Janus Pannonius Múzeum Évkönyve 37 (1992) (Pécs, 1993)

Régészet - Finnegan, Michael–Szalay, Ferenc: Population Distance Between Late Roman Period to 11th Century Arpadiam Age Populations as Determined by Non-metric Trait Analysis

97 In the present case, while the male sample is a little marginal with respect to side asymmetry, we have decided to pool the sides in the final analysis. Sex Dimorphism Chi square values for sex dimorphism, looking at right sides only and left side only, are presented in Tables 5 and 6 respectively. In right side comparisons, we find nine significant differences at the .01 level and 11 significant differences at the .05 level. Our expectation, due to chance alone, would be three significant differences at the .01 level and approximately 15 significant differences at the .05 level. In this case, these chance expectations are exceeded, showing that some significant differences are attributable to sex alone. In this particular testing, it was felt that seven traits had sample sizes less than or equal to 12 in either males or females, and that this may contribute greatly to the significance of some of these small sample sizes. On the right side the traits are not as randomly dispersed throughout the trait list or the samples. However, on the left side, a number of traits show significant differences in more than one sample. This is lead by the Highest Nuchal Line which showed significant differences in three samples as did the Zygomaxyllary Tuberosity. Ossicle at Lambda and Posterior Ethmoid Foramen each showed a significant difference in two populations and interestingly, the same two samples, Nagypall and Kékesd. The other significant values are more or less randomly distributed, both among the traits and among the samples. There are two samples, Székesfehérvár Street and the Eilend sample which showed but one significant difference each. Sample size alone could probably not have accounted for these differences. Left side comparisons (Table 6) show 22 significant differences at or above the .05 level. This exceeds, by 50%, the number of significant differences expected due to chance alone. However, in this sample, 12 of the significant differences involve traits and samples where the sample size was very small (less than 12). Additionally, while a number of traits showed a significant difference in more than one population (Ossicle at Lambda, Mental Foramen, Zygomaxyllary Tuberosity and Posterior Ethmoid Foramen), not all of these were in the same direction. For example, the Ossicle at Lambda showed a significant sex difference in the Nagypall sample where the higher frequency was found in males and which is contrasted with the higher female frequency for the significant difference in the Kékesd sample. As well, the Mental Foramen showed a significant diference in both Majs and Zengővárkony samples with the former showing a higher incidence in males while the latter showed a higher incidence in females. There is also one example, Zygomatic Tuberosity, which shows a significant difference in the István Square, Majs and Zengővárkony samples, all of which have a higher frequency in favor of the male sample. Additionally, it must be noted that in some samples left and right sides mimic each other while in other samples, they do not. For example, on the right side, the Highest Nuchal Line is significant at the .01 level in István Square and Nagypall samples and at the .05 level in the Zengővárkony sample, while with left sides only the István Square sample shows a significant difference between male and female, and this time at the .05 level rather than the .01 level. The Nagypall and Zengővárkony samples do not show a significant difference. Obviously, the central traits will show the same significant differences between male and female in the left and right tables as these are not bilateral traits. The question now is, what should we do about the sex dimorphic traits? We believe that we should pool the sexes in generating our frequencies for interpopulation analysis in that, for example on the right side, 12 traits have a sample size small enough to warrant questioning. If these examples were excluded, we would be below chance expectation for both the .05 and .01 levels of significance. Additionally, if we look at the separation in per cent (with respect to numbers of males and females), we find that 55.7 % of the sample is male and 44.3 of the sample is female. This difference in number does not exceed 15 % and we do not see this numerical difference as a major problem in this particular analysis. Additionally, the higher incidence frequency was not always with one sex or the other, but on some traits in some samples with one sex and with other traits in other samples with the other sex which should tend to homogenize the samples very slightly and as well, tend to homogenize or reduce the biological distance between samples. Thus, we are following the work of FlNNEGAN (1972, 1978) in pooling the sexes in this particular analysis. There are numerous other ways of handling sex dimorphic traits (GEHARTY, 1974 and JANTZ, 1970). Here, we will pool the sexes, understanding that nowhere do we have a split greater than 60-40 as shown in FINNEGAN and MARCSIK (1979).