Reconstruction and Comparative Analysis of a Large Late Pleistocene Modern Human Femur From Wajak, Indonesia

ABSTRACT

Objectives

A largely undescribed Late Pleistocene Homo sapiens femur from Wajak, Java, Indonesia (minimum age, 36.5 ka) is reconstructed. In conjunction with data derived from other fossil specimens, we test whether ecogeographic rules relating body size and shape to climate apply to late Middle to Late Pleistocene East Asian Homo.

Materials and Methods

Preserved portions of the Wajak femur were micro‐CT scanned and used to reconstruct the whole bone using a sample of modern human femora as a template. Body mass and stature were reconstructed from femoral head breadth and length. Similar data were derived from eight penecontemporaneous (40–28 ka) and five earlier (ca 260–112 ka) East Asian femora or tibiae, spanning a wide latitudinal range. Anthropometric data from 74 living Old World populations were also included in comparisons.

Results

Higher latitude individuals are heavier and relatively wider‐bodied than lower latitude individuals in both fossil groups and among modern populations. There is also a temporal decline in body mass and relative body breadth in both latitudinal zones. Stature shows much weaker associations with both climate and time period. The Wajak femur is long (468 mm) and likely male, with an estimated stature of 170.5 cm but a relatively linear body shape, consistent with these patterns.

Discussion

Ecogeographic rules that relate surface area/body mass to temperature appear to apply to East Asian late Middle to Late Pleistocene Homo as well as modern humans. The temporal decline in body mass and breadth may be due to technological changes during the Late Pleistocene.

Both climate (latitude) and time (earlier: black; later: gray) affect body size and shape among late Middle to later Late Pleistocene Homo in East Asia.

1. Introduction

Variation in body size and shape within the genus Homo is considerable, among both living and past populations (Hruschka et al. 2015; Roberts 1978; Ruff 1994). This has in part been attributed to ecogeographic trends in body form widely observed among endotherms, the most well‐known of which are Bergmann's and Allen's “Rules.” Bergmann's (1847) Rule states that within a geographically widely‐dispersed species, populations in colder regions will be larger in body size than those in warmer regions, while Allen's (1877) Rule states that under the same conditions, populations in colder regions will have relatively shorter extremities than those in warmer regions. The proposed physiological basis for these rules is the same: increases in overall body size and decreases in extremity (usually limb) length both decrease surface area to body mass and thus help to conserve heat, while the opposite is true for smaller bodies and relatively longer extremities (Mayr 1956; Roberts 1978; Schreider 1964). While some aspects of the rules have been critiqued over the years, their general validity has been repeatedly demonstrated in more recent large‐scale studies of mammals and birds (e.g., Ashton et al. 2000; Meiri and Dayan 2003). The same is true for modern humans (Foster and Collard 2013; Roberts 1978; Ruff 1994), even when factors such as population structure (genetic drift, gene flow) and nutrition are controlled for (Hruschka et al. 2015; Pomeroy et al. 2021; Roseman and Auerbach 2015). Furthermore, the same ecogeographic trends are apparent among fossil Homo (Holliday 1997a, 1997b; Ruff 1994; Trinkaus 1981; Weaver 2009; Will et al. 2021), including temporal changes in body form linked to past climatic changes (Stibel 2023), as also demonstrated in other mammals (Millien et al. 2006). However, factors other than climatic adaptation may also contribute to geographic clines in body form (e.g., Bogin et al. 2022; Walker et al. 2011; Weaver and Steudel‐Numbers 2005).

General temporal trends in body size among fossil hominins have also been extensively documented in recent studies. While results of investigations differ in some details, in part due to methodological issues (Ruff et al. 2018), most studies agree that there was an overall increase in body mass among early Homo relative to earlier and contemporaneous australopiths (with the exception of the diminutive H. naledi and H. floresiensis), and that body mass in H. sapiens has decreased on average over the past 30,000 years (Holliday 2012; Ruff et al. 2018; Ruff and Wood 2023; Stibel 2023; Will et al. 2017). This recent decrease in average body mass is apparent in several broad regions of the world, both tropical and higher latitude (Ruff et al. 2022).

In this paper, building on previous studies of the East Asian Late Pleistocene fossil record (Curnoe et al. 2021; Ruff et al. 2015, 2022), we report the reconstruction of a large H. sapiens femur from the site of Wajak, Java, Indonesia (minimum age 36.5 ± 3.2 ka [thousands of years]) (Storm et al. 2013), which has previously existed only as fragments (Dubois 1922; Storm 1995). From this reconstruction we estimate both the body mass and stature, and body mass index (BMI: body mass/stature²), a measure of body shape, of this individual. We carry out comparisons with eight penecontemporaneous (40–28 ka) modern human individuals from East Asia, spanning a geographic range from the tropics to higher latitudes (40° N). We also compare this sample to five East Asian Homo sp. individuals dating to the early Late Pleistocene (ca. 115–112 ka) or late Middle Pleistocene (260 ka), also spanning a wide latitudinal range. To permit inclusion of three tibial specimens from Ngandong and Sambungmacan, a new technique is developed for estimating body mass from tibial diaphyseal cross‐sectional size. Using results of these comparisons, we assess whether ecogeographic and temporal trends in body size and shape among East Asian hominins from this time range conform to expectations based on wider comparisons among fossil and living humans, and discuss implications for interpreting the Late Pleistocene East Asian fossil record.

2. Materials and Methods

2.1. Wajak Postcrania

2.1.1. Specimens

The Wajak (or Wadjak) site was first discovered in 1888 in a limestone quarry in Java, Indonesia (Figure 1), with additional excavations carried out in 1890 by Dubois (Dubois 1922; Storm 1995; Theunissen et al. 1990). The site is most famous for two partial skulls, Wajak 1 and 2, which have figured prominently in discussions of the population history of Southeast Asia and Australia (e.g., Dubois 1922; Jacob 1967; Storm 1995; Storm and Nelson 1992; Storm et al. 2013; Wolpoff et al. 1984). Both specimens are robust representatives of early anatomically modern humans (Storm et al. 2013). Human postcranial remains were also found at the site, but aside from a brief mention by Dubois (1922) went undescribed for more than 100 years, when a complete inventory was carried out (Storm 1995). Direct association of the postcrania with either of the skulls is uncertain, although later U‐series dating of Wajak 2 and a femoral specimen produced overlapping age estimates (Storm et al. 2013) (also see Discussion). The average age of the femoral specimen (W‐H‐29), included in the present analyses, was 36.5 ± 3.2 ka, which should be viewed as a minimum age estimate (Storm et al. 2013). All specimens were studied at Naturalis Biodiversity Center (Leiden, The Netherlands) as part of the Dubois Collection. As of July, 2026, the collection is being transported to Indonesia and is expected to be housed at Museum Nasional (Jakarta, Indonesia) from August, 2026 onwards.

FIGURE 1.

Maps showing location of the Wajak site (red star) in Java, Indonesia. Modified from Setiyabudi et al. (2020), with permission of the authors.

