Jiangzhai and BanPo (Shaanxi, PRC): new ideas from old bones

by Mary Jackes and Gao Qiang

The skeletal biology of the Chinese Neolithic has received very little attention outside China. Black (1928) compared Neolithic skeletal remains from northern and western China with recent skeletons: a superbly detailed analysis of post-cranials. The Neolithic material he worked on had been discovered by Andersson (1973 [1934]) and excavated by Zdansky (ibid.: 166) in the Yellow River valley. One Chinese researcher (Yan) published during the 1960s and 1970s, but the mass of work in China has appeared since about 1980. In general, the publications have been confined to descriptions and measurement of skulls with, more recently, use of multivariate analyses (e.g. Zhang et al. 1982; Chen 1989). There has been some consideration of the division of the Chinese into northern and southern groups, a division considered to have great time depth. Howells (1983, 1993) proposed that the modern North Chinese population was already firmly in place by the Neolithic period. Our research suggests that the Neolithic population of North China was heterogeneous and certainly not all identical with modern mainland East Asians, and that differences cannot be ascribed simply to a north/south division of the East Asian population.

In the recent past Chinese archaeologists placed a marked emphasis on the excavation of Neolithic sites. During the 1950s and the Cultural Revolution, huge Neolithic villages were uncovered. Such excavations brought to life the early achievements of the peasantry. Thus, when vestiges of feudalism were being destroyed, workers were putting enormous efforts into the excavation, preservation and reconstruction of villages such as BanPo, which contained 45 houses and has been partially preserved. The pottery, together with the bone and stone artifacts from the village are magnificent and certainly merit the construction of the BanPo Museum near Xi'an.

Several hundred skeletons were excavated at BanPo, but only nine are contained in the museum, all in sealed cases and unavailable for study. Some skeletons were sent to Beijing and measurements of 13 skulls, 35 mandibles and 6 femora were published in 1960 (Yan et al. 1960). The most accessible publication in the west comprises a reprinting of the original cranial measurements (Wu and Zhang 1985). These remnants are all that is left of 248 individuals, mostly primary extended burials, a very few of them secondary or multiple burials. The youngest children, 73 of them, were placed separately in urns.

BanPo is the best known among the many sites in the area of the junction of the Yellow River and its major tributary the Wei River, representing early agriculture and permanent settlements in Northern China. This is the "nuclear area" where Chinese civilization arose and from which it spread, and the Yangshao Neolithic period is often considered to mark the beginning of this civilization. The particular combination of loess soils, river valleys and adjacent mountains and plains led to perfect conditions for the intensification of millet agriculture.

Other references to the Northern Chinese Neolithic in western osteological publications are included in studies by Brace (Brace and Hunt 1990; Li et al. 1991), using a specific set of measurements of the mid-facial region. The earlier paper refers to up to 33 skulls held in Beijing, no doubt including BanPo, but clearly from other sites as well, while the 1991 paper refers to skulls from the Neolithic of Xi'an, measured by GQ. These skulls were excavated from the site of Jiangzhai, not BanPo.

Jiangzhai was a huge and well-preserved site a few kilometres from BanPo on a terrace above a tributary of the Wei River, itself the tributary of the Huang Ho or Yellow River. Both BanPo and Jiangzhai date from 6,000 to 7,000 cal BP (Chang 1992), and are thus contemporaneous, as well as contiguous.

Jiangzhai was discovered in 1972, and over the next eight years, during 11 seasons of excavation, over 17,000 square metres of surface was excavated (Gong 1988).

The earliest levels (Stage I) include the complete layout of a Neolithic village with 120 houses, over 200 hearths and more than 300 large storage pits, as well as animal pens, all separated from the area of kilns. Five groupings of houses were arranged around a central plaza of 4,000 square metres. The cemetery, with three distinct sections, was separated from the living and working areas by ditches and palisades. In each area of the cemetery there were around 26 urn burials and 50 single primary burials, though a few burials were secondary or multiple. Infants were also buried in funeral urns among the houses.

A later cemetery contained nearly 300 graves: mostly secondary, these were collective burials of about 20 individuals, both adults and children. The long bones in some of the multiple graves are described as having been laid carefully to the north and south of each skull, with the short bones laid on the east side of the skull. It seems that nearly 2,300 individuals were excavated from this stage (Stage II).

