How and Why Humans Grow Thin Skulls: Experimental Evidence for Systemic Cortical Robusticity

AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 101:217-236 (1996)

DANIEL E. LIEBERMAN Department of Anthropology, Rutgers Uniuersity, Douglas Campns, New Brunswick, New Jersey 08903-0270 KEY WORDS Skull, Neurocranium, Cortical bone, Robusticity, Homo, Pigs, Armadillos

ABSTRACT To what extent is cranial vault thickness (CVT) a character that is strongly linked to the genome, or to what extent does it reflect the activity of an individual prior to skeletal maturity? Experimental data from pigs and armadillos indicate that CVT increases more rapidly in exercised juveniles than in genetically similar controls, despite the low levels of strain generated by chewing or locomotion in the neurocranium. CVT increases in these individuals appear to be a consequence of systemic cortical bone growth induced by exercise. In addition, an analysis of the variability in vault thickness in the genus Homo demonstrates that, until the Holocene, there has been only a slight, general decrease in vault thickness over time with no consistent significant differences between archaic and early anatomically modern humans from the Late Pleistocene. Although there may be some genetic component to variation in CVT, exercise-related, non-genetically heritable stimuli appear to account for most of the variance between individuals. The thick cranial vaults of most hunter-gatherers and early agriculturalists suggests that they may have experienced higher levels of sustained exercise relative to body mass than the majority of recent, post-industrial humans. 0 1996 Wiley-Liss, Inc.

Nature is not greatly concerned over the precise Neanderthals. Cranial vault thickness thickness of the cranium. (CVT) is frequently used as a character to Todd (1924:255) make inferences about the phylogenetic relationships among recent taxa of Homo (e.g., Why modern humans have generally thin- Stringer, 1984, 1987; Thorne and Wolpoff, ner bones and, in particular, thinner skulls 1981; Groves, 1989; Frayer et al., 1993), and than archaic humans has long been a subject to make inferences about the behavioral dif- of speculation for clinical and hominid ferences between modern and archaic hu- palaeontological research. The degree of cor- mans (e.g., Coon, 1962; Brace, 1979). tical robusticity throughout the skeleton CVT is an interesting character to exam- and, more specifically, in the cranial vault ine in depth because the vault grows some- is often considered a major morphological what differently than long bones that have distinction between anatomically modern traditionally been the focus of most research Homo sapiens and earlier taxa ofHomo (e.g., on cortical robusticity. Unlike bones that Wolpoff, 1980; Stringer, 1988). Weidenreich form within a cartilagenous framework, the (1943), for example, noted that the mean pa- bones of the neurocranium-the parietals rietal thickness at bregma of modern Euro- peans, 5.5 mm, is 60% thinner than that of the Sinanthropus pekinensis fossils from Zhoukoudian and 40% thinner than that of Received August 29, 1995; accepted April 29, 1996.

0 1996 WILEY-LISS, INC 2 18 D.E. LIEBERMAN and the squamous portions of the occipital, vaults are a possible adaptation for pro- frontal and temporal-all develop intra- tecting the skull from injury, suggesting that membranously from the membranes that “the technological innovations of the Late surround the inside and outside, respec- Stone Age, including the use of efficient long- tively, of the vault. Early in development, distance hunting weapons such as bows with osteoblasts in these membranes rapidly de- poisoned arrows reduced the necessity for posit highly vascularized woven bone within large size and skeletal robustness in hunting numerous ossification centers (Ohtsuki, populations.” According to this hypothesis, 1977).As growth slows after birth, both the the recent decline in vault thickness most inner membrane, the endocranium, and the likely reflects a relaxation of selection pres- outer membrane, the pericranium, switch to sures to maintain thick vaults, presumably depositing vascularized lamellar bone, form- because thin vaults are metabolically less ing the inner and outer tables of the neuro- expensive to grow or support. Similar argu- cranium (Sperber, 1989). Osteoclasts and ments are found in Brace (1979:67), Stringer then osteoblasts begin to invade the center (1988:268), and elsewhere. of the vault at around age four, resorbing and A second hypothesis is that differences in remodeling the woven bone and the earliest CVT reflect nonheritable, in vivo responses formed lamellae into the trabecular bone of to mechanical force. Bone tissue interacts the diploe, which comes to contain haemo- dynamically with its mechanical environ- poietic cells (Williams et al., 1989). It is im- ment. Force applied to a bone (quantified portant to note that, unlike most limb bones per unit area as stress, a), generates strain and other cranial bones, both the endocran- (deformation, E) whose cumulative effects, ial and pericranial surfaces of the superior if sufficient in magnitude, can damage its half of the vault constitute depository microstructure and mechanical integrity growth fields in human and nonhuman pri- (Carter, 1987; Martin and Burr, 1989, Mar- mates (Duterloo and Enlow, 1970; Enlow, tin, 1992.). Controlled experiments on both 1990).Consequently, the vast majority of the limb bones that form endochondrally (e.g., neurocranial cavity does not expand though Biewener et al., 1986;Woo et al., 1981;Rubin drift (resorption of the inner layer and depo- and Lanyon, 1984, 1985) and facial bones sition of the outer layer) but instead expands that form intramembranously (e.g., Corruc- from tension-induced growth within the su- cini and Beecher, 1982, 1984; Bouvier and tures of the vault. A small amount of drift Hylander, 1981;Yamada and Kimmel, 1991) does occur, however, at the sutural margins demonstrate that high levels of strains in- of the neurocranial bones, preventing duce local osteoblastic responses that in- steeply angled sutures (Enlow, 1990). The crease cortical bone mass. Such remodeling bones of the superior half of the vault, there- can be adaptive because it increases the dis- fore, can grow only thicker and do so inde- tribution of mass in the plane(s) of deforma- pendently of any increases in cranial ca- tion, thereby reducing the amount of strain pacity. generated by a given force (Biewener et al., This paper asks to what extent as a char- 1986; Frost, 1986,1987,1988). The most im- acter CVT is strongly linked to the genome portant sources of mechanical force on the and to what extent it reflects epigenetic re- cranial vault are clearly those from chewing. sponses to exogenetic stimuli. Despite infor- It is a reasonable hypothesis that strains mation on the proximate cellular processes from the tension produced by the m. tempo- by which CVT develops, there is little under- rulis on the outer table of the vault or strains standing of the mechanisms that cause some generated from other chewing-related forces individuals or taxa to have thicker cranial generated elsewhere in the cranium may af- vaults than others. Three hypotheses have been proposed to account for variability in nonpathological cranial vault thickness in ‘A variety of pathologies, including haemolytic blood dysplasias humans.l The most common hypothesis is associated with sickle cell anemia and thalassemia, can cause unusual thickening of the diploic layer of the cranial vault (Webb, that CVT is subject to selection. Wolpoff 1990). Such cases, however, are rare in the fossil record and are (1980:333),for example, points out that thick not considered here. CRANIAL. VAULT THICKNESS 2 19

fect the rate of bone growth in the cranial selection. If, however, CVT variation is an vault (Weidenreich, 1941; Washburn, 1947; in vivo, exogenetic response to higher me- Moss, 1954; Moss and Young, 1960; Naw- chanical forces, then such strains should be rocki, 1991, 1992). As Hylander (1986) has sufficient to replicate them in laboratory shown, chewing harder foods generates more conditions. In addition, this hypothesis strain in the mandible and face than chew- would predict that humans who eat soft diets ing soft foods, so variation in CVT may re- will have thinner vaults than those who eat flect responses to diets of different hardness. harder diets. Finally, if CVT is merely a con- Locomotion may also generate strains in the sequence of elevated circulating hormones cranium, resulting in the prediction that hu- that stimulate osteogensis, then thickened mans who run more may have thicker skulls cranial vaults will develop in laboratory ani- (Bhanba, 1961). mals who have elevated GH levels, and such Finally, a few researchers (e.g., Twiessel- individuals will also have an overall high mann, 1941; Kennedy, 1985; Nawrocki, degree of cortical robusticity. An additional, 1991; Nelson and Gauld, 1994) have pointed related question is whether a thin cranial out that differences in the levels of certain vault is a derived character of anatomically circulating hormones may also influence modern humans or whether it is a more re- CVT. All bone growth is mediated by hor- cent phenomenon. Although generally cited mones on both local and systemic levels. as a defining character of modern humans, Growth hormone (GH) in particular has there has been little systematic investiga- well-documented effects on CVT and overall tion of variability in CVT among recent hu- cortical robusticity. For example, acromega- man taxa (see, however, Nawrocki, 1991). lics who have persistent and high levels of This paper, therefore, presents the results GH have significantly thicker cranial vaults of two interrelated analyses. First, it reports than normal individuals, whereas sufferers details of controlled laboratory studies on from GH deficiencies (e.g., hypopituitarism) the effects of exercise on local and systemic have thinner-than-average cranial vaults cortical bone growth in nonhuman animals. (Randall, 1989; Vogl et al., 1993; Pirinen et These experiments demonstrate that signifi- al., 1994). Other circulating hormones such cant differences in overall cortical robusti- as insulin-like growth factor (IGF-I), para- city, including the thickness of the cranial thyroid hormone (PTH), calcitonin, and even vault, can occur in genetically identical and/ insulin could also have similar effects (Bu- or closely related animals who experience chanan and Preece, 1992).Levels of circulat- different levels of exercise during their de- ing hormones such as GH might be higher velopment. These results indicate that there in taxa with thicker vaults because of the is a strong nonheritable component to CVT influence of selection on the activities of the variability. Second, these conclusions are endocrine system or because of other non- supported by an analysis of changes in corti- heritable factors that induce endocrine cal vault thickness in Pleistocene Homo responses. Exercise, for example, substan- which reveals that, despite a slight trend tially elevates circulating GH levels in hu- towards thinner skulls, there are not many mans (Felsing et al., 1992). significant differences over time or between Each of these hypotheses, which are not taxa until the Holocene. exclusive, makes predictions that can be tested in the laboratory on nonhuman animals and/or using comparative data on liv- MATERIALS AND METHODS ing and fossil human samples. If CVT varia- The processes that cause variation in CVT tion is a consequence of selection in relation were studied in controlled laboratory experi- to cultural factors, then human crania from ments in which confounding factors such as Upper Palaeolithic contexts should have age, nutrition, sex, and genetic variance thinner vaults than those from Middle or were controlled among the subjects. I report Lower Palaeolithic contexts. In addition, here the results of separate experiments on CVT would have to be a predominantly heri- two species of laboratory animals: the min- table trait for it to be a target of natural iature domestic pig (Sus scrofa) and the 220 D.E. LIEBERMAN