Four pieces of a left femur were among the material recovered from the site, including a proximal end (W‐H‐27), a diaphyseal section near midshaft (W‐H‐29), a more distal shaft fragment (W‐H‐31), and a shaft fragment just proximal to the condylar region (W‐H‐28) (Figure 2). All are very similar in color and apparent fossilization and are filled with a dark brown matrix, when visible (see Figure 3). As shown below, although there are no direct adjoining surfaces, they fit well together within a single left femoral template. In terms of size, apparent fossilization, and coloration, there are no other femoral specimens from the site that match these (see Text S1 and Figure S1). Some gray plasticine was added to the distal surface of W‐H‐29 and the proximal surface of W‐H‐31 at some point in the past, holding these fragments together (Figure 2); the two fragments have since been separated. A piece was removed from the proximal end of W‐H‐29 for chemical and dating analyses (Storm et al. 2013), leaving a smooth transverse surface that was used here for cross‐sectional analyses (Figure 3; also see below).

FIGURE 2.

Wajak femoral specimens. Top row: Anterior view; bottom row: Posterior view. Proximal is above and distal below in all specimens; medial is to the left in anterior view, to the right in posterior view. Gray material visible in W‐H‐29 and 31 is plasticine added at some point in the past.

FIGURE 3.

Cross section from cut proximal surface of W‐H‐29. From left to right: Original physical section; physical section with displaced pieces anatomically realigned; binary image used in cross‐sectional analyses, with matrix and trabecular bone removed and cracks filled.

All of the fragments are relatively well‐preserved, although there is some cracking and minor distortion (Figure 2). The femoral head on W‐H‐27 is pitted and is missing much of the posterior aspect. An original femoral head superoinferior (S‐I) breadth of about 45.5 mm is estimated from the remaining surfaces; this is also consistent with the best fit of a template to the specimen (see below). The neck is broken near its midpoint and is joined together by a tight, glued junction. The superior aspect of the neck is damaged but the greater trochanter is well preserved. The lesser trochanter is largely worn away but the outline of its base is still clear. Shaft breadths 1.5 cm distal to the distal edge of the lesser trochanter, at approximately 80% of biomechanical length from the distal end (see below), are 32.6 mm mediolaterally (M‐L) and 27.4 mm anteroposteriorly (A‐P). The total length of the specimen, measured parallel to the shaft long axis, is 134 mm.

W‐H‐29 measures 83 mm in length and has several longitudinal cracks that have somewhat displaced the cortex in places, although the periosteal surface is generally well‐preserved. This is illustrated in Figure 3, which shows the original cut proximal surface and the restored surface, used in cross‐sectional analyses. A clear linea aspera is visible proximally (Figure 3) but is worn away distally.

W‐H‐31 is a short segment, 53 mm in length. A large chunk of the distal cortex is missing anterolaterally and the morphology of the proximal end is obscured by the added plasticine. W‐H‐28 measures 63 mm in length and includes part of the popliteal surface posteriorly and mediolateral widening and cortical thinning distally, indicating that it was positioned just proximal to the formation of the condyles.

Other postcranial specimens recovered from Wajak are described and illustrated in Text S1 and Figure S1. They include another fragmentary left proximal femur, well‐preserved right and left proximal tibiae, a well‐preserved right talus, two femoral shaft fragments, the distal half of a left humerus (badly preserved at its distal end), portions of the shaft and distal end of a left radius, a small piece of a distal tibial shaft, and fragments of a fibula(e) and other foot bones. With the exception of the two proximal tibiae and talus, none of these specimens are as heavily fossilized as the four specimens described above. The humerus also appears to be from a much smaller individual.

2.1.2. Femoral Reconstruction

All four pieces of the main study femur (W‐H‐27, 28, 29, 31) were micro‐CT scanned at the Naturalis Biodiversity Center (Leiden, The Netherlands) using a NeoScan N80 scanner. NeoScan acquisitions used 360° rotation at 0.2° steps (≈1800 projections), 6‐frame averaging, and a 2800 × 2400 px detector. The tube was set to 110 kV/146 μA with a 1.0 mm Cu filter. Isotropic voxel sizes were 25.0–37.7 μm.

CT scan data of the Wajak specimens were input into Avizo Lite 2020.3.1 to create fossil femoral surface models. A combination of thresholding and manual voxel selection was used to remove adherent external matrix and repair minor defects. Portions of the resulting surface models not representing the external bone surface were removed. CT scan data from a sample of 40 female and 40 male adult human femora from the Texas State University Donated Skeletal Collection, derived from 20th century Euro‐Americans, African Americans, and Hispanic Americans, were used to generate modern femoral surface models, as described previously (Sylvester et al. 2024). Surface models were then resampled to ~25,000 vertices to generate near‐equilateral triangles of uniform size using the GIBBON package for Matlab (Moerman 2018).

Variation in femoral shape was quantified using a 3D landmark analysis. A single femur from the reference sample was selected to serve as an initial template, which was downsampled to 5170 vertices. These vertices were more densely spaced on the proximal one‐third of the bone to aid in the reconstruction process described below. The template was scaled by length to match each sample femur and then rigidly aligned using the iterative closest point algorithm (Besl and McKay 1992). The template vertices were transferred to the surface of each femur using the Coherent Point Drift (CPD) algorithm as described in Sylvester et al. (2024). This approach is based on the Gaussian mixture model (soft clustering technique) and the expectation–maximization algorithm. Landmark displacement was controlled by λ and local deformation by β; parameters were set to λ = 1 and β staged at 100, 50, 30, 10, and 8. Deformed templates were treated as landmark configurations, and Generalized Procrustes Analysis (Schönemann 1966) was applied to remove the effects of size, location, and orientation. From these analyses, an average femur was generated and returned to natural size based on the average centroid size of the sample.

The Wajak proximal femur specimen (W‐H‐27) was used in conjunction with the average modern femur model to estimate femur length and the relative positions of other fragments (W‐H‐28, 29, 31). The average femur from the shape analysis was surface‐fitted to W‐H‐27 using a modified iterative closest point (mICP) algorithm. Point (vertex) correspondences were established by finding the nearest vertex on the average femur from each vertex on W‐H‐27, thus establishing fossil‐to‐average correspondences. This was followed by a Procrustes transformation (translation, rotation, scale) to best fit the average femur to the fossil fragment. This process continued until the root‐mean‐squared (RMS) error failed to decrease further. The other fragments were aligned to the transformed average femur manually based on available morphology.

This approach inherently assumes that the Wajak femur morphology falls within the range of that of modern human femora. As noted above, the modern human reference sample for these analyses was ethnically diverse, including Euro‐Americans, African Americans, and Hispanic Americans, so incorporated variation in body proportions associated with these groups (Martorell et al. 1988). Although the two Wajak skulls are relatively robust, there is no feature in either the skulls or the femur that distinguishes them from anatomically modern H. sapiens (Storm 1995; Storm et al. 2013; present study). Thus, it seems reasonable to presume that the modern human sample provides an appropriate group for reconstructing the length of the Wajak femur, although this must be acknowledged as a potential source of uncertainty. At present, there is no complete femur from the same general region (southeast Asia) and time period (40–28 ka) as Wajak that could serve as an effective alternative template.

