|
|
|
Anthropometric variation and population structure of the
island of Pag, Croatia Human Biology, Apr 1994 Our specific purpose here is to present data on phenotypic and underlying genetic anthropometric variation for the island of Pag (northern Dalmatia, Croatia) and to relate the data to past and present migration patterns, geography, and linguistics. The small, fairly environmentally homogeneous island of Pag has experienced differential migration into its various regions throughout its history, and consequently, reproductive isolation of the villages from each other and from the mainland has occurred to varying degrees. Significant anthropometry-geography-linguistics relationships would provide evidence of the action of geographic and cultural isolating factors in the formation of the population's genetic structure . Pag is a part of the Dalmatian population system, which has been under extensive biological and cultural study since 1971 (Rudan 1972, 1980; Chaventre and Rudan 1982; Bennett et al. 1983; Rudan, Angel et al. 1986, 1987; Rudan, Simic et al. 1987; Rudan, Finka et al. 1990; Rudan, Bennett et al. 1990; Simi and Rudan 1990; Simic et al. 1990). Anthropometric data have been collected throughout the northern and middle Dalmatian area (Smolej et al. 1983, 1984; Rudan, Roberts et al. 1986; Rudan, Angel et al. 1987; Rudan Finka et al. 1990; Rudan, Bennett et al. 1990; Smolej-Narancic and Simic 1989; Smolej-Narani et al. 1989, 1990). The ultimate objective of the ongoing holistic anthropological research is to clarify the microevolutionary and sociocultural processes responsible for the population structure and differentiation in the Dalmatian portion of the Balkan peninsula. ISLAND OF PAG Several previous reports have dealt with the geographic, demographic, and linguistic characteristics and the historical background of Pag in detail (Sujoldzic et al. 1987; Sujoldzic 1990). In what follows we briefly summarize these sources. Pag is part of the northern Dalmatian island group (eastern Adriatic, Croatia). It is parallel to Velebit, a mainland mountain massif, and is separated from Velebit by the Pag and Velebit channels (Figure 1). (Figure 1 omitted.) The island is of uneven shape and width and covers an area of 284.6 km sup 2. It is a little less than 60 km long and provides relatively homogeneous physical ecology for its populations. There are three main roads on the island today: Pag-Novalja-Lun, Pag-Povljana, and Pag-Gorica-Dinjiska, The road system was constructed after World War II along the previously existing system of roads and footpaths. Previous investigations of the island population have suggested the existence of marked heterogeneity among the populations of particular villages as a result of different directions of migration over the past few centuries. The ancestral Slavic (Croatian) inhabitants of the island, who settled on Pag between the sixth and eighth centuries, spoke the Chakavian dialect of the Croatian language, traditional on the Dalmatian coast. However, Pag was subject to the influence of the Stokavian dialect through the continuous migration of Croatian Stokavian-speaking settlers from the mainland to the southeastern part of Pag. Particularly significant migrations occurred in the sixteenth and seventeenth centuries, when the population from the Ravni Kotari region of the interior fled from Turkish invaders to the islands of the Zadar region. Thus, since the seventeenth century, Pag has had an older Croatian substratum (Chakavian) and a newer superstratum (Stokavian). The population structure of Pag was probably influenced not only by its shape but also by its administrative division into two parts since the eleventh century. Its northwestern part with Lun and Novalja administratively belonged to the island of Rab and the Rab Bishop's jurisdiction (today only Lun administratively belongs to Rab), whereas the southeastern part together with the town of Pag belonged to Zadar and the Nin Diocese. According to the official census of 1981, 7457 inhabitants live in 13 settlements on Pag. Despite the fact that part of the population emigrated from the island during this century, the population size constantly rose from 1857 to 1948. Emigration from the island to the mainland and overseas was greatest at the turn of the twentieth century, when the economically important wine industry collapsed because of a phylloxera epidemic. Despite the difficult economic conditions, the constant increase in population size in the island's villages from 1857 to 1948 was mainly the result of high natural population increase and immigration of new inhabitants from the mainland to the island. The main reason for immigration was the rapid development of the sea-salt industry in the southeastern part of Pag in the nineteenth century, when the Austrian authorities monopolized the production