In all, around 3,000 skeletons were uncovered at Jiangzhai, but not all of the cemetery areas were excavated. Unfortunately, the remains conserved in the BanPo Museum today represent elements of no more than about 56 individuals, often just skulls, covering a long period of time, from the first village with individual burials to the Stage II multiple burials. It is possible that about 500 skeletons were originally saved from the cemeteries, but these were reburied in the museum grounds. All that remains is a publication which lists ages and sexes of the individuals excavated. The reassessments undertaken in 1993 (Appendix 1) indicate that some errors may have been made during age and sex assessment, the consulting anatomist having been resident in another city during the excavations (GQ, personal communication to MJ, 1993).

What can be said about a 1% sample? Gong (personal communication to MJ, 1993) has stated that the skeletons were collected absolutely at random. The representative nature of the retained skeletal sample is important because the bones present an extraordinary array of trauma, in both Stage I and Stage II. This trauma can best be attributed to interpersonal violence. As Smith (1996:85) has pointed out, it is not sufficient to have fractures to the forearm, of the type called 'parry' fractures, commonly attributed to violence: "A violent aetiology for mid-shaft fractures becomes more tenable when potentially corroborative craniofacial injury data are considered." The Jiangzhai cases of trauma comprise the following:

- Two right clavicles were broken, one of an old male (M238, Stage II) and one of a female (M112, Stage II).

- A depressed fracture on the frontal of a young female (M75.2, Stage I) is particularly interesting. The circular depression is consonant with the many bola stones found in the Jiangzhai excavations.

- A broken and badly healed malar in a male (M112:13, Stage II).

- One male (M84:10, Stage II) had a blow to the chin with damage to the teeth and the tempero-mandibular joint.

- A young adult female (M84:10, Stage II) had a broken nose with significant alteration to the shape of the entire area; two other females of the same age also had broken noses (M112:10, M216:11, both Stage II).

- There were two fractured left elbow joints, in a female adolescent (M162, Stage I) and an old male (M151, Stage I).

- The left ulna of a young male (M150, Stage I) had a parry fracture, caused by raising the left arm to ward off a blow to the face.

- The left ulna of another young male (M275, Stage I) had necrotic bone at the mid-shaft, probably the result of a wound in the overlying soft tissue of the forearm, strongly suggesting a parry wound.

Of the adolescents and adults examined, about one third had bone changes suggesting violence. There was no tibial periostitis; no lower limb fractures were evident; no broken bones in the feet were present. The lack of evidence for the trauma associated with agriculturalists sustaining trauma to their hands, feet or shins was notable. The trauma in the Jiangzhai sample appeared to result from violence, apart from one vertebral compression fracture (the male postcranials from M155).

There are, however, other vertebral columns with trauma of a completely different sort, not indicating violence or major trauma.

The depressions - Schmorl's nodes - caused in the vertebral end plates by the rupturing of intervertebral discs, provide clear evidence of minor trauma. While there are small lumbar vertebral end plate defects in elderly males (M150, 151, 149), the adolescent female M181 and especially the slightly older M161 have extensive noding, both lumbar and thoracic. Such trauma results from heavy work, and since the nodes were especially obvious in the vertebrae of young females at Jiangzhai, we can hypothesize that women undertook the carrying of loads or heavy agricultural labour in the fields.

Figure 1 plots 13C data for Neolithic sites published by Cai and Chou (1984) along with estimated values for Jiangzhai. No stable isotope values have been determined for Jiangzhai, because, although we selected representative material, we were not able to export bone samples for analysis. The estimated Jiangzhai values are derived from other sites of the period between 7,500 and 4,500 cal BP in the same area of Northern China, assuming that the 13C value for Jiangzhai would be -13‰ to -14‰. The values for areas in which millet was not contributing to the diet would have been -20‰, as shown for Baishi village, a site in Shantong, closer to the coast.

We can certainly make the assumption that all the Jiangzhai individuals were agriculturalists. The question arises because one individual from BanPo showed evidence of a diet which did not depend upon millet. Furthermore, the sample of 250 faunal remains retained from the Jiangzhai Phase I and II excavations shows that hunting was important and that only around one third of the animal remains were of domestic pig (Fu 1988). Spotted deer and hornless river deer made an important contribution to the diet, and the assumption is that less than 50% of the diet was made up of C4 plants and of pigs eating C4 plants (Neolithic domestic pig from Taosi had the same 13C values as human beings; Cai and Chou 1984: 952).