2.2. Comparative Samples

2.2.1. East Asian Fossil Specimens

The 13 late Middle to Late Pleistocene East Asian fossil specimens included in the study are listed, along with the Wajak femur, in Table 1. Locations of fossil sites are shown in Figure 4. All lie east of 90° E. longitude and range from 8° S. to 41° N. latitude. Sites located north of 30° N. latitude are considered “high latitude,” and those located south of this “low latitude.” This boundary corresponds roughly to the division between the Palearctic and Oriental biogeographic zones of Asia, with the former characterized by cold winters and generally less rainfall, and the latter by milder winters and more rainfall (Dennell et al. 2020). In fact, all of the high latitude sites in the present study are located above 39° N. latitude, and most of the low latitude sites are within 8° of the equator, the exception being Liujiang (24° N.). They range in age from 260 to 28 ka, and for the present purposes are divided into two temporal groups: late Middle Pleistocene/early Late Pleistocene (LMP/ELP), 260–112 ka, and later Late Pleistocene (LLP), 40–28 ka. The first group includes specimens variously assigned to late Homo erectus or transitional between late H. erectus and H. sapiens , while the second group has been assigned to H. sapiens (Curnoe et al. 2016; Harvati 2009; Rizal et al. 2020; Ruff et al. 2022; Shang and Trinkaus 2010; also see Section 4). Because of the extreme effects of island endemism on body form in H. floresiensis and H. luzonensis (Brown et al. 2004; Détroit et al. 2019; Jungers et al. 2009; Mijares et al. 2010), representatives of these species are not included here, despite temporal and general geographic overlap with the present study samples.

TABLE 1.

Comparative fossil specimens.

Specimen	Sex	Element(s)	Date (ka) a	Period b	Latitude c	Body mass (kg)	Body mass method d	Bone length (mm) e	Stature (cm) f	BMI g	Sources h
Jinniushan 1	F	Pelvis, ulna	260	LMP/ELP	High	75.8	FHSI i , Bi‐il/stat	242 (radius) j	164.7	27.9	3/1/1
Trinil Femur 1		Femur	≈115	LMP/ELP	Low	62.6	FHSI	455	166.5	22.6	2/2/2
Sambungmacan 2		Tibia	≈112	LMP/ELP	Low	68.6	Tibial CA				6/1/−
Ngandong Tibia 13 (A)		Tibia	112	LMP/ELP	Low	76.4	Tibial CA	443	181.3	23.2	6/1/5
Ngandong Tibia 14 (B)		Tibia	112	LMP/ELP	Low	64.2	Tibial CA	377	163.5	24.0	6/1/4
Tianyuan 1	M	Femur	40	LLP	High	73.4	HDML	463	168.9	25.7	7/2/7
Wajak femur	M	Femur	36.5	LLP	Low	64.2	FHSI	468	170.5	22.1	8/1/1
Zhoukoudian UC 67		Femur	36	LLP	high	67.5	Femoral CA				9/2/−
Zhoukoudian UC 68		Femur	36	LLP	High	60.0	Femoral CA				9/2/−
Zhoukoudian UC 105		Femur	36	LLP	High	69.9	FHSI				9/2/−
Zhoukoudian UC 117		Femur	36	LLP	High	58.6	FHSI				9/2/−
Trinil 9	M	Femur	34.5	LLP	Low	55.4	Femoral CA/len	417	157.2	22.4	2/2/2
Deep Skull 1	F	Femur	34.5	LLP	Low	44.6	TPML	390	147.3	20.6	10/2/10
Liujiang 1	M	Femur	28	LLP	Low	50.9	FHSI	413	156.2	20.9	11/2/12

Note: Wajak specimen highlighted in bold.

^a Dates in thousands of years before present, corrected. Midpoint of range, if given. See Text S2 for more details.

^b LMP/ELP: late Middle Pleistocene/early Late Pleistocene; LLP: late Late Pleistocene.

^c High = > 30° latitude; low = ≤ 30° latitude.

^d FHSI: femoral head superoinferior breadth; Bi‐il/Stat: bi‐iliac breadth and stature; tibial CA: tibial 50% location cortical area; HDML: distal humeral mediolateral articular breadth; femoral CA: femoral 50% location cortical area; femoral CA/Len: femoral 50% location cortical area and maximum length. All formulae given in Ruff and Wood 2023 (world‐wide equations) except tibial CA—see text. Individual dimensions given in Ruff et al. 2022, except tibial CA (Trinkaus and Ruff 2012) and Wajak FHSI (present study). See Text S2 for additional information.

^e Maximum length of long bone elements noted.

^f See main text and Text S2 and Table S1 for stature derivations from bone lengths.

^g Body mass index: body mass (kg)/stature (m)².

^h Sources for Date/Body mass/Bone length: (1) present study; (2) Ruff et al. 2022; (3) Rosenberg et al. 2006; (4) Trinkaus and Ruff 2012; (5) Nelson, pers. comm.; (6) Rizal et al. 2020; (7) Shang and Trinkaus 2010; (8) Storm et al. 2013; (9) Li et al. 2018; (10) Curnoe et al. 2019; (11) Ge et al. 2024; (12) Wei et al. 2023.

ⁱ Femoral head breadth derived from acetabular breadth—see Text S2.

^j Radius length derived from ulna length of 260 mm—see text.

FIGURE 4.

Map of East Asia showing locations of fossil sites. See Table 1 for individual specimens (Deep Skull 1 is from Niah Cave). (Map courtesy of Mapswire (mapswire.com/maps/asia/)).

All of the comparative specimens preserve skeletal elements from which body mass can be estimated; specific estimation methods employed are listed in Table 1 and are further discussed below. Most of the specimens also include a long bone length from which stature can be estimated, as described in Section 2.4. Sex assignments are given where these have been established based on morphological indicators—pelvic morphology (Liujiang 1: Rosenberg 2002; Jinniushan 1: Rosenberg et al. 2006), cranial morphology (Deep Skull 1 (Niah Caves): Curnoe et al. 2021), or femoral cross‐sectional shape (Trinil 9: Ruff et al. 2022). The Tianyuan 1 individual was not assigned a sex by its original describers (Shang and Trinkaus 2010), but is assigned as a probable male here, based on its femoral cross‐sectional shape (even accounting for some pathological distortion: Ruff et al. 2022). Because there are so few specimens with assigned sexes, sex‐specific comparisons are not carried out here; however, sex assignments were considered, when possible, in choosing stature reconstruction formulae (Section 2.4).

2.2.2. Living Samples

In order to provide a broad comparative context for evaluating geographic variation in body size and shape, body mass, stature, and BMI for sex‐specific samples from 74 living Old World populations were obtained from the literature. The samples and sources are listed in Table S1. They are similar to those used in a previous analysis (Ruff et al. 2022), with the addition of several samples from South and West Asia, North Africa, Australia, the Pacific, and Arctic (the Arctic samples are Inupiats and Aleuts, who are considered Old World for the purposes of these comparisons, due to recent shared ancestry with northeastern Asian Native populations (Flegontov et al. 2019)). For comparisons with the fossil specimens, they were also placed into high (> 30° N. latitude) and low latitude groups. Most of these data were collected in the 1960's and 1970's, prior to the general increase in relative body mass characteristic of more recent populations (Katzmarzyk and Leonard 1998; Leonard and Katzmarzyk 2010; Ng et al. 2014). Data for two of the “living” samples—the Aeta and Andamanese—were derived from very recent skeletal material, as described previously (Ruff et al. 2022). To avoid increases in body mass that normally occur with age among adults, data for younger adults (under 40 years) were preferentially selected if available. For other notes on the samples, see Ruff et al. (2022).

