curr. Anthropol. 53, S278e Lieberman, D.E., Pilbeam, D.R., Wrangham, R.W., 2009. The transition from Austral- S298. opithecus to Homo. In: Shea, J.J., Lieberman, D.E. (Eds.), Transitions in Prehistory: Ashizawa, K., Kumakura, C., Kusumoto, A., Narasaki, S., 1997. Relative foot size and Essays in Honor of Ofer Bar-Yosef. Oxbow Publications, Oxford, pp. 1e22. shape to general body size in Javanese, Filipinas, and Japanese with special Lieberman, D.E., Venkadesan, M., Werbel, W.A., Daoud, A.I., D’Andrea, S., Davis, I.S., reference to habitual footwear types. Annls. Hum. Biol. 24, 117e129. Mang’eni, R.O., Pitsiladis, Y., 2010. Foot strike patterns and collision forces in Auerbach, B.M., Sylvester, A.D., 2011. Allometry and apparent paradoxes in human habitually barefoot versus shod runners. Nature 463, 531e535. limb proportions: implications for scaling factors. Am. J. Phys. Anthropol. 144, Lockwood, C.A., Richmond, B.G., Jungers, W.L., Kimbel, W.H., 1996. Randomization 382e391. procedures and sexual dimorphism in Australopithecus afarensis. J. Hum. Evol. Barnett, C., 1962. The normal orientation of the human hallux and the effect of 31, 537e548. footwear. J. Anat. 96, 489e494. Lordkipanidze, D., Jashashvili, T., Vekua, A., Ponce de León, M.S., Zollikofer, C.P.E., Behrensmeyer, A.K., 2011. Conversations with Glynn’s ghost: the evolution of Rightmire, G.P., Pontzer, H., Ferring, R., Oms, O., Tappen, M., Bukhsianidze, M., paleolandscape research at East Turkana. In: Sept, J., Pilbeam, D. (Eds.), Casting Agustí, J., Kahlke, R., Kiladze, B., Martinez-Navarro, B., Mouskhelishvili, A., the Net Wide: Papers in Honor of Glynn Isaac and His Approach to Human Nioradze, M., Rook, L., 2007. Postcranial evidence from early Homo from Origins Research. Oxbow Books, Cambridge, pp. 21e40. Dmanisi, Georgia. Nature 449, 305e310. Behrensmeyer, A.K., LaPorte, L.F., 1981. Footprints of a Pleistocene hominid in Lovejoy, C.O., 2005. The natural history of human gait and posture Part 1. Spine and northern Kenya. Nature 289, 167e169. Pelvis. Gait Posture 21, 95e112. Bennett, M.R., Harris, J.W.K., Richmond, B.G., Braun, D.R., Mbua, E., Kiura, P., McHenry, H.M., 1978. Early hominid humerus from east Rudolf, Kenya. Science 180, Olago, D., Kibunjia, M., Omuombo, C., Behrensmeyer, A.K., Huddart, D., 739e741. Gonzalez, S., 2009. Early hominin foot morphology based on 1.5-million-year- McHenry, H.M., 1991. Femoral lengths and stature in Plio-Pleistocene hominins. Am. old footprints from Ileret, Kenya. Science 323, 1197e1201. J. Phys. Anthropol. 86, 445e454. Bobe, R., 2011. Fossil mammals and paleoenvironments in the Omo-Turkana Basin. McHenry, H.M., 1992. Body size and proportions in early hominids. Am. J. Phys. Evol. Anthropol. 20, 254e263. Anthropol. 87, 407e431. Carretero, J.M., Arsuaga, J.L., Martínez, I., Quam, R.M., Lorenzo, C., Gracia, A., McHenry, H.M., 1994. Behavioral ecological implications of early hominid body size. Ortega, A.I., 2004. Los humanos de la Sima de los Huesos (Sierra de Atapuerca) y J. Hum. Evol. 27, 77e87. la evolución del cuerpo en el género Homo. In: Baquedano, E., Rubio Jara, S. McHenry, H.M., Coffing, K., 2000. Australopithecus to Homo: Transformations in (Eds.), Miscelánea en homenaje a Emiliano Aguirre. Paleoantropología, Alcalá body and mind. A. Rev. Anthropol. 