of sea-salt on Pag. This significant economic activity played a specific role in the employment opportunities on the island. Since 1948 there has been a slow but steady decline in population size because of emigration from the island to urban centers on the main coast. During the last 20 years, decreased natality and increased mortality have led to negative natural population growth. The age structure of the population is characterized by a decrease of younger age groups, an increase in older age groups, and generally equalization in all age groups according to age and sex. SAMPLE AND METHODS Measurements were taken by one member of the research team during field investigations in June 1986. Nine-hundred fifteen persons were examined (405 males and 510 females), age 19 to 84 years. The subjects were the inhabitants of four settlements from the northwestern part of the island (1, Lun; 2, Stara Novalja; 3, Novalja; 4, Kolan), six settlements from the southeastern part of the island (5, Pag; 6, Gorica; 7, Dinjiska; 8, Miskovici; 9, Vlasici; 10, Povljana), and three settlements from the northern part of the island (11, Kustici; 12, Zubovici; 13, Metajna) (see Figure 1). For the sake of simplicity, the northwestern part of the island will be referred to as the western part of the island and the southeastern part will be referred to as the eastern part. (Figure 1 omitted.) The number of subjects in each settlement and their average age are shown in Table 1. Samples from four settlements are excluded from the analyses. These are samples from villages 6 (Gorica) and 8 (Miskovici), which have small population sizes, and samples from towns 3 (Novalja) and 5 (Pag), which are not equally representative of the total population (only 2% of the inhabitants were examined). In the remaining nine villages 16.5-35.3% of the total population was represented in the analyses. They comprised 361 males and 437 females. The age ranges and age distributions are similar in those nine samples. Because anthropometric measurements are influenced by age, it was necessary to estimate variation in the average age between the village samples. Variance analysis showed that only the female samples were significantly heterogeneous in mean age (p < 0.05). Consequently, the F values obtained from the variance analyses for anthropometric traits on the village and regional levels reported here were compared with the respective F values obtained after the adjustment for age performed by covariance analysis. Because these F values were established to be more or less identical, the original nonadjusted values of anthropometric traits were used in all variance and discriminant analyses. The 24 body and 14 head variables listed in Table 2 were all taken according to the IBP recommended technique (Weiner and Lourie 1969) using standard anthropometric instruments (Sieber-Hagner, Switzerland). Data processing was performed on a DATAMINI AT 386 (IBM PC-compatible) computer using SPSS/PC+ software (SPSS Inc. 1987) and additional new programs. Multivariate analyses of variance and discriminant analyses were carried out for the groups of body and head variables to examine the degree and the pattern of between-group phenotypic variation. To estimate the minimum degrees of genetic divergence in the island population using anthropometric data, the minimum F sub ST was derived following the analytical framework of Williams-Blangero and Blangero (1989). They have shown that, assuming additive genetic variation and an underlying polygenic model for quantitative traits, the minimum F sub ST based on phenotypic data is less than or equal to the genetic F sub ST. The minimum F sub ST was calculated according to where t is the number of traits, g is the number of groups, W sub i is the relative census size of the ith group, and C sub p(ii) is the phenotypic distance of the ith group from the population centroid. The C sub p(ii) are the diagonal elements of the standardized phenotypic codivergence matrix C sub p, obtained using phenotypic traits scaled by the inverse of the pooled within-group phenotypic covariance matrix. Biological distances between groups were assessed from anthropometric data using Mahalanobis generalized distances (D sup 2) given by (Equation omitted.) where (character omitted.)) is the vector of trait means for the ith (jth) group and P denotes the pooled within-subdivision phenotypic covariance matrix. These distances represent the minimum genetic distances between populations derived from the phenotypic variation (Williams-Blangero and Blangero 1989). Biological distance (D sup 2) matrices were compared with geographic and linguistic (H sub m) distance matrices using regression analysis. Because the elements of these matrices are not independent, Mantel's (1967) matrix comparison method [modified by Smouse et al. (1986)] was applied