2.2.3. Cross‐Sectional Sample

For comparative analyses of the cross‐sectional diaphyseal properties of the Wajak femur, properties for a sample of 56 Late Pleistocene anatomically modern humans (36.0–14.2 ka) of East Asia and Europe were obtained from the literature, along with bone lengths and articular dimensions used to estimate body mass (Ruff 2018b; Shackelford and Demeter 2012; Shackelford 2007; Shackelford; pers. comm.; Trinkaus and Ruff 2012). Note that Tianyuan 1 was not included in these analyses, due to pathological distortion of the diaphysis (Shang and Trinkaus 2010; Wei et al. 2016). Body mass was estimated using the same methods employed here (see below), except that no body masses were estimated using femoral cortical areas to avoid potential circularity when carrying out comparisons of cross‐sectional properties. All comparative data are given in Table S2.

2.3. Body Mass Estimation

Body masses of most of the fossil specimens in Table 1 were estimated from preserved skeletal dimensions using previously published equations based mainly on lower limb articular size or midshaft femoral cortical area, the latter combined with femoral length when possible (Ruff et al. 2022; Ruff and Wood 2023). The methods employed for each fossil are given in Table 1, with more details on individual fossils given in Text S2. All equations used in this study are shown in Table S3. In one case (Tianyuan 1), due to probable distortion of relevant lower limb dimensions, distal humeral articular width was used to estimate body mass (see Ruff et al. 2022 for explanation). Because pelvic material was available for Jinnuishan 1, both femoral head breadth (estimated from acetabular breadth) and a method combining bi‐iliac (maximum pelvic) breadth and estimated stature were used, taking an average of the two estimates. Due to the availability of new stature and body mass estimation equations, both the body mass and stature estimates for this specimen vary slightly from those published previously (Rosenberg et al. 2006) (see Text S2 for details). The femoral head breadth of the Wajak femur is estimated here as 45.5 mm, slightly different from an earlier estimate (46.0 mm; Ruff et al. 2022), resulting in a slightly lower body mass estimate of 64.2 kg (vs. 65.4 kg; Ruff et al. 2022) (see Text S2). Percent standard errors of estimate range between 7.4% and 10.5% for all body mass estimation equations (Table S3).

For three tibial specimens—Sambungmacan 2, Ngandong 13 and Ngandong 14—no articular breadth dimension was available. Following previously presented arguments (Ruff et al. 2022), here we develop a new body mass estimation equation based on tibial cross‐sectional cortical area measured at 50% of biomechanical length (or length, measured from the midpoint of the distal articular surface to the average proximal position of the midpoints of the tibial plateaus, parallel to the shaft longitudinal axis) (Ruff 2002). As in the previous study of the femur (Ruff et al. 2022), to avoid possible bias due to systematic changes in cross‐sectional size to body mass allometric relationships between earlier and recent (Holocene) Homo, only data obtained from Pleistocene Homo specimens were used in the analysis (Trinkaus and Ruff 2012). These included 47 individuals with body masses derived from femoral head breadth (n = 42), tibial plateau breadth (n = 2), distal humeral articular breadth (n = 2), or femoral cortical area and bone length (n = 1). The resulting equation is:

$$\begin{array}{cc}\text{Body mass}=\hfill & 0.0676\times \text{Cortical area}+40.0\hfill \\ \hfill & \left(r=0.678,\%\text{standard error of estimate}=9.8\%\right)\hfill \end{array}$$ (1)

Note that this formula, like previously developed formulae utilizing femoral cortical area (Ruff et al. 2022), are intended for use with Pleistocene Homo specimens only (see Ruff and Wood 2023 for further discussion). Unlike the femoral cortical area formulae, which also included an alternative multiple regression equation including femoral bone length (Ruff et al. 2022), no equation combining tibial cortical area and tibial bone length was employed here. Because of large systematic variation in intralimb length proportions and tibial length to stature proportions among modern and Late Pleistocene Homo (Holliday and Ruff 2001; Ruff and Walker 1993; Trinkaus 1981), sample composition (e.g., whether or not Neandertals are included) has a relatively large effect on equations that include tibial length, while effects are smaller for equations including femoral length. To avoid presupposing any particular linear proportion, then, we did not include tibial length in body mass estimations. Prediction error for the tibial cortical area equation above is similar to that for femoral cortical area alone (%SEE = 10.9%) and femoral cortical area and femoral length (%SEE = 9.3%) (Table S3).

2.4. Stature Estimation

Four different modern reference samples, representing a variety of body sizes and shapes, were used for stature estimation: (1) a Holocene European skeletal sample (Ruff et al. 2012), (2) a Thai skeletal sample (Gocha et al. 2013); (3) a Thai cadaveric sample (Mahakkanukrauh et al. 2011); and (4) a South African skeletal sample (Feldesman and Lundy 1988). Statures in the skeletal samples were derived through anatomical reconstruction (Fully 1956; Raxter et al. 2006). All studies carried out regressions of stature on femoral and tibial lengths—Reduced Major Axis in the European sample, and Least Squares in the others. When fossil specimens could be assigned to sex, formulae specific to that sex were used; otherwise, both male and female formulae were applied. However, to avoid extrapolation beyond the reference sample size range, only samples whose bone length range incorporated that of the fossil specimen were included for that specimen. Using these criteria, between two and 10 formulae were applicable to each fossil specimen, with the average value taken as the best estimate. Equations and individual stature estimations using each equation are given in Tables S4 and S5, respectively, with further information given in Text S2. For Jinniushan 1, stature was estimated from radial maximum length using a female Holocene European formula (Ruff et al. 2012), with radial length estimated from its preserved maximum ulnar length of 260 mm (Rosenberg et al. 2006) using a formula derived from a combined skeletal sample of East African and Pecos Pueblo individuals (Ruff 1995): radial length = 0.968 × ulnar length—9.9 (mm) (n = 99; r = 0.988; %SEE = 1.3%). A higher latitude sample formula (only) was used for Jinnuishan because it has been shown that she had body proportions characteristic of high latitude modern populations (Rosenberg et al. 2006). Similarly, a South African tibia formula (Feldesman and Lundy 1988) was not applied to Tianyuan 1 because the average crural (tibia/femur length) index of the African sample was much higher than that of the fossil, suggesting a different tibia length/stature proportion (other samples had crural indices more similar to that of Tianyuan 1). It should be noted that the new stature estimate for Trinil 9 differs slightly (by about 1 cm) from the previous estimate (Ruff et al. 2022), with a consequent very small change in BMI (from 22.8 to 22.4). New stature estimates for Jinnuishan 1 and Tianyuan 1 vary by 3–4 cm from previously published estimates (Rosenberg et al. 2006; Shang and Trinkaus 2010). More details can be found in Text S2.

2.5. Cross‐Sectional Properties

Cross‐sectional properties—areas, second moments of area, and section moduli—were derived for the Wajak cross section (Figure 3) using the MomentMacro routine (Ruff, n.d.). These properties summarize the geometric components of axial, bending, and torsional rigidity and strength of the femoral diaphysis (Ruff 2018a).

2.6. Statistical Analyses

Boxplots and scatterplots are used to visualize data distributions and relationships between study variables. Because sample sizes for the fossils are small (Table 1), non‐parametric permutation‐based 2‐way ANOVA was used to test for the effects of latitude and temporal period on body mass, stature, and BMI in the fossil samples, using the ImPerm package, aovp routine in R (R Core Team 2025). Permutation‐based methods are particularly appropriate when sample sizes are small and sample distributions may not conform to traditional parametric requirements; they are also more powerful than other traditional nonparametric methods (e.g., using ranked data) (Collingridge 2013). As noted earlier, because sex could not be determined a priori for most of the fossil specimens, sexes were pooled in these analyses. Parametric t‐tests were used to compare latitudinal samples among living humans. Probabilities of ≤ 0.05 were considered statistically significant. Graphing and parametric tests were carried out using SYSTAT (SYSTAT13 2009).

