Abstract: China is rich in Neolithie human skeletons as well as Paleolithic human remains. The former is even more abundant in number. Nevertheless, the evolutionary aspeets of these skeletons and their morphological differences from modern ones have not drawn sufficient attention. The present paper aims to examine the nature of morphologieal evolution in Holocene human populations of China. In literature were described about 20 Chinese Neolithie sites yielding more or less human crania, at most 30—40, at least 1—2. Among them, southern and northen populations were separated, and the North China group was further subdivided into eastern and western subgroups.
In order to involve as many groups and variables as possible, two techniques were used for preparing the data-sets. These are 1)combination of sample groups and 2) estimation of missing values.
There are two reasons for eombining sample groups. First of all, in some cases, there were more than one group sampled from the same population. Usually, one can not know which is more representative. Combining them is a better way than choosing at random, as long as there exists reasonable similarity among the samples. Secondly, combining is useful when any sample group does not contain sufficient number of samples, or has all variables represented. The member groups can be mutually complementary.
In this study, after combining samples there were still some missing values in the data-sets. And these missing values involved such important groups as the Zengpiyan group and such important variables as eranial base length and facial length. In order to maintain both these groups and these variables in the statistical study, the missing values had to be estimated. Stepwise regression analysis was used for that. The estimated values have been proved reliable by both the F-test and the correlation test.
Cluster analysis, factor analysis, diseriminant analysis and two way variance analysis were used in this study. The results produced by these four techniques are quite consistent.
1. Cluster analysisIn the dendrogram of the male groups based on Penrose's shape distances, Neolithic north and south groups, and modern north and south groups respectively form their own blocks. One exception is the Gansu-Henan Neolithie group, which was excavated by the Swedish geologist and archaeologist J. Andersson and studied by the Canadian anthropologist D. Blaek. The two regional branehes of the same period join with each other respeetively. Then, the Neolithie branch and the modern branch link at a higher level (greater distance) . This shows that there is a bigger difference in terms of time than of region. In other words, the secular change is generally more significant than the geographical variation. In addition, in the Neolithic branch and the modern branch, the artificially deformed groups are specially separated as expeeted.
2. Factor analyssIn the factor analyses of both the male and the female groups, almost all of the linear variables have large positive loadings for thefirst factor, especially facial length, bizygomatie breadth, eranial base length, basi-bregmatie height, minimum frontal breadth and nasal breadth. The first factor, therefore, can be called the size factor.
For the second factor, upper facial height and nasal height have large loadings. This is the facial height factor.
The coordinates of the first two factors marking the positions of the groups studied produced the same results as shown above. The two axes distinguish temporal and regional groups respectively. The first factor (X-axis) distinguishes between Neolithic and modern populations. The second factor (Y-axis) distinguishes between north and south populations. Once again, the Gansu-Henan Neolithie group is the only exception.
The aceuracy of the ehronology for at least some of the samples of the Gansu-Henan group has been questioned based on a review of the Neolithie cultural context. The above analyses support this doubt.
To compare the male and the female Hemudu specimens is also interesting. Each is represented by only one sample. The male sample (M23 is bigger than the other male groups on the average, while the female one (M 17) is smaller than the other female groups. It is not likely that the tremendous differences of these two samples reflect a extremely big sexual dimorphism of the population. They seem to be merely extreme samples in their respective sexes.
It is not proper to include the Gansu-Henan group and the Hemudu specimens in the further analyses.
3. Discriminant analysisThe purpose of the discriminant analyses is to detect the significance of the difference between the temporal groups (i. e. Neolithic and modern group) or the regional groups (i. e. north and south group) in each variables. The analyses of the male groups obtained exactly the same results as did the faotor analysis: there were very significant temporal differences in eranial base length, facial length, basi-bregmatie height, nasal breadth, bizygomatic breadth and minimum frontal breadth and very significant or significant regional differences in nasal height and upper facial height.
4. Two way variance analysisThe time sequence consists of early Neolithic, late Neolithie and recent times. The region sequence consists of northern, central-eastern and southern China. 9 male groups were chosen to fill in the tablet of the two sequences. The analyses of 8 variables show somewhat gradual changes over time in basi-bregmatic height, nasal breadth, minimum frontal breadth, eranial length and bizygomatic breadth, and somewhat gradual changes over region in upper facial height, nasal height and eranial length.
Orbital dimensions were not included in the above analyses, because it was noticed that significant errors had been made in measuring orbital height and breadth, and because not all groups contained the mean values of these two measurements. To test whether these two variables are still useful in further discussion, another seattergram was drawn with two axes representing orbital height and orbital breadth respectively. It is shown that there exist differences among the four bloeks: the Neolithic groups have wider and lower orbitals than the modern ones, while the northern groups have higher orbitals than the southern ones. This indieates that errors of measurement of the orbital dimensions do not affeet comparative study to sueh an extent as to obseure the differences in question over time and region.
Based on the above analyses, the secular change and the geographical variation in question can be summarized as follows.
Chinese Neolithic residents had larger erania than the modern population. This was reflected in cranial base length, facial length, eranial height, nasal breadth, bizygomatic breadth, minimum frontal breadth and orbital breadth. Their upper facial height and orbital height were slightly lower than in the modern population.
As to geographical variation, in both Neolithic and recent times northern Chinese inhabitants have higher upper facial height, orbital height and nasal height than southern Chinese ones.
Generally speaking, the seeular ehange is more significant than the geographical variation.
The present study is an attempt to consider seeular change and geographical variation at the same time. Obviously, this is often necessary. For instance, sereral Neolithic groups of Shaanxi Yang-shao Culture were characterized by relatively low facial height and relatively prominent prognathism ete. At first sight, they seemed to represent anomalous geographical variation. Taking secular change into account, however, one finds a somewhat inereasing tendency toward facial height and a somewhat decreasing tendeney to prognathism. Therefore, it is not reasonable to regard this Neolithic population as a separate variant of the southern Mongoloid racial stock. The statistical analyses illustrate that to study these two aspects of changes of populations at the same time is not only necessary but also possible if only these changes happened in different eharacters.
The statistical analyses were carried out under the instruction of Prof. Kazuro Hanihara during the author's stay in the University of Tokyo as visiting scholar. Gratitude is expressed to him and his colleagues, especially Dr. Makiko Kouchi and Mr. Kiyotaka Koizumi for their help. Thanks are also due to Prof. Wu Rukang for his encouragement throughout this study.

Key words: Neolithic population; Modern population; Crania; Metric traits