29, 125e146. de Henares, vol. III. Museo Arqueológico Regional, Spain, pp. 121e135. Ohman, J.C., Wood, C., Wood, B., Crompton, R., Gunter, M., Yu, L., Savage, R., Carretero, J., Rodríguez, L., García-González, R., Arsuaga, J., Gómez-Olivencia, A., Wang, W., 2002. Stature-at-death of KNM-WT 15000. Hum. Evol. 17, 129e142. Lorenzo, C., Bonmatí, A., Gracia, A., Martínez, I., Quam, R., 2012. Stature esti- Plavcan, J.M., Lockwood, C.A., Kimbel, W.H., Lague, M.R., Harmon, E.H., 2005. Sexual mation from complete long bones in the Middle Pleistocene humans from the dimorphism in Australopithecus afarensis revisited: how strong is the case for a Sima de los Huesos, Sierra de Atapuerca (Spain). J. Hum. Evol. 62, 242e255. human-like pattern of dimorphism? J. Hum. Evol. 48, 313e320. Charteris, J., Wall, J.C., Nottrodt, J.W., 1981. Functional reconstruction of gait from Pontzer, H., 2012. Ecological energetics in early Homo. Curr. Anthropol. 53, S346e the Pliocene hominid footprints at Laetoli, northern Tanzania. Nature 290, S358. 496e498. Pontzer, H., Rolian, C., Rightmire, G.P., Jashashvili, T., Ponce de León, M.S., Coolidge, H.J., Shea, B.T., 1982. External body dimensions of Pan paniscus and Pan Lordkipanidze, D., Zollikofer, C.P.E., 2010. Locomotor anatomy and biomechanics troglodytes chimpanzees. Primates 23, 245e251. of the Dmanisi hominins. J. Hum. Evol. 58, 492e504. D’Aout, K., Pataky, T.C., De Clercq, D., Aerts, P., 2009. The effects of habitual footwear Potts, R., Behrensmeyer, A.K., Deino, A., Ditchfield, P., Clark, J., 2004. Small mid- use: foot shape and function in native barefoot walkers. Footwear Sci. 1, 81e94. Pleistocene hominin associated with East African Acheulean technology. Sci- Feibel, C.S., Brown, F.H., McDougall, I., 1989. Stratigraphic context of fossil hominids ence 305, 75e78. from the Omo group deposits: northern Turkana Basin, Kenya and Ethiopia. Am. Raichlen, D.A., Pontzer, H., Sockol, M.D., 2008. The Laetoli footprints and early J. Phys. Anthropol. 78, 595e622. hominin locomotor kinematics. J. Hum. Evol. 54, 112e117. Feldesman, M.R., Lundy, J.K., Kleckner, J.G., 1989. The femur-stature ratio and esti- Rasband, W.S., 1997e2012. ImageJ. U. S. National Institutes of Health, Bethesda, mates of stature in mid- to late-Pleistocene fossil hominids. Am. J. Phys. Maryland, USA. http://imagej.nih.gov/ij/. Anthropol. 78, 219e220. Reno, P.L., Meindl, R.S., McCollum, M.A., Lovejoy, C.O., 2003. Sexual dimorphism in Feldesman, M.R., Kleckner, J.G., Lundy, J.K., 1990. The femur/stature ratio and esti- Australopithecus afarensis was similar to that of modern humans. Proc. Natl. mates of stature in mid- and late-Pleistocene fossil hominids. Am. J. Phys. Acad. Sci. 100, 9404e9409. Anthropol. 83, 359e372. Reno, P.L., McCollum, M.A., Meindl, R.S., Lovejoy, C.O., 2010. An enlarged postcranial Gordon, A.D., 2004. Evolution of body size and sexual size dimorphism in the Order sample confirms Australopithecus afarensis dimorphism was similar to modern Primates: Rensch’s rule, quantitative genetics, and phylogenetic effects. Ph.D. humans. Phil. Trans. R. Soc. B 365, 3355e3363. Dissertation, University of Texas, Austin. Richmond, B.G., Jungers, W.L., 1995. Size variation and sexual dimorphism in Aus- Gordon, A.D., Green, D.J., Richmond, B.G., 2008. Strong postcranial size dimorphism tralopithecus afarensis and living hominoids. J. Hum. Evol. 29, 229e245. in Australopithecus afarensis: results from two new resampling methods for Richmond, B.G., Aiello, L.C., Wood, B.A., 2002. Early hominin limb proportions. multivariate data sets with missing data. Am. J. Phys. Anthropol. 