3. Results

3.1. Reconstructed Wajak Femur

The Wajak femur reconstruction is shown in Figure 5. All four preserved fragments fit comfortably within the modern femoral template. Using the average femur model generated from the modern sample of 80 femora, the best estimate of the total maximum reconstructed length of the Wajak femur is 468 mm. An estimate of the error associated with this approach was generated by fitting each of the 80 femora separately and comparing results to that using the average femur model. The mean length estimate in this case is 470 mm, slightly longer than our original estimate. The 95% confidence interval on this mean estimate is ±5 mm. We prefer the more conservative 468 mm estimate, which also falls within this range. Even so, this is the longest femur in the East Asian Late Pleistocene fossil record, slightly (5 mm) longer than Tianyuan 1 (Table 1, Figure 6) (the East Asian sample in Figure 6 includes additional individuals to those listed in Table 1—see Table S2). When compared more broadly (Figure 6), the Wajak femur falls within the range of male and female Early Upper Paleolithic specimens from Europe, and just within the upper range of male European Late Upper Paleolithic specimens.

FIGURE 5.

Anterior and posterior views of reconstructed Wajak femur, with preserved portions aligned within the average modern femoral template.

FIGURE 6.

Femoral maximum length in East Asian Late Pleistocene (LP) and European Early (EUP) and Late (LUP) Upper Paleolithic samples (see Table S2 for individual specimens). Red: Females; blue: Males; star: Wajak. Sample sizes for males/females: East Asian LP: 4/8; European EUP: 10/6; European LUP: 14/7.

Using a previously published formula (Ruff et al. 2015, 2022), the estimated biomechanical length (the average distance from the superior surface of the femoral neck to the most distal points on the medial and lateral condyles, measured parallel to the shaft axis; Ruff 2002) of the Wajak femur is 439 mm (r = 0.986; standard error of estimate = 4.6 mm). The cut cross section at the proximal end of W‐H‐29, shown in Figure 3, is at about 252 mm, or 57% of biomechanical length, from the distal end. Studies of modern human femora have shown that cross‐sectional morphology varies little between 50% and 57% of biomechanical length, with second moments of area and section moduli differing on average by ≤ 5% (Mongle et al. 2015; Sladek et al. 2010). The general similarity in morphology over this region can also be observed in the Wajak femur (Figures 2 and 5; direct measurements indicate less than 1 mm variation in A‐P and M‐L external breadths). Therefore, comparisons of properties of the Wajak section can be validly made with those taken at the 50% location in the comparative samples. ¹

3.2. Cross‐Sectional Analyses

Cross‐sectional properties of the Wajak femoral cross section are given in Table 2. One characteristic of femoral midshaft regions that has consistently been shown to be sexually dimorphic in preindustrial populations is the ratio of A‐P to M‐L bending rigidity (I _x /I _y ), with males having higher indices, that is, relatively more A‐P elongated sections (Ruff 1987, 2018a; Ruff et al. 2022; Ruff, Wallace, Toya, et al. 2024). This index is 1.374 in the Wajak femur. When compared to other available East Asian Late Pleistocene specimens with assigned sexes (data from Table S2), Wajak falls close to the male median and above the distribution of female values (Figure 7a; nine females, four males in comparative sample). If the total sexable comparative sample is considered (Figure 7b; 23 females, 28 males), there is more overlap between males and females (although the sex difference is still significant: p < 0.01, t‐test), and Wajak falls within both sexes' ranges, but is outside the female interquartile range and much closer to the male median value. Therefore, the Wajak femur is likely to have belonged to a male. This is also consistent with its large size.

TABLE 2.

Wajak femoral cross‐sectional properties.

TA	CA	MA	% CA	I x	I y	I x /I y	I max	I min	Z x	Z y	J	Z p
620	461	158	74.5	33,661	24,491	1.374	33,787	24,365	2027	1820	58,152	3508

Note: TA, CA, MA, total subperiosteal, cortical, medullary areas (mm²); %CA, percent cortical area (CA/TA × 100); I _x , I _y , second moments of area about x and y axes (anteroposterior and mediolateral bending rigidities) (mm⁴); I _max, I _min, maximum and minimum second moments of area (rigidities); Z _x , Z _y , second moments of area about x and y axes (anteroposterior and mediolateral bending strengths) J: polar second moment of area ((twice) average bending rigidity and torsional rigidity) (mm⁴); Z _p , polar section modulus ((twice) average bending strength and torsional strength) (mm³).

FIGURE 7.

Anteroposterior to mediolateral bending rigidity at the femoral 50% (midshaft) section in (a) East Asian Late Pleistocene, and (b) combined East Asian Late Pleistocene and European Early and Late Upper Paleolithic samples (data from Table S2). Red: Females; blue: Males; star: Wajak. Sample sizes for males/females: (a): 4/9, (b): 28/23.

The section modulus, Z _p , is a good indicator of overall bending and torsional strength. The most appropriate body size parameter for evaluating relative differences in Z _p is body mass × bone length (Ruff 2018a). Figure 8 is a plot of log‐transformed femoral midshaft Z _p against body mass × femoral length in the Wajak femur, nine other East Asian Late Pleistocene femora, and 14 European Early Upper Paleolithic and 21 Late Upper Paleolithic femora (data from Table S2) (log transformation is necessary to maintain proportionality across different size ranges). A Reduced Major Axis regression line is run through the pooled sample (r = 0.890, p < 0.0001). The Wajak femur falls slightly above the line and overlaps both Early and Late Upper Paleolithic groups. It is absolutely stronger than any other East Asian Late Pleistocene femur yet analyzed, but is fairly typical in relative strength to Late Pleistocene femora in general. Thus, in terms of inferred mechanical loadings and activity levels (Ruff 2018a), it is close to average for this time period.

FIGURE 8.

Femoral 50% (midshaft) torsional/average bending strength relative to body size in Late Pleistocene samples (data from Table S2): East Asian Late Pleistocene (LP), European Late Upper Paleolithic (LUP), European Early Upper Paleolithic (EUP), and Wajak. Line is reduced major axis regression through total combined comparative sample.

3.3. Geographic and Temporal Comparisons

Box plots of body mass, stature, and BMI for the LMP/ELP, LLP, and living samples, by latitudinal group, are shown in Figure 9. The living samples consist of sex/population means (Table S1) and are subdivided by sex in these plots; the fossil samples are individuals, with sexes pooled (Table 1). The results of the 2‐way nonparametric permutation‐based ANOVAs with latitude and temporal period factors for the fossil samples are shown in Table 3. Results of t‐tests between high and low latitude living population samples are shown in Table 4.

FIGURE 9.

Body size and shape parameters by latitudinal zone and time period: (a) body mass; (b): stature; (c) body mass index. Brown: Late Middle/early Late Pleistocene (LMP/ELP) East Asians; pink: Later Late Pleistocene (LLP) East Asians (fossil samples pooled sex; data from Table 1). Red and blue: Living females and males, respectively, from world‐wide sample (data from Table S1). Horizontal line indicates a single fossil specimen. See Table 1 for composition of each fossil sample.

TABLE 3.