TABLE 1. Experimental protocols

Pigs (Sus scrofa) Armadillos (Dasypus nouemcinctus) N 2 females, 4 males 2 males 3 runners, 3 controls 1 runner, 1control Relatedness Inbred siblings Identical twins Age at start 1 month 2 months Age at stop 4 months 8 months Exercise 3.0 mph AM, 30 min 1 mph AM, 60 min 3.0 mph PM, 30 min Dyes Calcein, 1month Calcein, 2 month Tetracycline, 2 months Tetracycline, 4 months Alizarin, 3 months Alizarin, 6 months Longitudinal data Weight (kg) Weight (kg) Lateral X-rays of tibia, crania Strain data Tibia (proximal medial) Tibia (proximal medial) Supraoccipital (lateral to sagittal suture) common nine-banded armadillo (Dasypus tween, the animals were anaesthetized nouemcinctus). The protocols for both experi- (6 mgkg Telazol and 0.04 mg/kg Atropine) ments are summarized in Table 1. For a vari- in order to weigh them, take blood samples, ety of reasons (discussed below), more data and radiograph their heads and right hind was acquired from the six pigs; the data from limbs in lateral view to document longitudi- the two armadillos are less comprehensive nal growth rates in the cranium and tibia. but are included because the subjects were A record of longitudinal bone growth was genetically identical twins. also provided by intraperitoneal injections Pigs of fluorescent dyes that incorporate rapidly into bone mineral (Frost, 1969, 1983; Skin- The subjects were two female and four ner and Nalbadian, 1975). Calcein (20 mgf male piglets, all siblings from a single litter kg), which appears green under flourescent of an inbred strain of miniature swine light, was injected at age 1month; oxytetra- (Charles River Laboratories, Wilmington, cycline (70 mgkg), which appears orange MA). At weaning, 1 month after birth, the under flourescent light, was injected at age pigs were almost identical in weight (see Ta- 2 months, and alizarin red (50 mgkg),which ble 3) and were arbitrarily divided into an dyes the bone red, was injected at age 3 exercise group of runners and a control months. group, each of which consisted of one female Close to the end of the experiment, (Tokyo and two males. The exercise group was Sokki Kenkyujo Co., Tokyo, Japan) rect- trained to run on a treadmill for 30 min each 45” morning and 30 min each afternoon; after 1 angular strain gauges with 120 -t 0.5 ohm month they were able to run comfortably at resistance (FRA-1-11)were placed in two lo- a speed of 3.0 mph. These levels of exercise cations on one of the exercised animals. One approximate what a pig might habitually ex- gauge was placed on the dorso-medial aspect perience in wild conditions. In addition, the of the proximal end of the left tibia where exercise group was kept in a large pen there are no muscle attachments; the other (13.5 m2), whereas the controls were con- gauge was placed on the squamous portion fined to a smaller pen (2.25 m2).The control of the occipital (the supraoccipital) just lat- animals were not kept immobile but spent eral to the midsagittal line. For these proce- most of each day walking about their pen. dures, anaesthesia was induced by telazol All other aspects of their conditions were (6.0 mg/kg) and atropine (0.04 mg/kg) and identical, including diet, exposure to light, maintained with halothane (Muir and Hub- and temperature. The experiment ran for 3 bell, 1989). To affix the gauges, a roughly months, and the animals were sacrificed at 4 x 4 mm window was cut in the periosteum the age of 4 months, after the cessation of to expose the surface of the bone, the bone neural growth. At the beginning and end of surface was degreased with chloroform, and the experiment and every 2 weeks in be- the gauge was bonded with super-glue CRANIAL VAULT THICKNESS 22 1

(methyl-2-cyano-acrilate). Each gauge was supraoccipital just to the left of the sagittal connected with insulated wire to a Vishay suture and from the midshaft, proximal 2120 amplifier to form one arm of a third, and distal third of the left femur, tibia, Wheatstone bridge circuit. Bridge excitation fibia, and metatarsal bones. Sections were was 1V, and voltage outputs were recorded embedded in Epotek'" epoxy, cut with an on a Bell and Howell CPR 4010'" magnetic Isornet'" low speed saw, affixed to slides, and tape recorder at 15 ids. ground down to approximately 100 micron Strain levels were recorded 2 days after thickness with a Beuhler Petrothin'". Sec- surgery; at this time, the pig had no limp tions were examined in plain and cross-po- and was able to move its head freely and larized transmitted light and reflected without any evidence of pain. Recordings flourescent light using an Olympus'" SZH were made when the animal was running stereomicroscope and then were digitally at 3.0 mph and when it was chewing hard captured for computer image analysis using pellets. Gauges were periodically calibrated a color video camera connected to a Quick and balanced when the animal was not ac- Capture'" video input board in a Macintosh tive to record zero levels of strain. The sub- 1IT"computer. Each image was analyzed us- ject was filmed in lateral view in normal ing Image (version 1.52). light at 100 frame& with Kodak 16mm Plus-X reversal film (no. 7276) in order to Armadi Ilos correlate strain gauge activity with footfall The protocol for the armadillo experiment during running. A voltage pulse triggered by was similar to that used for the pigs but with the camera shutter was recorded by the tape several important differences. Only one pair recorder, allowing precise synchronization of of genetically identical armadillo twins was frames with the strain data. After each ex- available. As with the pigs, one armadillo periment, still X-ray photographs were exercised by running on a treadmill every taken to pinpoint gauge location and the ori- day for 60 min but at 1.0 mph. The duration entation of the gauge axes. Selected portions of the experiment was for 6 months after ofthe strain gauge data were played through weaning, which occurred at the age of 2 an A-D converter into a Macintosh 11" com- months. Dyes were administered in the same puter at 500 pointsh and analyzed using sequence and at the same doses but every 8 Labview I1 software (program developed by weeks. Close to the end of the experiment, a K. Johnson, Duke University). These data strain gauge was placed on the dorsomedial were integrated to calculate microstrain aspect of the proximal end of the left tibia units (p# of tension (EJ, compression (EJ, of the exercised animal as described above. shear and the orientation of tension Longitudinal weight data were recorded ev- in degrees relative to the axis of the A ele- ery 2 weeks, but lateral radiographs of the ment of each gauge (for fomulae and discus- skull were only taken at the end of the exper- sion, see Dally and Riley, 1978; Biewener, iment. Immediately after being sacrificed, 1992). the animals were defleshed and their skulls Immediately after the animals were and tibia fixed in a 5% gluteraldehyde solu- killed, the bones were defleshed and cleaned, tion buffered to pH 7.0. These bones were and samples from the supraoccipital and the then dehydrated in ethyl alcohol, cleared in right tibia were fixed in 10% formaldehyde xylene, embedded in Osteobed'" polymer, solution. The rest of the skeleton was and sectioned as described above. cleaned using dermestid beetles and degreased with a 25% ammonia solution. For Measurements and analysis each individual, sections were cut from the For the cranial vaults, total thickness was measured directly from sections of the anteromedial corner of the left supraoccipital (where a strain gauge was attached in the 'Strain (€1 is defined as AL/L, in which L is the original length pigs). In addition, a specially modified ver- of an object and AL is its change in length when a force is appIied. By convention, strain is calculated in dimensionless units of sion of Image 1.52 (by B. Guilford, Univer- microstrain (+E) which equal 1 X mdmm (or idin). sity of Arizona) was used to measure cortical 222 D.E. LIEBERMAN and medullary areas as well as the second the consequence of non-genetically heritable moment of inertia, I, around the mediolat- influences related to exercise. era1 (x) and dorsoventral (y) planes from cross-sections of the limb bones. I, and I, Comparative fossil hominid data are calculations of the distribution of mass In addition to testing the effects of strain around the neutral axis of the bone in these and exercise on vault thickness in pigs and planes (see Wainright et al., 1976; Ruff and armadillos, CVT values were compared in Hayes, 1983a; Biewener, 1992; Ruff, 1992). adult human and fossil hominid crania at Other measures of CVT were taken on the two locations: bregma, the intersection of the dried skulls using Mitutoyo'" digital calipers frontal and parietal bones where the coronal (accurate to 0.01 mm) at the following loca- and sagittal sutures meet, and the parietal tions: the center of the frontal along the mid- eminence (tuberosity), which marks the ini- sagittal axis, bregma (average of parietal tial center of ossification of the bone. The and frontal), the parietal eminences, and the fossil sample (Table 2), all from the genus maximum thickness of the nuchal region in Homo, are divided into five groups: 1) H. the midsagittal plane. A number of other erectus, 2) early archaic humans (which in- comparative measurements were taken on cludes non-Neanderthal fossils attributed to dried bones using calipers. These measure- Homo sp. indet. from Africa Asia and Eu- ments include the dorsoventral and medio- rope, 3) Neanderthals, 4) early anatomically lateral thickness of the last and second- modern H. sapiens from the Pleistocene, and to-last ribs and the first and fourth caudal 5) recent anatomically modern H. sapiens vertrebrae at their midpoints, the maximum from the Holocene. All the data from Pleisto- width of the mandibular corpus at dp, in cene fossils were generously provided by S. the pigs and MI in the armadillos, the total Nawrocki, who compiled them from pub- maximum width of the maxilla at M1, the lished measurements (for details see Naw- length of the mandible, the maximum medio- rocki, 1991). The comparative data on Holo- lateral width of the zygomatic arch at its cene modern human populations comes from midpoint, and the buccolingual and mesio- several published sources which are indi- distal dimensions of M' and dp, in the pigs cated in Table 2. and MI in the armadillos. In all cases, measurements were made twice and averaged; RESULTS wherever relevant, measurements were also Comparisons of cortical robusticity in averaged from the left and right sides. experimental animals Comparisons of bone thickness of various Before comparing the differences in vault characters in the exercised and control ani- thickness between the exercised and control mals these data were analyzed using Stat- animals, it is useful to examine the variation View 4.1, mostly as Mann-Whitney U-tests in robusticity elsewhere in the cranium and to avoid assuming that the data were nor- postcranium of the experimental subjects. mally distributed. The power of the U-test Table 3 summarizes the metrical variation is limited because of the necessarily small for tooth size, body mass, and several cranio- sample size for the pig experiment (three facial sites that are likely to generate and pairs). Consequently, a few of the P values or withstand masticatory forces. Tooth di- suggest a high degree of significance but do mensions are included for comparison be- not satisfy conventional 95%)degree confi- cause crown dimensions are highly heritable dence limits. No statistical analysis was pos- and form prior to eruption (Garn et al., sible for the armadillo data as there was only 1965). As one might expect for the geneti- one pair of subjects. Note, however, that the cally identical armadillos and the inbred sib- armadillos used in the experiment were ge- ling pigs, dental dimensions are statistically netically identical twins kept in exactly the indistinguishable between the two groups. same conditions with the exception of their In addition, several features of the mandible levels of daily exercise. It is reasonable to and maxilla, whose growth and shape are interpret any differences between them as known to be strongly influenced by mastica- CRANIAL VAULT THICKNESS 223