The caries rate was low according to the traditional view of a Neolithic population - for example, there was only a 7% caries rate in the lower molars. There were virtually no occlusal caries and almost all the caries observed were associated with impacted third molars. The rate of maxillary third molar agenesis (22%) or reduction was very high, and lower third molars were commonly impacted.

Most of the mandibular caries were directly attributable to these impacted third molars, and in some cases the lower second molars seem to have been lost as a consequence of third molar impaction.

But there was other premortem tooth loss which can be attributed to severe periodontal disease, assuming that the loss of the alveolar bone was so great that teeth lost attachment and simply fell out. This is most common with molars (Figure 2), but several jaws showed anterior tooth loss which could certainly not be interpreted as resulting from tooth evulsion.

Periodontal disease normally involves a build-up of plaque, bleeding of the gingiva, calculus below the gingiva, pocketing, recession of the alveolus and the presence of a number of micro-organisms. Clinical studies show that the height of the cemento-enamel junction above the alveolus is the best predictor of periodontal disease, and it is this "crestal bone loss" which leads to loss of attachment. A cemento-enamel junction height above the alveolus of 4 mm or more is established as an adequate predictor of the presence of periodontal disease. Modern Chinese have a high incidence of periodontal disease (Cao et al. 1990).

A soft diet leads to a build-up of plaque and this soft diet, in the absence of dental hygiene, tends to result in periodontal disease. The modern wheat-based diet in the Xi'an area is extremely soft, and periodontal disease is very evident in the farmers of today flocking into the city. It seems likely that the Neolithic millet-based diet was also relatively soft. The dental attrition scheme based on Smith's (1984) work, which had been effective in describing tooth wear in Mesolithic to mediaeval European samples (e.g. Lubell et al. 1994), was unsatisfactory for the Chinese material examined (Shaanxi Neolithic to Han Dynasty). Chinese Neolithic buccal crown height is often retained, despite fairly advanced attrition. The attrition involves a quite flat grinding away of surface features, so that fissures and fovea are gone. A thin layer of enamel still covers the dentin until an advanced age, suggesting that the diet was soft relative to that of Neolithic diets in other parts of the world. It is worth noting, however, that the Jiangzhai attrition seems to have occurred at a more rapid rate than that recorded for modern southern Chinese males, with a rice-based diet (Mo and Peng 1983) in which dentin exposure on first molars is delayed until the late twenties. Jiangzhai dentitions may have pinpoint dentin exposures on the first molars of young adolescents.

In summary, Jiangzhai is characterized by a diet of millet, sorghum, mulberries, nuts, with some pigs, hunted deer, and fish, and dental hygiene probably limited to toothpicks (we observed clear toothpick grooves). The data shown in Figure 2 gives ample evidence that the incidence of periodontal disease is very high, the diagnosis being based on:

1) marked recession of the alveolus around the roots of all the cheek teeth, especially upper molars;

2) alveolar bone which is markedly altered by inflammation of the overlying soft tissues;

3) deep pocketing especially around the lower molars;

4) tooth loss that seems to be related to extreme alveolar bone loss.

The average height of the cemento-enamel junction (CEJ) above the alveolar crest in many maxillary teeth of individuals over about 18 years of age approaches or exceeds 3 mm (Figure 2), the measurement being taken from the alveolar crest, not from the bottom of the pocket. The mean exceeds 4 mm for many maxillary teeth in older adults, and even in young adults the average CEJ height for the right second incisor is over 4 mm. These individuals would have a clinical diagnosis of periodontal disease in modern dental surgeries. The anterior dentition of almost all individuals is affected by bone changes and the diagnosis of periodontal disease is supported by the fact that the peaks of alveolar bone alteration accord with the peaks of cemento-enamel junction height and premortem tooth loss. Equivalent features are found on mandibular teeth (Figure 2), with older adult mean CEJ heights exceeding 4 mm for the right first premolar (P) and canine.

A highly-developed and stable farming economy seems likely. Why is there so much osteological evidence for interpersonal violence? Why the deep ditches and palisades which suggest that the Neolithic villages near Xi'an were strongly fortified?

Fu (1994) studied biological distance based on Jiangzhai cranial and dental non-metrical traits. The small sample size and the lack of comparative samples is unfortunate, since a paired study of analyses of diverse non-metrical traits and cranial measurements demonstrates that the former is much more powerful in establishing relationships among skeletal samples (Jackes and Lubell 1999).