Permutation‐based ANOVA results, latitudinal and temporal effects on body form in fossils.

Dependent variable	Factor	p
Body mass	Latitude	0.0136
	Temporal period	0.0098
Stature	Latitude	0.5952
	Temporal period	0.0754
Body mass index	Latitude	< 0.0001
	Temporal period	0.0008

TABLE 4.

Comparisons between high and low latitude groups among living samples.

	Body mass		Stature		BMI
	n (H/L) a	p b	n (H/L)	p	n (H/L)	p
Males	26/43	< 0.00001	26/41	0.0006	26/41	< 0.00001
Females	24/29	< 0.00001	24/28	0.005	24/27	0.00003

^a Number of high and low latitude population samples.

^b Probability, t‐test.

In all comparisons within temporal period groups (fossil LMP/ELP and LLP, living males and females), individuals or populations from higher latitudes have higher body masses than those from lower latitudes (Figure 9a). For the fossils, this is supported by 2‐way ANOVA: the effect of latitude is significant (p = 0.0136) when accounting for any temporal period effects (Table 3). Temporal period also has a significant effect on body mass among the fossils when accounting for latitude (p = 0.0098), with earlier (LMP/ELP) individuals being heavier than later (LLP) individuals. Latitudinal effects are highly significant within the living sample (p < 0.00001, t‐tests within sexes, Table 4), with higher latitude samples being heavier (Figure 9a). Within the latitudinal group, the living sample overlaps broadly in body mass with the LLP fossil sample, but is smaller than the LMP/EP sample (Figure 9a). Wajak is the heaviest individual in the LLP low latitude sample (Table 1).

Systematic variation in stature between latitudinal and temporal groups is generally much less marked (Figure 9b). Higher latitude living samples are taller on average than lower latitude samples, but with more overlap between groups than for body mass and at a lower level of significance (females, p = 0.005; males, p = 0.0006, t‐tests Table 4). There is no significant effect of either latitude or temporal period on stature within the fossils (Table 3), although the LLP low latitude sample tends to be shorter than other fossil samples (Figure 9b). Living samples overlap broadly with fossils from both temporal periods, within latitudinal groups. Wajak is again the largest (tallest) individual in the LLP low latitude sample (Table 1).

Like body mass, BMI is strongly affected by latitude, with higher latitude individuals and populations having higher values (Figure 9c). For the fossil samples, this is highly significant (p < 0.0001, controlling for temporal period) (Table 3). Temporal period also has a significant effect, with earlier LMP/ELP individuals having higher values than LLP individuals (p = 0.0008, controlling for latitude). As with body mass, there is no interaction between latitude and temporal period, indicating similar effects of latitude in both periods. The living sample shows a similar strong correspondence between latitude and BMI (p < 0.0001, t‐tests within sexes, Table 4). It has generally lower BMIs within the latitudinal group than the LMP/ELP fossils. The LLP fossil group shows more overlap in BMI with living samples, although it tends to be in the higher range of living values, within the latitudinal group (Figure 9c). Wajak falls near the middle of the LLP low latitude sample (Table 1).

An increase in BMI at higher latitudes, in all groups, suggests a relatively wider body. ² This can be tested by comparing bi‐iliac (maximum pelvic) breadth relative to stature against BMI. This has been done in Figure 10, using an anthropometric data set of 58 sex/population means that includes bi‐iliac breadth (Ruff et al. 2005). There is a moderate but highly significant correlation between bi‐iliac breadth/stature and BMI (r = 0.60, p < 0.00001). As expected, lower and higher latitude samples are well differentiated in both BMI and bi‐iliac breadth/stature (the high outlier in BMI among lower latitude samples is from a South Pacific population known to be characterized by this type of body form, which is adaptive for their oceanic environment—see Houghton 1990; Ruff 1994). As shown in Figure 9c, the fossil specimens fall on the higher end of variation in BMI among living samples for similar latitudinal zones, implying that they had relatively wide bodies compared to living populations in those regions. This is particularly marked for the earlier LMP/ELP specimens.

FIGURE 10.

Bi‐iliac (maximum pelvic) breadth/stature (both in cm) relative to body mass index in modern humans from low (≤ 30°) and high (> 30°) latitudes (data from Table S5). Line is reduced major axis regression through total combined sample.

A schematic representation of latitudinal and temporal variation in body size and inferred body shape among the East Asian fossil specimens is shown in Figure 11, based on results in Figures 9 and 10. The greater variation in body mass and relative body breadth than in stature is apparent, except for a slight decline in stature among more recent low latitude specimens. Both body mass and relative body breadth decline with decreasing latitude and time. Note that because latitude and temporal period both affect body form, it is necessary to consider both factors when interpreting patterns of variation, that is, the effects of latitude are best evaluated within a similar time range, and the effects of time within a similar latitudinal zone.

FIGURE 11.

East Asian map with schematic representation of latitudinal and temporal variation in body size and shape. Dark silhouettes: Late Middle/early Late Pleistocene; lighter silhouettes: Later Late Pleistocene. Red squares show fossil localities (also see Figure 3). (Map courtesy of Mapswire (mapswire.com/maps/asia/)).

4. Discussion

4.1. Climatic Effects

Our results support the application of ecogeographic rules linking body form to climate among Pleistocene East Asian hominins. In both late Middle/early Late Pleistocene and later Late Pleistocene time periods, individuals from higher latitudes are larger in body mass than those from lower latitudes, conforming to expectations based on Bergmann's Rule. Higher latitude individuals are also relatively wider‐bodied, which is another mechanism for reducing surface area/body mass in colder climates (Ruff 1994).

The application of Bergmann's Rule to recent humans has been strongly questioned, due to the confounding effects of nutrition and socioeconomic status on body size (Bogin et al. 2022). It is unlikely, however, that such factors were generally applicable to populations in the distant past, prior to the adoption of food production and large‐scale changes in socioeconomic structure occurring during the Holocene. There is evidence from broader studies that Bergmann's Rule applies across Pleistocene Homo world‐wide (Stibel 2023; Will et al. 2021), especially during the Late Pleistocene (Ruff et al. 2022). There is also evidence that body breadth is much less responsive to shorter‐term secular effects such as nutritional status than is stature (Ruff 1994). As shown here and elsewhere (Niskanen et al. 2018; Ruff 1994), body breadth is the principal factor behind ecogeographic and temporal variation (the latter within similar ecogeographic zones) in body mass. Thus, while other environmental influences undoubtedly also contribute to variation in human body size and shape, climatic adaptation is an underlying factor that must be considered in interpreting patterns of variation among past hominins.

Why stature shows a weaker relationship to climate than either body mass or body breadth (Gustafsson and Lindenfors 2009; Ruff 1994; Stibel 2023; and present study) can be explained by considering the effects of variation in these components of body form on surface area/body mass (SA/BM), the underlying physiological basis of Bergmann's and Allen's Rules. Using a “cylindrical model” of the body, it can be shown that variation in stature has no effect on SA/BM, while variation in body breadth always changes SA/BM—a larger breadth results in a smaller ratio (Ruff 1994). Thus, the primary mechanism for altering SA/BM among living humans is to change body breadth (which also changes body mass). This also appears to be the case among the East Asian Pleistocene samples examined here.