TABLE 2. Cranial vault thickness of fossils from TABLE 2. Continued the genus Homo (from Nawrocki (1991) unless otherwise indicated) Bregma Parietal eminence Fossil (mm) (mm) N Bregma Parietal eminence Fossil (mm) (mm) N Gibraltar 7.0 9.5 1 Neanderthal 7.5 10.0 1 H. erectus Amud 1 9.0 8.0 1 Olduvai Hominid 9 - 10.0 1 Tabun 1 4.0 5.0 1 Olduvai Hominid 12 10.0 7.5 1 Shanidar 1 - 8.0 1 Sale 8.0 6.4 1 Shanidar 2 - 8.2 1 Ternifine - 9.0 1 Shanidar 4 - 8.1 1 Trinil 9.0 9.0 1 Shanidar 5 - 9.0 1 Sangiran 2 8.8 11.0 1 Vindija 204 - 8.3 1 Sangiran 3 10.5 8.5 1 Vindija 26 1 5.9 - 1 Sangiran 4 5.5 9.3 1 Vindija 293 - 9.1 1 Sangiran 10 8.0 11.0 1 Krapina C - 8.5 1 Sangiran 12 9.0 9.5 1 Krapina D 8.5 7.0 1 Sangiran 13 10.0 - 1 Krapina E - 7.5 1 Sangiran 17 9.0 - 1 Krapina 16 7.0 7.5 1 Sangiran 18 11.0 10.5 1 Krapina Par. 5 - 7.5 1 Hexian - 13.5 1 Krapina Par. 20 - 8.0 1 Lantian 16.0 - 1 Krapina Par. 21 - 6.0 1 Zhoukoudian 1 - 5.0 1 Krapina Par. 32 - 8.0 1 Zhoukoudian 2 7.5 9.75 1 Krapina 34.1 - 7.0 1 Zhoukoudian 3 9.5 11.25 1 Early modern humans Zhoukoudian 4 10.5 10.0 1 Cro-Magnon 1 8.0 9.5 1 Zhoukoudian 5 9.0 - 1 Cro-Magnon 2 - 6.5 1 Zhoukoudian 6 - 10.5 Cro-Magnon 3 - 5.5 1 Zhoukoudian 10 10.0 9.25 1 Predmosti 1 - 5.0 1 Zhoukoudian 11 7.0 16.0 1 Predmosti 3 7.5 6.0 1 Zhoukoudian 12 9.5 8.75 1 Predmosti 4 - 6.0 1 Ngandong 1 9.0 - 1 Predmosti 9 4.5 6.0 1 Ngandong 3 10.0 8.0 1 Predmosti 10 - 5.0 1 Ngandong 5 7.0 - 1 Predmosti 14 - 7.5 1 Ngandong 6 12.0 - 1 Cotte de St. Brelade 12.0 6.0 1 Ngandong 9 9.0 - 1 Oberkassel 1 10.0 5.0 1 Ngandong 11 11.0 - 1 Oberkassel 2 8.0 7.0 1 Early archaic humans Mladec 1 3.5 3.5 1 Xujiayao 10 8.5 12.6 1 Mladec 5 - 7.0 1 Xujiayao 6 6.5 7.0 1 Mladec 6 - 8.0 1 Xujiayao 4.5 9.0 10.8 1 Mungo 1 4.5 - 1 Maba 7.0 9.0 1 Mungo 3 7.0 - 1 Dali - 12.0 1 Kow Swamp 16 7.5 - 1 Melka Kunture - 15.0 1 Keilor 9.0 - 1 Bod0 1 13.0 - 1 Tandou 2 8.0 - 1 Ndutu - 11.5 1 Wadjak 1 8.0 - 1 Flonsbad 12.0 12.0 1 Omo 1 8.0 - 1 Omo 2 9.0 - 1 Boskop - 14.0 1 Kabwe 8.8 9.5 1 Lukenya - 12.0 1 Laetoli 18 12.0 - 1 KRM41658 7.0 7.0 1 Bilzingsleben 9.0 - 1 Qafzeh 3 - 10.0 1 Swanscombe 7.0 10.5 1 Qafzeh 5 - 8.0 1 Steinheim 6.0 6.5 1 Qafzeh 6 - 8.0 1 Petralona 10.5 9.0 1 Qafzeh 7 - 5.0 1 Fontechevade 5 7.0 8.0 1 Qafzeh 9 6.0 6.5 1 Neanderthals Skhul2 - 10.5 1 Ehringsdorf 1 - 10.0 1 Skhul4 - 10.0 1 Ehringsdorf 2 - 17.0 1 Skhul5 7.5 4.5 1 Kulna - 11.3 1 Skhul9 - 11.0 1 La Chapelle 5.5 7.5 1 Holocene humans La Quina 5 5.0 5.3 1 French',5 5.4 5.7 200 La Quina 13 - 7.1 1 Yuendumu2 7.4 4.3 20 La Ferrassie 6.0 7.0 1 Africans' 6.7 7.7 64 Monte Circeo - 7.0 1 Chinese' 6.4 6.0 49 Le Moustier 6.0 6.8 1 Phillipines' 6.7 5.7 22 SPY 1 8.0 9.5 1 Hebrides' 7.0 6.9 16 SDV 2 7.0 9.0 1 American, ~hite~,~5.8a6 3.6 445 American, - 2.9 32 lTweisselmann, 1941. Belgian',' 5.3 5.7 200 'Brown et al., 1979. 3Tbdd, 1924. Byblos' 7.4 6.9 13 'Roche, 1953. Sialk Copper Age' 7.1 6.4 10 Postindustrial population. HastiBre' 6.4 6.7 24 CMeasuredat vertex. French Neolithic' 6.8 6.7 15 Sialk Iron Age' 5.7 5.3 20 Susa' 6.5 6.6 14 Palmyra' 6.4 72 19 224 D.E. LIEBERMAN