to assess significance levels. This technique randomly permutes the rows and columns of one matrix and then evaluates the regression value for the two matrices under comparison. In the present study 1000 permutations were used to generate an empirical significance level for the regression coefficients. Ceographic distances were defined as the road distances between the villages in kilometers. The Hemming linguistic distance (H sub m) matrix was taken from the previous report of Sujoldzic (1990), which dealt with linguistic diversity of the same village populations considered on the basis of 105 basic vocabulary words. RESULTS Analysis at the Village Level. Variance analysis applied separately to head and body variables showed a marked heterogeneity among nine villages of the island with overall discrimination significant at the p After analyzing the degree of phenotypic anthropometric variation in the island population, we studied its pattern. A discriminant analysis was carried out at the village level; the results indicate highly significant discrimination. Table 3 shows the first few discriminant functions and their relative significance. (Table 3 omitted.) For the head variables in both sexes the nine village populations are significantly discriminated by three functions. For the body variables four discriminant functions are significant for males and three are significant for females. These functions in combination account for 78-86% of the total discrimination. The effect of the first (I) and second (II) discriminant functions is shown in Figure 2 for head variables and in Figure 3 for body variables. For both sexes and for both sets of variables the first function tends to separate the villages of the eastern part of the island (villages 7, 9, and 10) from those of the western and northern parts of the island (villages 1, 2, and 4 and villages 11, 12, and 13); it is the strongest discriminator, accounting for 39-57% of the total variation. The second discriminant function separates relatively well the villages from the western part of the island (villages 1, 2, and 4) from those in the northern part (villages 11, 12, and 13), accounting for 18-20% of the total discrimination. The third discriminant function accounts for 9-10% of the total variation, but it is of little apparent anthropometric coherence. Analysis at the Regional Level. Discriminant analysis carried out at the regional level (western, eastern, and northern parts of the island) showed significant morphological discrimination among the three population groups (Table 4). Both discriminant functions are significant discriminators. The first accounts for 73-83% of the total discrimination and separates the eastern population group from the other two regional groups (see Figure 4 for the head variables and Figure 5 for the body variables). The second discriminant function separates the western from the northern population group. The results are identical in both sexes. Table 5 presents the results of the analysis of variance carried out for the three regional population groups. The groups are heterogeneous for most of the studied traits, and the results are almost identical in both sexes. Regarding head dimensions, the eastern population group has the smallest head length relative to head width and the greatest forehead and interorbital widths and morphological face and nose heights. Regarding body dimensions, the western group has larger values of all significantly heterogeneous traits than the other two groups. The degree of divergence at the regional level was compared with that at the village level using the minimum F sub ST, a summary measure of variation between subdivisions relative to the current total population. The minimum F sub ST was derived from all the head and body variables simultaneously (Table 6). In both sexes the minimum F sub ST value is 2.4 times lower at the regional level (males, 0.0096; females, 0.0086) than at the village level (males, 0.0237; females, 0.0201). In addition, all the minimum F sub ST estimates for the individual pairs of population groups are also lower than those for the nine villages, the values increasing from the west-north toward the east-north and the west-east pairs. Because the minimum FST is an estimator of the minimum degree of genetic divergence among subdivisions, the obtained results imply that regional population groups are not the smallest units of mating and that the divergence exists primarily at the village level. Within-Region Analysis. Further minimum F sub ST values were derived within each region to find out the amount of among-village variation in the regions (Table 6). The minimum F sub ST values are the lowest for the eastern region, indicating that the village populations from the eastern part of the island are less differentiated than those from the western and