For this paper we have recalculated and reanalyzed distance statistics on the Jiangzhai data (Fu 1994:45), together with that of 11 sets of data derived from Turner (1987, 1990). The dental traits considered, together with the incidence (k/N, where k is the number of teeth showing the trait and N is the number of teeth observed) determined by Fu (1994) are:
1. shovelling of the upper central incisor 4/7
2. double shovelling of the upper central incisor 5/7
3. single rooted upper first premolar 18/24
4. enamel extension of upper first molar 12/30
5. peg, reduced, congenital absence of upper third molar 13/30
6. deflecting wrinkle lower first molar 2/6
7. lower first molar with three roots 3/8
8. lower second molar with four cusps 8/23

The samples are: 1. Amur (see Turner 1987, Uichi, Tungus), 2. Mongol (see Turner 1990, Urga and Mongol 2), 3. Jiangzhai, 4. Anyang (royal tombs, 1990), 5. Northeast Siberia (1987, 1990, Chukchi etc.), 6. Lake Baikal (1987, 1990), 7. recent Japan (1987, includes Kanto), 8. South China (1987, Cantonese and general China), 9. Hong Kong (1987), 10. recent Thailand (1987), 11. Jomon (1987), 12. Ainu Hokkaido (see Turner 1990).

The method of analysis is one by which proportions of samples exhibiting a trait are expressed, not as percentages, but as theta (θ) values symmetrical around 0, such that 50% is expressed by a theta of zero. The data are thus transformed to approximate a normal distribution. Sjøvold (1977) has determined that the Anscombe formula is the best modification for calculation of theta, most accurately transforming the incidences of traits, except perhaps when incidences are extremely high or low, and stabilizing the variance well. The use of the Anscombe formula is accepted by zoologists using dental traits in distance studies (e.g. Suchentrunk et al. 1994, Suchentrunk and Flux 1996). Use of the Freeman and Tukey transformation provides slightly better variance stabilization when the sample size is very small and incidences are correspondingly low. In general the sample sizes here are adequate, but we have calculated both the Anscombe and the Freeman and Tukey transformations because the Jiangzhai data are only marginally adequate.

The summed divergences between two sites divided by the number of traits constitute the mean measure of divergence (MMD). According to Sjøvold (1977:23), the Anscombe formula introduces a slight dependence on sample size, but since the formula for the MMD utilizes a correction term (V) which serves to "ensure that the MMD is an unbiased estimate of the true population divergence, independent of sample size" (see Sjøvold 1977: 15,110), this is not of major importance.

De Souza and Houghton (1977) and especially Sjøvold (1977) have commented on the problems caused by incorrect or variant formulae in the literature (a pitfall difficult to avoid, compare formulae 1.8.1 and 2.6.1. in Sjøvold 1977). For this reason, we provide the formulae used here (Appendix 2), checked against Sjøvold's analyses for calculation accuracy. We are using bilaterally expressed traits and basing our analysis on the number of observable sides, because of the poor sample available from Jiangzhai. Ideally, under these circumstances, the variance formula should be modified to take into account the correlation between sides and the number of bilaterally observable individuals (Sjøvold 1977). Since this is impossible when using published data, we note only that all data have been treated in the same way and are thus comparable.

An MMD value that is more than twice its standard deviation is significant (Sjøvold 1977:17,30). Table 1 gives the Anscombe MMD values above the diagonal, and non-significant values (MMD-(2*SD) is negative) are printed in bold. Below the diagonal an approximation of the sample size is given by the total of the mean number of observations made in the comparisons of each pair of sites over the eight traits.

In Figure 3 we illustrate a dendrogram derived from hierarchical cluster analysis using Ward's method of the matrix of Z calculated with the Anscombe transformation. Z is discussed in both de Souza and Houghton (1977) and Sjøvold (1977), the latter warning against uncritical use of the statistic (pages 29-30).

Standardization based on MMD/SD has been suggested when diverse sample sizes are used (Sofaer et al.1986), and the tree obtained from standardized values based on the Anscombe formula is completely identical with that in Figure 3 (the tree derived from the Freeman and Tukey formula is also essentially identical). The concordance of the Z and standardized trees is not unexpected, given that the two values are highly correlated. We prefer, however, to use Z because this value is more highly correlated with the degree of significance, (MMD-(2*SD), called DI for short in Jackes 1988) than are the standardized values. Z may suitably be used to generate dendrograms, while DI is not appropriate.