TABLE 3. Comparative dental, cranial and bodvweight data from exDerimenta1 oi~sand armadillos Pigs Armadillos Controls (n = 3) S.D. Runners (n = 3) S.D. P' %diff Control Runner %diff Dental characters (mm) MLb-1 9.9 0.1 9.8 0.1 0.51 1.0 1.6 1.6 0.0 M, m-d 13.5 0.1 13.6 0.1 0.51 0.2 2.2 2.3 4.5 dp4 b-1 7.9 0.1 8.0 0.1 0.01 1.3 na na na dp? m-d 18.2 0.2 18.3 0.3 0.83 0.5 na na na Maxillary and mandibular characters (mm) Corpus width at dpml 15.9 0.1 15.8 0.1 0.28 0.6 4.2 4.2 0.0 Maxilla width at MI 49.5 0.2 49.3 0.5 0.66 0.5 19.4 19.2 1.0 Mandible length 156.3 1.5 154.7 1.2 0.19 1.0 68.0 67.9 0.1 Mid-zygomatic width 6.4 0.2 6.3 0.3 0.99 0.6 1.6 1.5 4.6 Body weights (kg) Experiment start 4.1 0.2 4.1 0.5 0.51 1.5 0.25 0.25 0 Experiment middle 10.8 0.9 10.9 0.9 0.99 0.8 1.1 1.13 2.7 Experiment end 28.0 1.04 26.5 2.3 0.51 5.6 2.88 2.79 3.2 'Mann-Whitney U-Test. tory forces, were also statistically indistin- of inertia, I, around the mediolateral (x) and guishable in size between the exercised and dorsoventral (y) axes and the minimum (Imln) control animals of both species. These char- and maximum (Imax)moment areas. As Table acters include the width of the mandibular 4 indicates, the limb bones of the exercised corpus at dp, in the pigs and M, in the arma- animals are not only significantly thicker in dillos, the width of the maxilla at M1, the cortical area and linear dimensions, but they length of the mandible, and the mediolateral also have a significantly greater distribution thickness of the zygomatic arch at its mid- of mass around the neutral axis of the bone. point (see Herring, 1993). As with the teeth, With the exception of the size of the medul- such metrical similarities are predicted lary cavity, the high degree of statistical sig- among the subjects because of their high ge- nificance for these measures is to be expected netic similarity and because they ate the given the well-established principle that same diets. Animals fed artifically softened bones model in response to higher levels of diets develop maxillae and mandibles that habitual strain, particularly in growing ani- are significantly less tall, wide, and deep mals (Lanyon, 1984; Biewener et al., 1986; than those of controls fed hard but otherwise Lieberman and Crompton, in press). Similar nutritionally identical diets (Corruccini and responses to strain have been documented Beecher, 1982;Bouvier and Hylander, 1981). in many species, including pigs (Woo et al., Table 3 also summarizes longitudinal weight 1981) and turkeys (Lanyon, 1984; Lanyon et data. It is important to note that there were al., 1986;Rubin and Lanyon, 1985;Loitz and never any statistically significant differ- Zernicke, 1992) as well as humans (Jones et ences in weight for the pigs or armadillos al., 1977; Ruff et al., 1994). during the experiment. While the exercised Perhaps the most surprising metrical dif- animals may have had slightly more bone ferences in cortical robusticity between the and muscle mass in their limbs (see below), exercised and control animals are found in these differences were apparently offset by the cranial vault. Table 5 presents CVT data lowered fat deposits. at five locations: the center of the frontal Measures of cortical robusticity in the along the midsagittal axis, bregma, the pari- limbs of the exercised and control animals, etal eminences, the maximum thickness of however, reveal predictable contrasts. Un- the nuchal region in the midsagittal plane, like gnathic characters, the weight-bearing and the anteromedial corner of the left su- limb bones of the exercised animals were praoccipital (where a strain gauge was significantly thicker than those of the con- attached in the pig). Despite the small sam- trols in both species. Table 4 summarizes ple sizes, there are highly significant con- several midshaft dimensions for the tibia in- trasts in thickness between the two groups cluding cortical area and the second moment at each location. On average, the vaults of CRANIAL VAULT THICKNESS 225

TABLE 4. Comparatiue midshaft tibia data from experimental pigs and armadillos Pigs Armadillos Controls (n = 3) S.D. Runners (n = 3) S.D. P' %diff Control Runner %diff Tibia midshaft I, (mm4) 843.7 21.7 1,523.3 222.8 0.05 80.5 107.0 224.0 209.3 I, (mm4) 1,506.7 158.9 2,130.0 185.2 0.05 41.4 38.9 65.1 67.4 Imin (mm') 820.3 26.3 1,480.0 198.0 0.05 80.4 114.0 230.0 201.8 Imax (mm') 1,53 0.0 148.0 2,173.3 198.6 0.05 42.0 31.7 59.7 88.3 Cortical area (mm2) 96.8 3.1 119.5 0.4 0.05 23.4 23.4 31.7 35.5 Medullary area (mm2) 22.9 1.0 25.3 3.6 0.28 10.5 2.0 3.3 60.7 ' Mann-Whitney U-Test.

TABLE 5. Comparatiue cranial uault, rib and uertebral data from experimental pigs and armadillos Pigs Armadillos Controls (n = 3) S.D. Runners (n = 3) S.D. P' %diff Control Runner %diff Cranial thickness (mm) Frontal, center, midsagittal 8.3 1.5 11.0 1.0 0.08 32.5 0.74 0.93 24.3 Bregma 6.6 0.8 8.8 0.4 0.05 33.3 0.66 0.80 21.2 Euryon 6.2 1.2 7.5 0.5 0.19 21.0 1.30 1.50 15.4 Supraoccipital 8.1 0.9 10.9 0.5 0.05 34.6 1.07 1.31 22.4 Nuchal (maximum) 10.2 0.2 12.6 1.1 0.05 23.5 0.34 0.50 47.0 Ribs, midpoint dimensions (mm) Fifteenth rib dorsoventral 3.8 0.1 4.6 0.3 0.05 20.9 na na na Fifteenth rib mediolateral 8.8 0.2 5.8 0.5 0.83 1.7 na na na Eighth rib dorsoventral na na na na na na 2.5 2.9 16.8 Eighth rib mediolateral na na na na na na 4.6 5.3 15.2 Sixteenth rib dorsoventral 3.3 0.1 4.0 0.2 0.05 20.3 na na na Sixteenth rib mediolateral 4.5 0.2 5.8 0.2 0.05 28.2 na na na Ninth rib dorsoventral na na na na na na 2.2 2.4 9.0 Ninth rib mediolateral na na na na na na 4.2 4.4 4.8 Caudal vertebrae (mm) C1 mediolateral 24.5 2.7 31.2 1.5 0.05 27.1 10.5 11.8 12.4 C1 dorsoventral 10.5 0.5 11.8 1.0 0.13 12.7 9.4 10.2 8.5 C4 mediolateral 14.4 0.4 16.2 0.9 0.05 12.3 9.2 10.5 14.1 C4 dorsoventral 6.4 0.2 7.2 0.2 0.05 11.5 8.6 9.2 7.0 ~ __ ' Mann-Whitney U-Test. the exercised animals are about 28% thicker lateral and dorsoventral thickness of the last than those of the controls. Analysis of para- and penultimate ribs at their midpoints. sagittal sections through the parietal bones These characters make no significant contri- just lateral to the sagittal suture demon- bution to supporting body mass during loco- strated the vault bones of the two groups motion because in neither species is the tail to be histologically indistinguishable. The used when running, and the most caudal ribs greater vault thickness of the exercised ani- do not extend to the ventral aspect of the mals is entirely a consequence of more lamel- abdomen. With the exception of one dimen- lar deposition, particularly from the peri- sion (the mediolateral width of the fifteenth cranium. rib in the pigs), the contrasts in thickness Finally, Table 5 also presents the differ- between the two groups for these characters ences in cortical robusticity between the ex- are highly statistically significant. Many of ercised and control animals for several loca- the differences in robusticity between the tions elsewhere in the postcranium that are exercised and control animals therefore ap- unlikely to experience high levels of strain pear to be systemic. from habitual masticatory or locomotor activities. These characters are the mediolat- Strain. Table 6 summarizes the results of era1 and dorsoventral thickness of the first the strain gauge analyses. The strain levels and fourth caudal vertebrae and the medio- in the tibia for the pig and the armadillo, 226 D.E. LIEBERMAN

TABLE 6. Strain gauge data from experimental pigs and armadillos

Tibia strains (p~) Cranium strains (p~) Subject Activity n Tension Compression Shear Tension Compression Shear

Pig Running 3.0 mph 24 1,243.0 t- 446.0 -561.3 2 172.4 1,804.3 I602.5 73.3 t- 21.9 -64.2 t 11.4 137.4 2 28.4 Armadillo Running 1.0 mph 21 1,260.5 t 193.7 -230.8 i 78.0 1,491.3 t- 181.2 na na na Pip Chewing hard food 23 na na na 86.7 t- 45.2 -86.0 t 29.3 172.7 i 51.0

approximately 1,500-2,000 WE of shear, are in CVT at bregma and the parietal emi- very similar to each other and to those docu- nences, respectively, over time for the adult mented for the tibia of other medium-sized fossils and populations listed in Table 2. The mammals (Biewener and Taylor, 1986; Lie- data from the 17 recent human populations berman and Crompton, in press). Such levels are included only as sample means in order of strain are well within the range of mini- to avoid statistical biases from the large mum effective strains that Frost (1986) pre- number of recent human skulls available for dicts will elicit the type of bone modeling analysis. Figure 1A,B clearly indicates that responses reported above. In contrast, the while there has been a slight, general trend peak levels of strain measured in the pig towards thinner skulls over the last 1million vault during locomotion and chewing, both years, the enormous degree of variability well under 200 WE of shear, are roughly an prior to the Holocene precludes any strong order of magnitude lower than the levels of relationship between time and vault thick- strain measured in the tibia, despite the ness at bregma and only a slight trend at the close location of the gauge to both the nuchal parietal eminences. Including the Holocene crest and to the posterior fibers of the m. populations, r2 is 0.13 for bregma and 0.28 temporalis. According to most models (Frost, for the parietal eminences; excluding the Ho- 1986; Carter, 1987; Beaupre et al., 1990; locene populations, r2for a least squares re- Turner, 1992), such low strains might be ex- gression analysis is 0.19 for bregma and 0.35 pected to induce resorption. In vivo strain for the parietal eminences. levels were not measured in the armadillo If one breaks down CVT by taxon, as in cranium, but it is unlikely that locomotion Figure 2A,B, it is apparent that there are generated high strains in the vault given few simple statistically significant differ- their head posture (like pigs, armadillos do ences between Late Pleistocene hominid not move their heads in synchrony with their taxa in terms of vault thickness (see also limbs as they run). The greater CVT in the Nawrocki, 1991). A single factor analysis of exercised animals is, therefore, either a re- variance (ANOVA) reveals that at bregmaH. sult of a lower threshold at which loading erectus fossils are thicker than Neanderthals induces bone growth in the cranial vault than in the tibia or from systemic responses (P < 0.001) but not other early archaic to exercise unrelated to strain (discussed humans and that neither the Neanderthals below). nor early anatomically modern humans have significantly thicker vaults at bregma Comparative hominid data than Holocene populations. Non-Neander- In addition to comparing differences in thal archaic humans, however, do have sig- cortical robusticity in laboratory animals as nificantly thicker vaults at bregma than described above, hypotheses concerning the Pleistocene anatomically modern humans processes by which CVT develops in humans (P < 0.02). At the parietal eminence, H. can also be tested with data from the homi- erectus fossils are not significantly thicker nid fossil record and from recent humans. I than either Neanderthals or other archaic studied how CVT within the genus Homo humans, and Neanderthals are not signifi- has changed over time and tested whether cantly thicker than Late Pleistocene early CVTs differ significantly between taxa, and/ anatomically modern humans. It is interest- or between humans with contrasting subsis- ing to note, however, that early archaic hu- tence strategies. Figure 1A,B plots changes mans have significantly thicker parietal em- CRANIAL VAULT THICKNESS 227