The use of broad latitudinal zones (above or below 30° N. latitude) rather than specific paleoclimatic data to test this model introduces some uncertainty in the comparisons. However, in a taxonomically wider study of Bergmann's Rule, analyses using latitude and temperature gave consistent results (Ashton et al. 2000), and latitude has been used as a proxy for temperature in a number of previous investigations of hominin morphology (Foster and Collard 2013; Roseman and Auerbach 2015; Ruff 1994). Foster and Collard (2013) reported that body size trends among modern humans were only significant over a wide latitudinal range—in their study, ≥ 50°. The fossil samples in the present study cover slightly less than this range, with the high latitude individuals all deriving from sites about 40° N. latitude, and almost all of the low latitude individuals deriving from sites close to the equator (Figure 1). It is possible that Pleistocene Homo shows greater effects of climate on body form than living humans (Will et al. 2021), in part because of greater cultural buffering among recent humans (Ruff 1994). Nevertheless, when our living samples, which include much finer gradations of latitude (Table S1), are divided at 30° N. latitude (mean low latitude samples: 11°; mean high latitude samples: 48°), they show similar, highly statistically significant differences in body mass and BMI. Thus, our relatively simple binary method for assessing climatic effects was probably adequate in the present context, and in any event was necessary given the nature of the available fossil data.

Sample sizes for the fossils are small, particularly at high latitudes, where only single individuals with both body mass and stature estimates are represented in each temporal group. Thus, our conclusions must be regarded as tentative until larger samples become available. The trends apparent in the present data set are fully consistent with those observed in recent populations and with theoretical predictions; however, with no obvious deviations.

Sexual dimorphism in body size is another factor that may have influenced some of the observed patterns, especially given the small fossil sample sizes and thus heightened chance of nonrandom sampling of males and females. However, the one early high latitude individual, Jinnuishan 1, is a female, suggesting that the trend for increased body mass in earlier and higher latitude individuals may actually be underestimated. While most fossil specimens in the present study cannot be definitively sexed (Table 1), other evidence suggests that there is likely a mix of males and females within the larger samples (Ruff et al. 2022; Weidenreich 1951). Sexual dimorphism in body mass is similar to or only slightly increased in late Middle and Late Pleistocene Homo relative to living humans (Ruff et al. 2018, 2022), so the effects of possible non‐equal representation of the sexes should not have been exacerbated in our samples.

4.2. Temporal Effects

There is a decline in body breadth and body mass between our late Middle/early Late Pleistocene (ca. 260–112 ka) and later Late Pleistocene (ca. 40–28 ka) East Asian groups, within both climatic zones. This result is consistent with broader world‐wide temporal trends during the Late Pleistocene (Holliday 2012; Hooijer 1950; Ruff et al. 2022, 1997; Ruff and Wood 2023; Stibel 2023). It is unlikely to have been caused by climatic changes, since temperatures world‐wide generally declined during the Late Pleistocene and were lower (or at least, not higher) in our later group than in our earlier group (Klein 1989; Stibel 2023). While other environmental factors such as nutrition and population demography may have contributed to changes in body size in the Terminal Late Pleistocene (e.g., Formicola and Holt 2007; Frayer 1981), perhaps a more important factor operating throughout the Late Pleistocene was advancement in technology that led to less selective pressure for maintaining large bodies (Churchill 1993; Frayer 1984). For example, there is accumulating evidence for use of the bow‐and‐arrow in both Africa and Eurasia, including tropical South Asia, prior to 40,000 BP (Backwell et al. 2018; Langley et al. 2020; Langley and Lombard 2025; Lombard and Phillipson 2010; Lombard and Shea 2021; Plisson et al. 2025). Such a change in technology, along with other innovations like the spear‐thrower, would have the effect of placing the individual wielding it further from large prey, with its attendant dangers (Berger and Trinkaus 1995; Shea 2009), and necessitating less overall physical size and strength to utilize effectively (Brues 1959; Frayer 1981). There is evidence for a slight reduction in sexual dimorphism in body mass from the Middle Pleistocene to the later Late Pleistocene (Early/Middle Upper Paleolithic) (Ruff et al. 2018), which might also reflect such a technological change (Frayer 1981, 1984). The development of more effective footwear and warm clothing in the later Late Pleistocene (Dennell et al. 2020; Trinkaus and Shang 2008) is another important technological advance that may have reduced the selective advantage of large body size in colder climates. Such broad temporal trends need to be evaluated in the context of regional variation (e.g., Li 2023; Nishiaki et al. 2021).

Of course, population movement and replacement/interbreeding may also have affected the temporal patterns that we observed. Recent archaeological, paleontological, and genetic evidence indicates a complex population history for East Asia during the Late Pleistocene, with multiple migrations and a lack of regional continuity (Bae et al. 2017; Dennell et al. 2020; Freidline et al. 2023; Fu et al. 2025; Sikora 2017). Still, the fact that declines in body mass occurred in multiple regions of the Old World—Subsaharan Africa, Europe, and East Asia—during the Late Pleistocene (Ruff et al. 2022) argues for some common underlying factor. ³ This could in part be attributed to broader evolutionary developments, that is, the spread of anatomically modern humans—as noted earlier, our temporally earlier group included specimens characterized as late H. erectus or transitional between H. erectus and H. sapiens , while our later group is attributed to H. sapiens (and the contribution of Denisovans to either temporal group is also uncertain; Fu et al. 2025). In any event, the fact that latitudinal distinctions between body forms persisted across time periods is strong evidence that climate had a consistent effect on morphology despite any changes resulting from other environmental or long‐term evolutionary trends.

4.3. Wajak Femur

The reconstructed Wajak femur is very long in comparison to most other East Asian Pleistocene femora—slightly longer than Tianyuan 1 and substantially longer than femora from Trinil (Ruff et al. 2015) and Sangiran (Kresna 11) (Puymerail et al. 2012), as well as other more recent specimens (Ruff et al. 2022). The femur associated with Ngandong Tibia A was probably even longer, though—using a crural (tibia/femur length) index within the range of modern human populations (78.4–85.8; Ruff and Walker 1993) would imply a femur maximum length of over 500 mm, equivalent to the longest Early Upper Paleolithic femora from Europe (Figure 6). This, and the long Wajak femur, may seem counterintuitive given their latitudinal position. However, as discussed above, stature is poorly correlated with latitude or temperature among modern humans, because of its minimal effects on surface area/body mass. Modern tropical populations vary greatly in stature, overlapping substantially with higher latitude populations. However, because of their relatively narrower bodies, body mass indices remain below those of higher latitude populations (Figure 9). Based on body mass estimates, the same can be inferred for the lower latitude specimens in our study, including Wajak and Ngandong 13. BMI is somewhat higher in both low and high latitude East Asian Pleistocene Homo than modern humans from the same latitudinal zones (this can also be demonstrated among just East Asian modern humans—see Table S1), and BMI declines between our two temporal groups, suggesting a long‐term decrease in relative body breadth in this region. However, some unusually short East Asian Late Pleistocene individuals/populations have relatively wide bodies (Ruff et al. 2022; also see data in Shackelford and Demeter 2012), which is not incompatible with the “cylindrical model” (Ruff 1994), so interpretation of BMI in isolation must be done with caution.