Is it a true homogeneity of the samples which produces low values of MMD, DI and Z? Does the position of Jiangzhai in the dendrogram (Figure 3) and the strong similarity with Lake Baikal, based on the standardized MMD value, reflect no more than the sample sizes? Lake Baikal has very high variances, suggestive of poor sample sizes, and yet the Lake Baikal sample sizes are mostly well above the limit for k/N suggested by Sjøvold (1977:23), of 2/15 when using the Anscombe formula. With the data available to us we can do little more than to point out that the two most similar samples in the entire data set are likely to be Mongolia and Jiangzhai, followed by Hong Kong and Southern China, each pair insignificantly different based on Z as defined by de Souza and Houghton (1977). Z requires proof that the traits are independent, and is perturbed according to Sjøvold (1977:29) whenever observations have to be recorded as missing, but at the very least Z provides a more acceptable ranking of values, in terms of geography, than do the standardized values.

Heirarchical cluster analyses (using SPSS with between and within groups average linkage, Wards, centroid and median methods) were performed on values generated by both the Anscombe and Freeman Tukey transformations (MMD, MMD/SD, Z and MMD-(2*SD) values entered into matrices. In all of the several methods of analysis, Turner's distinction between Sundodonty and Sinodonty is well defined. In 18 of the 40 dendrograms Jomon falls into a cluster by itself, but the weight of the evidence is very slightly in favour of Jomon joining the Thai-Ainu pair as in Figure 3. Other groupings are quite stable: Recent Japan consistently joins the close pairing of South China and Hong Kong. NE Siberia and Amur form a pair, joining immediately, as in Figure 3, with the other northern samples in most cases.

We have also undertaken analyses based on the raw Anscombe theta values, utilizing average linkage (between groups), city block measure, and standardizing to a standard deviation of 1 by traits. The results from such an analysis make excellent sense, with Hong Kong and Southern China closest, joined by Recent Japan (as in Figure 3), then by a grouping of Mongolia and the Anyang royal tombs. The NE Siberia and Amur samples group and join, followed by Jiangzhai and then by Lake Baikal. The relationship of the Thai, Ainu and Jomon samples to each other and to the Sinodont grouping is as in Figure 3.

This result serves to underline that the close linking of Jiangzhai and Lake Baikal and their position in Figure 3 may result from their small sample sizes. Good data provide consistent and clear results; and, while poor data may be manipulated in a number of ways, it is very susceptible to the preconceptions of researchers, and should be viewed sceptically. There can be no doubt that Jiangzhai finds its partners among northern mainland Asian populations, but the exact relationship with the Anyang and the samples from the area of Ulan Bator may require further study.

Fu (1994) also analysed non-metrical cranial traits, but the paucity of publications providing comparative material restricted the value of study. Ossenberg of Queens University, Kingston, Ontario very kindly provided Fu with unpublished data on Japanese and Siberian non-metrical cranial traits and we have analyzed an Anscombe Z matrix on eight cranial traits which give consistently high incidence rates and sample sizes. The samples from the area of Lake Baikal and the Amur River cluster with the Ainu. Jomon groups with NE Siberia and, more distantly, with Jiangzhai. Japanese populations (Kanto and Kinki) cluster completely separately from the other samples. The two sets of data do not, then, provide comparable results and we can again say only that the Jiangzhai nonmetrical cranial data set suggests northern connections.

The available literature provides significantly more data for cranial metrical studies. As a matter of interest rather than proof, considering the small numbers provided by some samples, a series of analyses was undertaken on male cranial samples of six skulls or more, from Ulan Bator and Sakhalin in the north to Hainan Island in the south (see Appendix 3 for sources of data). Taking data from publications is, of course, not ideal. The choice of measurements was dictated by those available in the literature; fortunately they provided relatively uncorrelated variables, and in multivariate analyses the variables' highest correlations are well distributed across the discriminant functions (Table 2). Furthermore, the measurements chosen are standard and easy to take accurately. In several cases researchers have published separately on the same samples (e.g. the Shang sacrificial victims from Anyang, and Hainan and Atayal samples: Howells 1989; Pietrusewsky 1995; Han and Pan 1986) and this has allowed monitoring of whether data on these measurements can be taken from diverse published sources. In all test classification analyses, measurements by different researchers on the same samples fell together. The Jiangzhai measurements used here are concordant in their placement in all analyses with those for Jiangzhai Stages I and II reported by GQ in Gong (1988). In all other cases when similar or equivalent samples are used, no discrepancies are apparent.