18 16 A 14 h E 0 0 A WE 12

6 4 2 0 200 400 600 800 1000 A Time (Ka)

18 16 A A 14 12 10 A 8 6 4 2 0 200 400 600 800 1000 B Time (Ka) Fig. 1. Change in thickness over time at bregma (A) and the parietal eminences (B) for taxa within the genus Homo. See text for discussion of measurements and Table 2 for specimens included. Least square regression for bregma is 7.23 + O.O6x,r2 = 0.19; least square regression for the parietal eminences is 6.7 + 0.007~.r2 = 0.35. 228 D.E. LIEBERMAN inences than both Neanderthals (P < 0.008) Upper Palaeolithic technologies, nor are and Pleistocene modern humans (P< early anatomically modern humans from the 0.0002) and that Pleistocene modern hu- Pleistocene consistently more gracile in mans have significantly thicker parietal em- terms of vault thickness than contemporary inences than Holocene modern humans or earlier archaic human populations such (P < 0.02). as the Neanderthals. A thin cranial vault is clearly not a derived The hypothesis that differences in vault character unique to modern Homo sapiens thickness among hominid taxa are attribut- relative to other taxa within the genus able to local responses to loading from either Homo. Subsistence strategy, however, does running or chewing is neither rejected nor appear to have an effect on CVT. Among the strongly supported. The neurocranium must post-Neolithic Holocene populations from experience some strain from the contrac- Europe (and North America) and the Middle tions of the m. temporalis that attaches along East listed in Table 2, seven are preindus- much of its surface, presumably generating trial farmers and four are postindustrial (as inferiorly and laterally directed tensile indicated in Table 2). A Student’s t-test com- forces in the bones and sutures of the vault. parison of these two groups reveals that the However, the strain levels recorded in the populations of preindustrial farmers have pig vault in this experiment-less than significantly thicker vaults (P < 0.02) at 150 p.~of shear-are low in comparison with both bregma and the parietal eminence than strains recorded elsewhere in their postcra- the more recent, industrial populations. nium (Lieberman and Crompton, in press). Brown et al. (1979) have also documented If the differences in CVT within taxa noted significant differences in cranial vault thick- above are a consequence of strain-induced ness between recent sedentary Australian osteogenesis, then one must conclude that aborigines and earlier, more mobile aborigi- lower amounts of force are necessary to gen- nes. These data do not support the null hy- erate bone growth in the pericranium and pothesis that CVT does not vary by subsis- endocranium. One might predict the cranial tence strategy. Pleistocene hunter-gatherers vault to have a much higher safety factor tend to have thicker skulls than Holocene than other parts of the skull or postcranium farmers (not enough data, however, are because of the high cost of breaking a currently available to compare Holocene bone in this region (Hylander and Johnson, hunter-gatherers with farmers), and recent 1992; Hylander et al., 1992; Lieberman and postindustrial populations tend to have thin- Crompton, in press). ner skulls than preindustrial farming popu- This safety factor hypothesis deserves fur- lations. ther investigation but must remain tenta- tive in the absence of stronger support and/ DISCUSSION The fact that genetically identical armadillos and sibling pigs can develop differ- Fig. 2. Differences in vault thickness at bregma (A) ences in vault thickness equivalent to those and the parietal eminences (B) for taxa within the genus Homo. Boxes show the standard error, with a vertical between recent and Pleistocene humans line at the mean; the tails show the standard deviation (20-30% in most regions of the vault) sug- from the mean, and open circles show the range. Ac- gests that CVT in these species is probably cording to a single factor ANOVA, Pleistocene modern not a very genetically heritable character. humans are not significantly thinner than Neanderthals The hypothesis that recent humans have at bregma or the parietal eminences. Holocene modern humans, however, are significantly (P < 0.05) thinner thin cranial vaults because of a relaxation at the parietal eminences than all earlier taxa, including of selection to maintain thick vaults is, Pleistocene modern humans. At bregma, Holocene mod- therefore, unlikely. In addition, the fossil re- ern humans are significantly (P < 0.05) thinner than H. cord provides further evidence against the erectus and non-Neanderthal archaic humans but are not significantly thinner than either the Neanderthals adaptation hypothesis. As discussed above, or Pleistocene modern humans. See text for discussion there is no apparent reduction in CVT asso- of measurements and Table 2 for the fossils included in ciated with the transition from Middle to each taxon. CRANIAL VAULT THICKNESS 229

H. erectus - O 0 (n=25) a+ 9.0 (s.d. 2.2) Early Archaic humans - (n=14) OCCIII-IO 6.7 (s.d. 1.4) Neanderthals - (n=13) 0 47bO 7.4 (s.d. 2.0) Pleistocene modern humans - (n=17) O4Ib O 6.5 (s.d. 0.7) Holocene modern humans I @-ID) (n= 16) 1 1 1 1 1

9.7 (s.d. 2.3)

H. erectus O (n=2 1) 10.3 (s.d. 2.4) Early Archaic humans (n=13) O4rI-l O 8.3 (s.d. 2.2) Neanderthals (n=30) Omo 7.4 (s.d. 2.5) Pleistocene modern humans (n=27) OaII- O 1 5.8 (s.d. 1.4) Holocene modern humans (n=17) J.mol , I , I , 4 6 8 10 12 14 16 18 B Vault Thickness at Parietal eminence (mm.) 230 D.E. LIEBERMAN or more rigorous testing. For instance, the by circulating hormones, it should not be osteoblasts in different regions of the skele- surprising that exercise can induce systemic ton may be activated by different strain osteogenic activity. The hormone most likely threshold levels. Preliminary studies, how- to mediate this phenomenon is GH, which ever, indicate that neurocranial osteoblasts the anterior pituitary synthesizes in re- appear to be less sensitive to strain than in sponse to thyroid hormone as well as to corti- the postcranial osteoblasts. Rawlinson et al. costeroids. In most mammals, the anterior (1995), for example, found that in vitro pituitary secretes GH every 3-4 h in a pulsa- strains between 100 p~ and 1,000 p~fail to tile fashion, with the largest peaks occurring stimulate osteoblasts from rat calvarium but during sleep. GH, which has numerous func- do stimulate cells from ulnae. Moreover, tions, is critical for inducing systemic bone other recent studies in several mammal spe- growth by stimulating the synthesis of IGF-I cies, including primates, have reported simi- and IGF-I1 (the somatomedins) that mediate larly low strain levels in the sutures and many of its effects on cells, as well as by bones of the neurocranium. Behrents et al.’s directly activating DNA synthesis in a vari- (1978) in vivo study of strain around the ety of related skeletal cell types including sagittal suture in macaques during maxi- fibroblasts, chondroblasts, and myoblasts mum, bilateral contraction of the temporalis (see Daughaday, 1989). muscles recorded maximum tensions of ap- GH, moreover, is strongly linked to exer- proximately 180 pe across the suture and cise and may, therefore, account for the over- 100 p~just lateral to the suture in the pari- all differences in bone thickness between the etal bone. Sugimura et al. (1984) recorded exercised and control animals. Even moder- on average 59 PEof tension along the sagittal ate levels of exercise significantly increase suture in dogs. Iwasaki (1989) consistently secretions of GH secretion (Borer, 1980; recorded peak shearing strain of less than Poehlman and Copeland, 1990). The magni- 70 p~ in the temporal and parietal regions tude of the GH response is related to work of the cranial vault in adult and infant ma- intensity, so that subjects who exercise regu- caques during chewing. In addition, it ap- larly and strenuously have higher circulat- pears that several regions of the cranium ing GH levels than more sedentary individu- other than the vault do not experience the als (Lasarre et al., 1974; Naveri, 1985; Van high strains that are common within the Helder et al., 1986; Felsing et al., 1992).The face, mandible, and most of the postcranium. effects of higher levels of GH on systemic The low strain levels reported above in the bone growth, including the cranial vault, are pig occipital are consistent with those re- well established in mammals. For example, corded by Hylander and colleagues in the exogenous GH injected in mice causes them supraorbital torus in macaques and baboons to develop thicker, longer crania and postcra- (Hylander et al., 1991; Hylander and John- nia, particularly at the site of muscle inser- son, 1992). The browridge is, thus, more tions (Vogl et al., 1993). Humans with defi- likely to be a developmental consequence of cient levels of growth hormone prior to the spatial separation of the splanchnocran- skeletal maturity (e.g., hypopituitarism) de- ium and neurocranium (Shea, 1985; Picq velop dwarfism unless they are treated with and Hylander, 1989; Ravosa, 1988, 1991) regular GH injections (Brook, 1989). Simi- than a beam to provide resistance against larly, individuals with abnormally high GH twisting or bending forces (e.g., Endo, 1966; levels develop gigantism (e.g., acromegaly). Greaves, 1985; Russell, 1985). Acromegalics are not only taller than indi- Finally, the hypothesis that systemic hor- viduals with normal GH levels, but they mones are the primary cause of variation in also develop extreme cortical thickening CVT is not rejected but is instead partially throughout the skeleton, including the cra- supported by this study. As demonstrated nial vault; in contrast, individuals with GH above, the pigs and armadillos who exercised deficiencies have abnormally thin cortical daily experienced more systemic cortical bone development in the skull and postcra- growth than control animals throughout the nium (Brasel et al., 1965; Randall, 1989; skeleton. Since all bone growth is mediated Pirinen et al., 1994). CRANIAL VAULT THICKNESS 23 1