One consistent observation regarding Pleistocene East Asian samples is that very small‐bodied individuals are found only in lower latitudes (along with taller and somewhat wider‐bodied individuals). Examples include H. florensiensis, Indonesia (Brown et al. 2004; Jungers et al. 2009), H. luzonensis , Philippines (Détroit et al. 2013, 2019), and H. sapiens from Minatogawa, Okinawa (Suzuki and Hanihara 1982) and Tam Hang, Laos (Shackelford and Demeter 2012). The first three samples are from islands and at least the first two of these are likely products of island endemism, so environmental effects in addition to climate were likely operative. Island endemism can also not be excluded as a possible contributing factor to the patterns of variation that we observed here. However, the lack of evidence for any very small‐bodied individuals or populations from higher latitudes—in the past and in the present—suggests the possibility of greater limitations on body form in cold environments (also see Will et al. 2021).

Analyses of the midshaft cross‐sectional geometry of the Wajak femur support its attribution as a male, which is consistent with its long length. Its overall strength relative to body size is normal compared to other later Late Pleistocene anatomically modern humans from East Asia and Europe. This supports other evidence for a mobile, hunting‐gathering subsistence economy in populations from this general region and time period (Barker et al. 2007).

Overall, our results support our previous suggestion that a tall‐bodied human population (represented by Wajak) and a small‐bodied human population (represented by Trinil 9) may have co‐occurred in Java in the Late Pleistocene, and possibly in an even larger area including Palawan (Tabon Cave) and Borneo (Niah Cave) (Ruff et al. 2022). This pattern of two co‐existing morphologically distinct human populations may be due to intra‐species niche partitioning, driving differences in body form as an adaptation to their respective local environments and/or foraging strategies. Barker et al. (2007) established that the small‐bodied Niah Cave inhabitants exploited a range of terrestrial, terrestrial‐arboreal, and aquatic vertebrates as well as plant resources from different habitats including tropical forest, swamp, open woodland, scrub, and the shores of rivers and lakes. It might be expected that the foraging strategy of the tall‐bodied Wajak population, likely living along the Indian Ocean coast of southern Java, may have been different from that of the small‐bodied population. However, so far the ecological niche of the Wajak population is unknown and deserves further study.

Finally, the body mass estimate based on the Wajak femur can be used to calculate relative cranial capacity (encephalization), in conjunction with the Wajak I and II cranial remains. As noted earlier, there is no direct evidence linking the Wajak cranial material with the postcrania recovered from the site, although dates determined for the femur analyzed here and the Wajak II specimen overlap in their 95% confidence limits (Storm et al. 2013) (no dates were obtained for Wajak I). Our personal observations indicate a similar degree of fossilization, color, and general condition of all three specimens. Furthermore, both skulls are likely from males (Storm 1995), as is our assignment of the femur.

As noted by Storm and Nelson (1992), cranial capacity estimates of the more complete Wajak I specimen have varied depending on measurement technique, from 1475 cc (Coon 1962) to 1550 cc (Dubois 1922) to 1633 cc (Jacob 1967). Taking an average of these three estimates (1553 cc), converting to brain mass (1493 cc), and calculating encephalization quotient (EQ) using the formulae in Ruff et al. (1997), with a body mass estimate of 64.2 kg, produces an EQ of 5.63. This is very similar to (within 0.5 SD of the mean of) both a sample of very recent humans and Early and Late Upper Paleolithic samples (Ruff et al. 1997). From comparisons of corresponding dimensions in Wajak I and II, it is apparent that Wajak II is from a larger individual than Wajak I (Dubois 1922; Storm 1995), although as noted above both are probably from males. If the Wajak femur is associated with Wajak II rather than Wajak I (or from an individual within their size range), and cranial capacity was correspondingly larger, the EQ may have been somewhat larger as well but still very likely well within the range expected for an anatomically modern human.

5. Conclusions

Although fossil sample sizes are small, ecogeographic rules whereby body size and shape are adapted to climatic conditions appear to apply to available East Asian late Middle/Early Late Pleistocene (ca 115–112 ka) and later Late Pleistocene (28–40 ka) Homo. In both time periods, estimated body mass and BMI are larger in higher latitude individuals than in lower latitude individuals, while stature shows less consistent trends. These patterns are similar to those observed among living humans, and are concordant with theoretical expectations based on varying surface area to body mass in response to temperature. There is also a temporal trend of decreasing body mass and relative body breadth between our fossil samples from both latitudinal zones, which may be related to increasing technological sophistication and reduced selection pressure for large body size through the Late Pleistocene. The reconstructed femur from Wajak, Indonesia (≥ 36.5 ka) is very long, but still conforms to these expectations through having an inferred relatively narrow body form. Its strength relative to body size is similar to that of European Upper Paleolithic femora, implying broadly similar behavioral patterns. Our results show that both time and climate, and possibly niche partitioning, must be considered in interpreting variation in body size and shape in East Asian Pleistocene Homo.

Author Contributions

Josephine C. Joordens: conceptualization, funding acquisition, writing – review and editing. Adam D. Sylvester: investigation, methodology, visualization, writing – original draft. Eduard Pop: conceptualization, funding acquisition, investigation, resources, writing – review and editing. Paul Storm: conceptualization, resources, writing – review and editing. Christopher B. Ruff: conceptualization, formal analysis, methodology, visualization, writing – original draft. Sofwan Noerwidi: resources.

Funding

Supported by Dutch Research Council NWO grant OCENW.M20.152 to E.P. and J.C.J.

Ethics Statement

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: (a–c) Superior (top row) and anterior (bottom row) views of left femur and right and left tibiae. (a) W‐H‐13583‐1 left femur (in superior view, anterior is to the left). Anteroposterior (A‐P) breadth of head: ≥ 44 mm. (b) W‐H‐32 right tibia (in superior view, medial is to the left, anterior above). Preserved mediolateral (M‐L) breadth of proximal surface: 72.6 mm; estimated original breadth: 73 mm. Breadths at distal surface: A‐P (maximum): 38.8 mm; M‐L (perpendicular to maximum): 28.0 mm. (c):W‐H‐33 left tibia (in superior view, medial is to the right, anterior above). Preserved mediolateral (M‐L) breadth of proximal surface: 74.0 mm; estimated original breadth: 75 mm. (d) W‐H‐40 right talus, superior (top) and medial (bottom) views (in superior view, medial is below, anterior to the left). Total length: 56 mm; M‐L breadths across superior surface: anterior: 34.0 mm; midpoint: 33.5 mm; posterior: 29.5 mm. (e, f) Anterior views of femoral midshaft (e, W‐H‐13584‐5) and distal shaft (f, W‐H‐13584‐2) fragments (distal is below). (g) Anterior view of distal (W‐H‐13584‐7) and shaft (W‐H‐13584‐6) fragments of left humerus. Breadths at proximal end of distal fragment: A‐P: 20.0 mm; M‐L: 17.2 mm. Distal end is fragmented and encased in matrix. (h) Anterior view of distal (W‐H‐10021‐44) and shaft (W‐H‐10021‐26, two conjoining pieces) fragments of left radius. Breadths at break between two shaft pieces: A‐P: 9.9 mm; M‐L: 16.1 mm. Styloid process is completely worn away. For humerus and radius, distal is below, medial to the right.

Text S1: Other postcranial specimens from Wajak.

Table S1: Sex/sample means and sources for anthropometric data in living samples.

Table S2: Comparative sample for cross‐sectional analyses.

Table S3: Body mass estimation equations used in present study.

Table S4: Stature estimation equations.

Table S5: Living sample for bi‐iliac breadth/stature to Body Mass Index (BMI) comparison in Figure 10.

Data Availability Statement

All fossil and modern data used in the study are available in Supporting Information.