The analysis illustrated here (Figure 4) is based on a grouping determined using Ward's method (squared Euclidean measure) cluster analysis of 47 samples of male skulls (Table 3). Five clusters were clearly differentiated in the resultant dendrogram, The two major clusters discriminate between some Neolithic samples (32, 35, 36, 37, 38, 46; see Appendix 3) and all other samples. The larger grouping (cluster 1-4) is subdivided into, firstly, Jomon, Yayoi and Kofun Japan (cluster 1 on Figure 4) and, secondly, all other samples (apart from those in cluster 5) both mainland Asia and mediaeval and recent Japan. A noteworthy feature is the tight clustering of modern mainland Asians, north and south (cluster 3) , in comparison with the great dispersion of the Neolithic and Bronze Age samples (clusters 4 and 5, as well as samples 31 and 34 from cluster 3). Stepwise discriminant function analysis was performed with group membership determined by a five cluster solution and the first two canonical functions explain 88% of the variance.

Martin (1957-66) measurements 17 (cranial height), 9 (minimum frontal breadth) and 5 (basi-cranial length) contribute most to the discrimination. Cluster means (Table 4) demonstrate the differences among the clusters. Cluster membership is recorded in Appendix 3.

The first and second discriminant functions were plotted in order to illustrate the dispersion produced by this type of analysis of the cranial measurements ( Figure 4). The great heterogeneity of the Neolithic mainland sites is immediately obvious, BanPo and Yongdenxian (at b on Figure 4) clustering with modern East Asian populations, but others, with Jiangzhai the outlier, dispersing in a quite separate grouping. The relationship of the Anyang skulls remains enigmatic, with the small tombs 1 and 3 in cluster 3, overlapping at a on Figure 4, while the small tomb 2 lies at b and the sacrificial victims, though in cluster 4, being represented by the solid circle lying to the right of b.

The existence of Neolithic heterogeneity may well explain a contradiction in the western literature. Howells (1983), using 11 cranial measurements and seven of the northern Neolithic samples and the Anyang Bronze Age (sacrificial victims) sample used here, concluded that the northern Chinese Neolithic population was homogeneous and "apparently indistinguishable from modern Chinese". On the other hand, Brace (Li et al. 1991) clearly separated the Chinese Neolithic (the Jiangzhai sample) from the Anyang sacrificial sample and from general mainland Asia. Without a number of good samples from northern Chinese Neolithic villages this contradiction cannot be clarified, but we suggest that both conclusions are correct based on Figure 4, and that the answer to this problem lies in the heterogeneity of the northern Chinese Neolithic populations.

We have then, based on the small samples retained from the Neolithic villages of the Wei valley, evidence that soft diets were well established by around 7000 cal BP, and that the fortified settlements were required in reponse to extreme levels of violence, resulting perhaps from movements of populations into the river valleys. Movement of population from the north is a possibility, given the evidence presented here and speculation that innovations are likely to have been introduced in the more challenging environment of Inner Mongolia (Bettinger et al. 1994), rather than in the ideal conditions of the Wei valley. The evidence for population heterogeneity is slight, but sufficient to provide the impetus for a new approach to the treatment of human skeletal remains on archaeological sites, Neolithic to Bronze Age, in the Wei and Yellow River Valleys. The first step must be the analysis of skeletal material from the Peiligang Neolithic, predating the Yangshao Neolithic of BanPo and Jiangzhai.

We must hope that the rapid development in the area of Xi'an allows for the discovery of new cemetery sites, and that these will be excavated with the greatest care. There are many questions of interest. Some of the questions have been considered by Chinese archaeologists. For example, why did burial practices alter so significantly and can skeletal biologists determine the basis of burial in the different sections of a cemetery? Gao and Lee (1993) summarized some of the work on this in considering Shijia, a site excavated in 1976 and dated at 5700 cal BP, which contained some 730 individuals, all but three in 40 secondary burials. Few skeletons were retained and there has been no publication as yet on the osteology. It will be an enormous loss to anthropology and archaeology if the potential of human skeletal biology to provide information on the past is overlooked as China races into the next century. Sites like Jiangzhai, BanPo and Shijia could have provided detailed information on genetics and population relationships within and between villages from 5500 cal BP to 7000 cal BP.