It is, therefore, reasonable to suggest that on approximately neonatal archaic humans higher GH levels induced by physical activ- are available for La Ferrassie 4b (Heim, ity could cause the differences in CVT be- 1982) and for Hortus 1 and Hortus lb (de tween the exercised and nonexercised ani- Lumley, 1973)-a11 of which fall within the mals reported above and that similar range for modern European values of CVT differences in activity levels in humans also at birth, with the one exception of the maxi- produce variation in CVT. As demonstrated mum thickness of La Ferrassie 4b, which is above, the most important factor that influ- 0.5 mm thicker (Minugh-Purvis, 1988). ences CVT in human taxa appears to be sub- There are no fossils of H. erectus neonates, sistence strategy since agriculturalists have but the only known skull of an H. erectus thinner vaults than hunter-gatherers (al- infant, the Modjokerto fossil, has cranial though it is difficult to factor time out of this walls that are very thin, “in the parietal re- comparison) and postindustrial populations gion up to 3 mm., and elsewhere even less” have significantly thinner vaults than farm- (LeGros Clark 1978:99).The Modjokerto fos- ing populations. As one would predict, Ruff sil thus falls within the range of modern val- and coworkers (e.g., Ruff and Hayes, ues for infants between 2 and 3 years, which 1983a,b; Ruff, 1992; Ruff et al., 1993, 1994) is likely to be a reasonable estimate of its have documented a similar relationship be- age. tween subsistence strategy and limb bone If, as these limited data suggest, CVT at cortical bone robusticity that appears to be birth is similar for all taxa of the genus relatively independent of taxonomy. Ruff et Homo, then most of the variance in robusti- al. (1993:21)note that “earlymodern H. sapi- city must develop later during childhood. In ens are closer in shaft robusticity to archaic particular, one might expect the pattern of H. sapiens than they are to recent humans.” systemic growth differences in cortical bone In other words, human populations who get thickness to occur in populations with high less habitual exercise because of technologi- levels of exercise prior to skeletal maturity cal advances not only have thinner weight- when the systemic effects of GH on growth bearing limb bones but also have less cortical are most profound. After adolescence, GH robusticity throughout the skeleton. While receptors at many sites of bone growth are nutritional and/or general health factors as- blocked, leading to growth plate fusion and sociated with these subsistence shifts un- to a general deceleration of overall growth doubtedly occurred that must also be consid- levels (Isaakson et al., 1982; Armstrong, ered3,these would probably have competing, 1988). This age-effect hypothesis remains to opposite effects on CVT since better nutri- be tested, but there are several lines of evi- tion leads to increased rather than decreased dence which provide some support. Brown rates of bone growth in the postcranium and et al. (1979) demonstrated that adult Aus- cranium (Israel, 1978). tralian aborigines tend to have thicker The hypothesis that systemic cortical ro- vaults than adult Americans of European busticity is primarily a consequence of the descent because their vaults grow at a faster effects of exercise suggests several predic- rate during adolescence. For example, the tions that can be tested using the fossil rec- rate of growth of CVT at vertex (Fig. 3A) is ord. First, if bone thickness is a character about the same in Australian aborigines and with low heritability that is not subject to Americans until about the age of 12, when strong selective pressures, then neonatal there is a substantial decrease in growth and young H. erectus and archaic humans rate in Americans but not Aborigines. Simi- should have cranial vault and postcranial lar differences in growth rate after childhood bones that are as thin as those of recent but prior to adulthood also appear to account humans. Measurements of vault thickness for the thicker skulls of Neanderthals and perhaps H. erectus, perhaps beginning as early as 3-4 years of age. Minugh-Purvis (1988) compared parietal thickness at ’Early farmers may have had decreased levels of nutrition compared with contemporary hunter-gatherers, but these popu- bregma and the parietal eminences in Nean- lations are not considered here. derthals and recent humans divided into 232 D.E. LIEBERMAN broad age classes (Fig. 3B). Although the estimates give H. erectus and archaic hu- sample sizes are unavoidably small, they mans significantly thicker vaults relative to clearly indicate that the differences in CVT body mass than those predicted by the best- between the two groups are the result of a fit line for all anthropoids. Although Gauld continued rapid growth rate after infancy in (1993) has suggested that H. erectus and ar- Neanderthals when recent humans begin to chaic humans may have had heavier body experience a slower growth rate. Zollikofer masses than their postcranial dimensions et al. (1995) reached similar conclusions predict, such an argument clearly cannot based on their computerized reconstruction apply to recent Australian aborigines and of the Devil's Tower Neanderthal infant. It is probably not to early anatomically modern significant to note that Pleistocene modern humans whose body masses were almost cer- humans from Europe, like the Neanderthals, tainly within modern human ranges. Per- also had rates of cortical thickening more haps we should not ask why recent humans rapid than recent Europeans (Minugh- tend to have thin skulls but why H. erectus Purvis, 1988). In recent populations, vault and Pleistocene humans have such thick thickness increases very slowly after the age skulls. The answer may be that hunter-gath- of roughly 20 in both sexes until approxi- erers, from H. erectus until recent times, ex- mately 50 or 60 years of age (Todd, 1924; perienced relatively longer durations of sus- Young, 1957; Israel, 1973; Adeloye et al., tained exercise relative to body mass than 1975). other anthropoids or recent humans. Alter- The development of cortical robusticity for natively or additionally, robusticity in non- the rest of the postcranium in archaic hu- stressed bones such as the vault might be a mans appears to mirror the apparent pat- consequence of the longer duration of skele- tern in the vault. Ruff et al. (1994) analyzed tal immaturity in the genus Homo, which cortical bone robusticity in two juvenile Ne- would tend to increase the effects of exer- anderthal postcranial skeletons for which cised-induced systemic growth prior to there is reasonable age data, Teshik Tash adulthood. and La Ferrassie 6. While both have thick In order to test the hypothesis that overall cortices, the ratio of their femoral cortex area differences in robusticity are a consequence standardized to femoral length is only of GH-mediated epigenetic responses to ex- slightly above that of recent humans. Some ercise, particularly exercise that occurs prior linear measurements of neonatal postcrania to skeletal maturity, it will be necessary to from La Ferrassie, however, are greater than integrate controlled developmental studies those of recent Europeans (Heim, 1982). with measurements of in vivo levels of GH Since young archaic humans appear to be and other osteogenic hormones. In addition, only slightly more robust than young ana- it is not known to what extent systemic bone tomically modern humans, one cannot reject growth can occur during adulthood. Most the hypothesis that there may be some ge- studies on the effects of exercise on adult netic component to intertaxon variations in bone modeling and remodeling have focused systemic robusticity, but the above data sug- on Haversian remodeling, calcium exchange, gest that the source of the variation appears and changes in limb bone cortex in diaphyses to be mostly nongenetic. and trabecular architecture in epiphyses The hypothesis that CVT is primarily a (see Currey, 1984; Martin and Burr, 1989). consequence of systemic bone growth is also Finally, we do not know whether or how dif- supported by the allometric relationship be- ferent regions of the skeleton respond to tween CVT and body mass. Gauld (1992) and disimilar strain levels. Nelson and Gauld (1994) have shown that The phenomenon of systemic bone growth CVT tends to scale positively with body mass in response to exercise merits further consid- in anthropoid primates, in spite of the fact eration for several reasons. From a clinical that body mass clearly does not transmit perspective, such research may help us to through the skull. Gauld (1992,19931,more- evaluate approaches to preventing or treat- over, has shown that while recent, thin- ing osteoporosis in humans. While there is skulled humans fit this interspecific regres- consensus that weight-bearing exercise has sion well, postcranially based body mass important local effects on bone growth be- CRANIAL VAULT THICKNESS 233 A 8 r

6 E -4E x Aborigines (Male and Female) >2 0 Recent Europeans (Female) 0 Recent Europeans (Male)

I I I I 0.00 6.0 12.0 18.0 24.0 Age (years) B

0 38 1 m WE 'I) $6 - c Y @ 24- 0 Neanderthals, parietal eminence b - f Neanderthals, bregma -u 0 Recent humans, parietal eminence - 0 $2 80 Recent humans, pregma I I I I I I I

