Model-bound and model-free approach in the holistic analysis of populationstructure: Example from the island of Pag, Croatia

In anthropological investigations of rural populations of the eastern Adria tic region, Croatia, various measures of biological distances (anthropometr ic body and head distances, physiological, dermatoglyphic and radiogrammetr ic bone distances), socio-cultural (linguistic), bio-cultural (migrational kinship) and genetic distances between inhabitants of nine villages of the island of Fag, Croatia were calculated. A model-bound (Malecot's isolation by distance) and a model-free approach (factor analysis of the structure of correlation matrices among different distance measures) were applied. Ling uistic and genetic distances (determined for both sexes), along with most o f the analyzed traits in females (kinship, anthropometric body and head dis tances, physiological distances), revealed a good "fit" to the isolation by distance model. Patterns of correlations among 17 distance matrices in bot h sexes were analyzed, and factor analysis was performed over the correlati on matrix obtained (after exclusion of some traits), explaining 85.9% of to tal variance by 11 retained variables that yielded five factors. A higher-o rder varimax rotation of the factor correlation matrix indicated that the m ajor source of population variability was neither sex-related nor variable- specific, bur rather lay in the interaction of population intrinsic variati on and the extrinsic homogenizing effects of their environment. The authors discuss and compare the application of model-hound and model-free approach in the study of population structure, stressing the dependence of the rate of (micro)evolutionary processes upon the historical processes that favore d or restricted gene flow. Results of different processes of homogenization and/or selective inertia were discussed, on strictly biological, bio-cultu ral and socio-cultural traits in the formation of population structure.