135, 311e328. J. Hum. Evol. 43, 529e548. Graves, R.R., Lupo, A.C., McCarthy, R.C., Wescott, D.J., Cunningham, D.L., 2010. Just Richmond, B.G., Bennett, M.R., Harris, J.W.K., Behrensmeyer, A.K., Braun, D.R., how strapping was KNM-WT 15000? J. Hum. Evol. 59, 542e554. Carnation, S., Chirchir, H., Green, D.J., Kiura, P., Mbua, E., 2010. The anatomy of Grine, F.E., Jungers, W.L., Tobias, P.V., Pearson, O.M., 1995. Fossil Homo femur from footprints from Koobi Fora, Kenya. Am. J. Phys. Anthropol. S50, 197. Berg Aukas, northern Namibia. Am. J. Phys. Anthropol. 97, 151e185. Ruff, C., 2002. Variation in human body size and shape. A. Rev. Anthropol. 31, Haeusler, M., Schiess, R., Boeni, T., 2011. New vertebral and rib material point to 211e232. modern bauplan of the Nariokotome Homo erectus skeleton. J. Hum. Evol. 61, Ruff, C.B., 2003. Long bone articular and diaphyseal structure in Old World monkeys 575e582. and apes. II: estimation of body mass. Am. J. Phys. Anthropol. 120, 16e37. Hartwig-Scherer, S., 1993. Body weight prediction in early fossil hominids: towards Ruff, C., 2007. Body size prediction from juvenile skeletal remains. Am. J. Phys. a taxon-“independent” approach. Am. J. Phys. Anthropol. 92, 17e36. Anthropol. 133, 698e716. Hatala, K.G., Dingwall, H.L., Wunderlich, R.E., Richmond, B.G., 2013. Variation in foot Ruff, C., Walker, A., 1993. Body size and body shape. In: Walker, A., Leakey, R. (Eds.), strike patterns during running among habitually barefoot populations. PLoS The Nariokotome Homo erectus Skeleton. Harvard University Press, Cambridge, One 8, e52548. pp. 234e265. Hoffman, P., 1905. Conclusions drawn from a comparative study of the feet of Ruff, C.B., Trinkaus, E., Holliday, T.W., 1997. Body mass and encephalization in barefooted and shoe-wearing peoples. Am. J. Orthoped. Surg. 3, 105e137. Pleistocene Homo. Nature 387, 173e176. Holliday, T.W., Ruff, C.B., 1997. Ecogeographical patterning and stature prediction in Scott, J.E., Stroik, L.K., 2006. Bootstrap tests of significance and the case for hu- fossil hominids: comment on Feldesman and Fountain (1996). Am. J. Phys. manlike skeletal-size dimorphism in Australopithecus afarensis. J. Hum. Evol. 51, Anthropol. 103, 137e140. 422e428. 568 H.L. Dingwall et al. / Journal of Human Evolution 64 (2013) 556e568

Silverman, N.J., Richmond, B.G., Wood, B.A., 2001. Testing the taxonomic integrity of Trinkaus, E., Shang, H., 2008. Anatomical evidence for the antiquity of human Paranthropus boisei sensu stricto. Am. J. Phys. Anthropol. 115, 167e178. footwear: Tianyuan and Sunghir. J. Archaeol. Sci. 35, 1928e1933. Sim-Fook, L., Hodgson, A.R., 1958. A comparison of foot forms among the non- Vaughan, C.L., Blaszcyzk, M.B., 2008. Dynamic similarity predicts gait parameters shoe and shoe-wearing Chinese populations. J. Bone Joint Surg. 40, 1058e for Homo ﬂoresiensis and the Laetoli hominins. Am. J. Hum. Biol. 20, 312e316. 1062. Wells, L.H., 1931. The foot of the South African native. Am. J. Phys. Anthropol. 15, Simpson, S.W., Quade, J., Levin, N.E., Butler, R., Dupont-Nivet, G., Everett, M., 185e289. Semaw, S., 2008. A female Homo erectus pelvis from Gona, Ethiopia. Science White, T., Suwa, G., 1987. Hominid footprints at Laetoli: facts and interpretations. 322, 1089e1092. Am. J. Phys. Anthropol. 72, 485e514. Spoor, F., Leakey, M.G., Gathogo, P.N., Brown, F.H., Antón, S.C., McDougall, I., Wood, B., Collard, M., 1999. The human genus. Science 284, 65e71. Kiarie, C., Manthi, F.K., Leakey, L.N., 2007. Implications of new early Homo fossils Wood, B.A., Richmond, B.G., 2000. Human evolution: taxonomy and paleobiology. from Ileret, east of Lake Turkana, Kenya. Nature 448, 688e691. J. Anat. 196, 19e60.