Fig. 3. A: Longitudinal increases in vault thickness at vertex in Australian Aborigines and Europeans. Data from Roche (1953) and Brown et al. (1979). B: Cross-sectional data on increases in vault thickness at bregma and the parietal eminence for Neanderthals and recent humans. Data from Minugh-Purvis (1988). cause of strain, its importance for systemic (Lane et al., 1986). In addition, the phe- growth has not been widely appreciated. The nomenon of exercise-induced systemic bone high rates of osteoporosis among old people growth is important for interpreting certain in industrial populations may result from a aspects of the hominid fossil record. For one, sedentary lifestyle not only during adult- it is clear that CVT is an inappropriate char- hood but also during childhood and adoles- acter to use for phylogenetic studies of the cence, with lasting consequences on cortical relationships of recent humans because it is bone thickness. Exercise-related systemic neither a derived character of anatomically cortical robusticity may, therefore, help ex- modern humans nor highly heritable (Lie- plain why the single best predictor for a sub- berman, 1995). The above data suggest that ject's likelihood to develop severe osteoporo- early modern and archaic humans, despite sis is how active shehe was is early in life their anatomical contrasts, may not have 234 D.E. LIEBERMAN been as different in terms of overall exercise on cortical bone structure in macaques (Mucuca mu- as other studies have concluded (e.g., Lieber- Zatta). J. Morphol. 167:l-12. Brace CL (1979) The Stages of Human Evolution, 2nd man, 1993; Lieberman and Shea, 1994). The ed. Englewood Cliffs, NJ: Prentice-Hall. thickness of the cranial vault may be similar Brasel J, Wright J, Wilkins L, and Blizzard R (1965)An in archaic and modern hunter-gatherers be- evaluation of seventy-five patients with hypopitu- cause, as a subsistence strategy, it demands itarism beginning in childhood. Am. J. Med. 383: frequent and regular exercise from a rela- 484498. Brook CGD (1989) Growth hormone deficiency: Fea- tively early age. In other words, it appears tures, assessment, and management. In L.J. DeGroot that young hunter-gatherers, regardless of (ed.): Endocrinology, Vol. 1,2nd ed. Philadelphia: W.B. their anatomical modernity, obtained lots Saunders, pp. 351-361. of exercise. Brown T, Pinkerton SK, and Lambert W (1979) Thick- ness of the cranial vault in Australian aboriginals. Arch. Phys. Anthropol. in Oceania 14:54-71. ACKNOWLEGMENTS Buchannan CR, and Preece MA(1992) Hormonal control I am especially grateful to A.W. Crompton of bone growth. In BK Hall (ed.): Bone, Vol. 6: Bone and F.A. Jenkins, Jr., for their support and Growth. Boca Raton: CRC Press, pp. 53-89. Carter DR (1987) Mechanical loading history and skele- help with the experiments and to S. Naw- tal biology. J. Biomech. 20:1095-1109. rocki for generously sharing data from his Coon CS (1962) The Origin of Races. New York: Alfred 1991 dissertation and for his detailed com- A. Knopf. ments on an earlier draft of this paper. Addi- Corruccini RS, and Beecher RM (1982) Occlusal varia- tional thanks go to A. Glazer and T. Ower- tion related to soft diet in a nonhuman primate. Sci- ence 218:74-76. kowicz for running the animals and to D. Corruccini RS, and Beecher RM (1984) Occlusofacial Pilbeam, R. Trivers, B. Wood, and two anony- morphological integration lowered in baboons raised mous reviewers for their suggestions. Fund- on soft diet. J. Craniofac. Gen. Dev. Biol. 4:135-142. ing was made possible by grants from the Currey J (1984) The Mechanical Adaptations of Bones. Milton Fund (Harvard Medical School) and Princeton: Princeton University Press. Dally JW,and Riley WF (1978) Experimental Stress the American School of Prehistoric Re- Analysis, 2nd ed. New York: McGraw Hill. search. Daughaday WH (1989) Growth hormone: Normal synthesis, secretion, control, and mechanisms of action. LITERATURE CITED In LJ DeGroot (ed.): Endocrinology, Vol. 1, 2nd ed. Philadelphia: Saunders, pp. 318-329. Adeloye A, Kattan K, and Silverman FN (1975) Thick- Duterloo HS, and DH Enlow (1970)A comparative study ness of the normal skull in the American blacks and of cranial growth in Homo and Macacn. Am. J. whites. Am. J. Phys. Anthropol. 43:23-30. Anat. 127:357-368. Armstrong PF (1988)Cell kinetics and control of physeal Endo B (1966) Experimental studies on the mechanical growth. In HK Uhthoff and JK Wiley (eds.): Behavior significance of the form of the human facial skeleton. of the Growth Plate. New York: Raven Press, pp. 233- J. Facult. Sci. Univ. Tokyo, Section 3, 5:l-106. 236. Enlow DH (1990) Facial Growth, 3rd ed. Philadelphia: Beaupr6 GS, Orr TE, and Carter DR (1990)An approach Saunders. for time-dependent modeling and remodeling- theoretical development. J. Orthop. Res. 8:651-661. Felsing NE, Brasel JA, and Cooper D (1992) Effect of Behrents RG, Carlson DS, and Abdelnour T (1978) In low and high intensity exercise on circulating growth uiuo analysis of bone strain about the sagittal suture hormone in men. J. Clin. Endocrinol. Metab. 75: in Macaca mulatta during masticatory movements. J. 157-162. Dent. Res. 57:904-908. Frayer DH, Wolpoff MH, Thorne AG, Smith FH, and G Bhanba JK (1961) Longitudinal and cephalometric Pope (1993) Theories of modern human origins: The roentgenographic study of face and cranium in rela- palaeontological test. Am. Anthropol. 95:14-50. tion to body height. J. Am. Dent. Assoc. 63:776-799. Frost HM (1969)Tetracycline based histological analysis Biewener AA (1992) Overview of structural mechanics. of bone remodeling. Calcif. Tissue Int. 3:211-217. In A Beiwener (ed.): Biomechanics-Structures and Frost HM (1983) Bone histomorphometry: Choice of Systems: A Practical Approach. Oxford: Oxford Uni- marking agent and labeling schedule. In HM Frost versity Press, pp. 1-20. (ed):Bone Histomorphometry. Boca Raton: CRC Press, Biewener AA, and Taylor CR (1986)Bone strain: A deter- pp. 38-52. minant of gait or speed? J. Exp. Biol. 123:383-400. Frost HM (1986) The Intermediary Organization of the Biewener AA, Swartz SM, and Bertram JEA (1986)Bone Skeleton. Boca Raton: CRC. Press. modeling during growth: Dynamic strain equilibrium Frost HM (1987) Bone “mass” and the “mechanostat”: in the chick tibiotarsus. Calcif. Tissue Int. 39:390- A proposal. Anat. Rec. 219:l-9. 395. Frost HM (1988) Vital biomechanics: Proposed general Borer KT (1980) Characteristics of growth-inducing ex- concepts for skeletal adaptations to mechanical usage. ercise. Physiol. Behav. 24:713-720. Calcif. Tissue Int. 42:145-156. Bouvier M, and Hylander WL (1981)Effect ofbone strain Garn SM, Lewis AB, and Kerewsky R (1965) Genetic, CRANIAL VAULT THICKNESS 235

nutritional and maturational correlates of dental de- of bone loss during calcium insufficiency by controlled velopment J. Dent. Res. 44:228-242. dynamic loading. Calcif. Tissue Int. 38209-216. Gauld SE (1992) An Allometric Study of Anthropoid Lassarre C, Girard F, Durand J, and Raynaud J (1974) Cranial Bone Thickness: Implications ofbody size esti- Kinetics of human growth hormone during submaxi- mation in early hominid species. Ph.D. Dissertation, ma1 exercise. J. Appl. Physiol. 37:826-830. University of California, Los Angeles. Ann Arbor: LeGros Clark WE (1978)The Fossil Evidence for Human Michigan Microfilms. Evolution, 3rd ed. Chicago: University of Chicago Gauld SE (1993) Body size of Asian Homo erectus: Esti- Press. mation based on prediction models utilizing measures Lieberman DE (1993) The rise and fall of hunter-gath- of cranial bone thickness (abstract). Am. J. Phys. An- erer seasonal mobility: The case of the southern thropol. Suppl 16:93. Levant. Curr. Anthropol. 34r599-631. Greaves WS (1985) The mammalian post-orbital bar as Lieberman DE (1995) Testing hypotheses about recent a torsion-resisting helical strut. J. Zool. SOC.London human evolution from skulls: Integrating develop- 207:125-136. ment, function, and phylogeny. Curr. Anthropol. Groves CP (1989) A Theory of Human and Primate Evo- 36:159-197. lution. Oxford: Oxford University Press. Lieberman DE, and Crompton AW (in press) Responses Heim JL (1982) Les Enfants Neandrthaliens de la Fer- of vertebrate bones to stress. In CR Taylor and E rassie. Paris: Masson. Weibel (eds.): Optimality in Biology: Controversies Herring SW (1993) Epigenetic and functional influences Over Symmorphosis. Cambridge: Cambridge Univer- on skull growth. In J Hanken and BK Hall (eds.): The sity Press. Skull, Vol 1. Chicago: University of Chicago Press, Lieberman DE, and Shea JJ (1994) Behavioral differ- pp. 153-206. ences between archaic and modern humans in the Hylander WL (1986) In vivo bone strain as an indicator Levantine Mousterian. Am. Anthropol. 96:300-332. of masticatory force in Macaca fascilularis. Arch. Oral Loitz BJ, and Zernicke RF (1992) Strenuous exercise- Biol. 31:149-157. induced remodeling of mature bone: Relationships be- Hylander WL, and Johnson KR (1992) Strain gradients tween in vivo strains and bone mechanics. J. Exp. in the craniofacial region of primates. In Z Davidovitch Biol. 17O:l-18. (ed.): The Biological Mechanisms of Tooth Movement. Lumley, MA de (1973)Anteneanderthaliens et Neander- Columbus, OH: Ohio State University College of Den- thaliens du Bassin Mediterranean Occidental Euro- tistry, pp. 559669. peen. Etudes Quaternaires, Memoires 2. Marseilles: Hylander WL, Picq P, and Johnson KR (1991) Mastica- Universite de Provence. tory-stress hypotheses and the supraorbital region of Martin RB (1992)A theory of fatigue damage accumula- primates. Am. J. Phys. Anthropol. 86:l-36. tion and repair in cortical bone. J. Orthop. Res. 10: Hylander WL, Picq P, and Johnson KR (1992) Bone 818-825. strain and the supraorbital region of primates. In DS Martin RB, and Burr DB (1989) Structure, Function, Carlson and SA Goldstein (eds.) Bone Biodynamics in and Adaptation of Compact Bone. New York Raven Orthodontic and Orthopaedic Treatment, Vol. 27 Ann Press. Arbor MI: Craniofacial Growth Series, Center for Hu- Minugh-Purvis N (1988) Patterns of Craniofacial man Growth and Development, pp. 315-349. Growth and Development in Upper Pleistocene Homi- Isaakson OGP, Jansson JO, and Gause IAM (1982) nids. Ph.D. Dissertation, University of Pennsylvania. Growth hormone stimulates longitudinal bone growth Ann Arbor: Michigan Microfilms. directly. Science 216:1237-1238. Moss ML (1954) Growth of the calvaria in the rat. Am. Israel H (1973) Age factor and the pattern of change of J. Anat. 98:191-204. craniofacial structures. Am. J. Phys. Anthropol. 39: Moss ML, and Young RW (1960) A functional approach 111-128. to craniology. Am. J. Phys. Anthropol. 18:281-292. Israel H (1978) The fundamentals of cranial and facial Muir WM, and Hubbell JAE (1989) Handbook of Veteri- growth. In F Falkner and JM Tanner (eds.): Human Growth, Vol. 2: Postnatal Growth. New York Plenum, nary Anesthesia. St. Louis: CV Mosby. Naveri H (1985) Blood hormone and metabolic levels pp. 357-380. Iwasaki K (1989)Dynamic responses in adult and infant during graded cycle ergometer exercise. Scand. J. Lab. monkey craniums during occlusion and mastication. Clin. Invest. 45:559-603. J. Osaka Dent. Univ. 23:77-97. Nawrocki SP (1991) A Biomechanical Model of Cranial Jones HH, Priest JD, Hayes WC, Tichenor CC, and Na- Vault Thickness in Archaic Homo. Ph.D. Dissertation, del A (1977)Humeral hypertrophy in response to exer- SUNY Binghamton. cise. J. Bone Joint Surg. [Am.] 59:204-208. Nawrocki SP (1992) Cranial thickness and skull bio- Kennedy GE (1985) Bone thickness in Homo erectus. J. mechanics in Archaic Homo, Am. J. Phys. Anthropol. Hum. Evol. 14:699-708. Suppl. 14:127. Lane JM, Healy JH, Vigorita VJ, and Werntz JR (1986) Nelson AJ, and Gauld SE (1994) Covariance relationship Orthopedic management of osteoporosis: Effects of nu- between measures of cranial and post-cranial bone trition and exercise on the skeleton. In HK Uhthoff thickness and body mass. Am. J. Phys. Anthropol. (ed.): Current Concepts on Bone Fragility. Berlin: Suppl. 18~151-152. Springer-Verlag, pp. 429-447. Ohtsuki F (1977) Developmental changes of the cranial Lanyon LE (1984) Functional strain as a determinant bone thickness in the human fetal period. Am. J. Phys. for bone remodeling. Calcif. Tissue Int. 36:556- Anthropol. 46:141-154. 561. Picq PG, and Hylander WL (1989)Endo’s stress analysis Lanyon LE, Rubin CT, and Banst G (1986) Modulation of the primate skull and the functional significance of 236 D.E. LIEBERMAN

