Dr. T. Tóth szerk.: Studia historico-anthropologica (Anthropologia Hungarica 21. Budapest, 1990)

At this point we could accept this configuration of series as a well-founded one but the indefinite results yielded by the Mahalanobis distance made us to carry out some further examinations. The number of variables was reduced to four (MARTIN 1,8, 17, 48) and we have redone the previous series of analysis. Quite independently from the data transformation, distance measure and the group distance used the same results occurred again. The dendrograms produced by Mahalanobis distance with Minkowski (Fig. 2e) and Maximum methods (Fig. 2f) were essentially the same as all the others. But taking into account the relatively small number of elements and the large number of variables it was not a surprise. We had no reliable variance-covariance matrix (VAN VARK 1984). The variance-covariance matrix estimated on the basis of the present sample contains a good number of uncertainties that decreases with the reduction of the number of variables. In the next step we tried to establish what kind of information could be derived from the analysis of long bone measurements. Applying standardized data, group average method, Minkowski (Fig. 2g) and Maximum distances (Fig. 2h) it is obvious that six series are in a group. Series 1 and 10 form another loose but characteristic group. The position of series 5 (Tiszavasvári) seems to be peculiar both in cranial and in postcranial analyses. But it is not a surprise because Tiszavasvári was an early Avar population. The separate standing of series 6, Toponár, deserved attention. On the basis of craniological clusterings Toponár is similar to the Transdanubian series and Arpadian ones. Archaeological finds indicated remarkably different social standing for these cemeteries. For example Fészerlak consisted of poor common people while Toponár was the burial place of a privileged Avar noble community and its servants. This difference was also evident in the long bone measurements. Toponár was characterized by the longest and most developed postcranial material of the whole sample. At the end some words about the application possibility of the postcranial indices. Certain characteristics of the populations could be read out from these when presenting them in a linear way. The main virtue of indices - especially indices of stature and body weight - is their descriptiveness. However, postcranial indices neither separately nor together seem to be able to serve as a basis for the classification of populations. Cluster analyses carried out according to postcranial indices produced the possible widest variations of results. DISCUSSION We know all too well that it is very hard to establish validity of cluster analysis. Nevertheless, by utilizing the great number of hierarchic cluster analyses of different nature that can supplement and check up one another, some markedly outlined conclusions could be drawn even from relatively small samples. When answering the questions put at the beginning we state that there are regional differences among the samples analysed. We found two characteristic groups. One of them is made up by the three Transdanubian and the two Arpadian samples. The other group contains three samples from the region between the rivers Danube and Tisza. Series 1 (Alattyán) was loosely connected to them. Tiszavasvári did not belong to any groups. Skull measurements presented a more characteristic image than long bone measurements did. Populations living under similar social conditions were relatively closer to each other in their body constitutions than in cranial configurations. On the other hand populations leading different way of life were more markedly classified by their body constitutions than by their cranial configurations. The differences between the results of cranial cluster analyses and those of postcranial ones seem to indicate the differences in the way of life and diets. It seems to be the case even when archaeological finds do not produce clear evidences for it.