the supraorbital region. Am. J. Phys. Anthropol. erectus and the existence of the species in Africa and 79:393-398. Europe. Cour. Forsch. Ins. Senckenberg 69:131-143. Pirinen S, Majurin A, Lenko HL, and Koski K (1994) Stringer CB (1987) A numerical cladistic analysis for Craniofacial features in patients with deficient and the genus Homo. J. Hum. Evol. 16:135-146. excessive growth hormone. J. Craniofac. Genet. Dev. Stringer CB (1988) Homo sapiens. In I Tattersall, E Biol. 14:144-152. Delson, and W Van Couvering (eds.):The Encyclopedia Poehlman ET, and Copeland KC (1990) Influence of of Human Evolution and Prehistory. New York: Gar- physical activity on insulin-like growth factor I in land, pp. 267-274. healthy younger and older men. J. Clin. Endocrinol. Sugimura T, Inada J, Sawa S, and Kakudo Y (1984) Metab. 71:1468-1473. Dynamic responses ofthe skull caused by loss of occlu- Randall RV (1989) Acromegaly and gigantism. In LJ sal force J. Osaka Dent. Univ. 18:29-42. DeGroot (ed.): Endocrinology, Vol. 1,Znd ed. Philadel- Thorne AG, and Wolpoff MH (1981) Regional continuity phia: W.B. Saunders, pp. 330-350. in Australian Pleistocene hominid evolution. Am. J. Ravosa MJ (1988) Browridge development in Cerco- Phys. Anthropol. 55:337-349. pithecidae: A test of two models. Am. J. Phys. Anthro- Todd TW (1924) Thickness of the white male cranium pol. 76:535-555. Anat. Rec. 27245-256. Ravosa MJ (1991)Interspecific perspective on mechani- Turner AS (1992)Functional determinantsof bone struc- cal and nonmechanical models of primate circumorbi- ture: Beyond Wolff‘s law of bone transformation. tal morphology. Am. J. Phys. Anthropol. 86:369-396. Bone 13:403-409. Rawlinson CF, Mosley JR, Suswillo RFL, Pitsillides AA, Twiesselmann F (1941) Methode pour l’evaluation de and Lanyeon LE (1995) Calvarial and limb bone cells l’epaisseur des parois cranienne. Bull. Musee Royal in organ and monolayer culture do not show the same #Histoire Nat. Belgique 17:l-33. early responses to dynamic mechanical strain. J. Bone Van Helder WP, Casey K, Goode RC, and Radomski WM Miner. Res. 10:1225-1232. (1986) Growth hormone regulation in two types of Roche AF (1953) Increase in cranial thickness during exercise of equal oxygen uptake. Eur. J. Appl. Phys- growth. Hum. Biol. 25:81-92. iol. 55:236-239. Rubin CT, and Lanyon LE (1984) Regulation of bone Vogl C, Atchley WR, Cowley DR, Crenshaw P, Murray J, formation by applied dynamic loads. J. Bone Joint and Pomp D (1993) The epigenetic influence of growth Surg. 66:397402. hormone on skeletal development. Growth Dev. Rubin CT, and Lanyon LE (1985) Regulation of bone Aging 57t163-182. mass by mechanical strain magnitude. Calcif. Tissue Wainright SA, Biggs BA, Currey JD, and Gosline JM Int. 37:411-417. (1976) Mechanical Design in Organisms. Princeton: Ruff CF (1992) Biomechanical analysis of archaeological Princeton University Press. human skeletal samples. In SR Saunders and A Kat- Washburn SL (1947)The relation of the temporal muscle zenberg (eds.): Skeletal Biology of Past Peoples: Re- to the form of the skull. Anat. Rec. 99239-248. search Methods. New York: Wiley-Liss, pp. 37-58. Webb S (1990) Cranial thickening in an Australian hom- Ruff CB, and Hayes WC (1983a) Cross-sectional geome- inid as a possible palaeoepidemiological indicator. Am. try of Pecos Pueblo femora and tibiae-a biomechani- J. Phys. Anthropol. 82:403-411. cal investigation: I. Method and general pattern of Weidenreich F (1941)The brain and its r61e in the phylo- variation. Am. J. Phys. Anthropol. 60:359-381. genetic transformation of the human skull. Trans. Am. Ruff CB, and Hayes WC (198310) Cross-sectional geome- Phil. SOC.31:321-442. try of Pecos Pueblo femora and tibiae-a biomechani- Weidenreich F (1943) The skull of Sinanthropus peki- cal investigation: 11. Sex, age, and side differences, nensis: A comparative study of a primitive hominid Am. J. Phys. Anthropol. 60:383400. skull. Palaeontologica Sinica, NS D 1O:l-484. Ruff CB, Trinkaus E, Walker A, and Larsen CS (1993) Williams PL, Warwick R, Dyson M, and Bannister LH Postcranial robusticity in Homo I: Temporal trends (1989) Gray’s Anatomy, 37th edition. Edinburgh: and mechanical interpretation. Am. J. Phys. Anthro- Churchill Livingstone. pol. 91:21-53. Wolpoff MH (1980)Paleoanthropology. New York: Knopf. Ruff CB, Walker, A, and Trinkaus E (1994) Postcranial Woo SLY, Kuei SC. hiel D, Gomez MA, Hayes WC, robusticity in Homo 111: Ontogeny. Am. J. Phys. An- White FC, and Akeson WH (1981) The effect of pro- thropol. 93:35-54. longed physical training on the properties of long bone: Russell MD (1985) The supraorbital torus: “A most re- A study of Wolff‘s law. J. Bone Joint Surg. 635’80-787. markable peculiarity.” Curr. Anthropol. 26t337-360. Yamada K, and Kimmel DB (1991) The effect of dietary Shea BT (1985) On aspects of skull form in African apes consistency on bone mass and turnover in the growing and orangutans, with implications for hominid evolu- rat mandible. Arch. Oral Biol. 36:129-138. tion. Am. J. Phys. Anthropol. 68:329-342. Young RW (1957) Postnatal growth of the frontal and Skinner HC, and Nalbadian J (1975) Tetracyclines and parietal bones in white males. Am. J. Phys. Anthro- mineralized tissues: Review and perspectives. Yale J. pol. 15:367-386. Biol. Med. 48:377-393. Zollikofer CPE, Ponce de Leon MS, Martin RS, and Sperber GH (1989) Craniofacial Embryology, 4th ed. Stucki P (1995) Neanderthal computer skulls. Na- London: Wright. ture 375283-285, Stringer CB (1984) The definition of the species Homo