northern parts. Distance Matrix Correlations. The matrix of Mahalanobis distances (D sup 2) between the studied village populations derived for 38 anthropometric head and body variables is shown in Table 7. To analyze anthropometric-geographic-linguistic correspondences in the Pag population, we used the matrices of geographic and linguistic (H sub m) distances as predictors of the anthropometric distance matrix. The analysis was carried out for eight of the nine anthropometrically studied populations. Village 11 (Kustici) was not included because it had not been studied linguistically (Sujoldzic, 1990). Table 8 shows the results of the regression analysis using the Mantel matrix comparison technique. The partial correlation of geography and anthropometric distance is significant in both sexes (p [missing text] DISCUSSION Data on anthropometric variability have often been incorporated into population structure analyses with the ultimate wish of elucidating the effects of different microevolutionary factors on modulating the underlying genetic variation among the populations [reviewed by Relethford and Lees (1982)]. Inferences made using this approach are based on the premise that phenotypic variation adequately reflects genetic variation. This approach was used in studies by, for example, Friedlaender (1975) in Malaysia; Rothhammer and Spielman (1972), Spielman (1973), Neel et al. (1974), Spielman and Smouse (1976), and Stinson (1990) in South America; Vasulu and Pal (1989) and Majumdar et al. (1990) in India; Sokal and Winkler (1987) and Winkler and Sokal (1987) in Kenya; Relethford (1983, 1988) and Relethford et al. (1980, 1981) in Ireland; and Rudan, Roberts et al. (1986), Rudan, Angel et al. (1987), Rudan, Finka et al. (1990), Rudan, Bennett et al. (1990), Smolej-Naranic and Simic (1989), and Smolej-Narancic et al. (1989, 1990) in Croatia. Recent advances in human quantitative genetic methodologies have resulted in a more explicit approach to genetic structure based on anthropometric traits. This approach has been applied in studies by, for example, Williams-Blangero (1989), Williams-Blangero and Blangero (1989, 1990), and Relethford and Blangero (1990) in Nepal and Relethford and Blangero (1990) and Relethford (1991) in Ireland. Both the traditional and more novel approaches to population structure analysis conform to the utility of anthropometric traits in microevolutionary studies. Using the traditional and new methods, we have demonstrated that interpopulational anthropometric variation exists on the island of Pag. The island population is divided into three population groups: western, eastern, and northern, each of which is composed of several village populations. In general, phenotypic heterogeneity among the three groups is due to the specific head dimensions of the eastern group and the body dimensions of the western group. The eastern group, which is more brachycephalic and has a longer face, differs considerably in the head habitus from the mutually more similar western and northern groups. With regard to the body habitus, the western group, which has significantly larger body dimensions, differs considerably from the eastern and northern groups. A significant contribution to the morphological heterogeneity among the three population groups is given by the traits with presumably strong "genetic" components of variation, such as some head dimensions and longitudinal body dimensions (Osborne and DeGeorge 1959; Hiernaux 1963; McHenry and Giles 1971). In addition, several of the least inheritable traits also show significant group heterogeneity, such as body and extremity circumferences, skinfold thicknesses, and body mass, which are particularly responsive to the effects of, for example, nutrition and physical activity. We assume, therefore, that the phenotypic heterogeneity of the three groups living in the same biotope reflects underlying genetic differences in these quantitative traits. This assumption is supported by previous studies of other quantitative variables (morphometric dimensions of metacarpal bones and dermatoglyphics of the digitopalmar complex), which have shown the possibility of discrimination of the island's subpopulations into three basic groups (Kusec 1990; Milicic et al. 1990). Borot et al.'s (1991) study of monogenic traits (GM and KM immunoglobulin allotypes) in the Pag population gave results that are also in favor of the mentioned hypothesis. With a smaller sample of 691 subjects from the same villages grouped into five village groups, Borot et al. (1991) found a genetic distinction of the group of northern villages from the other western and eastern groups. In addition to these biological patterns, anthropometric variation was related to the past and present migration patterns and to geographic distances and linguistic characteristics, which may act as isolation mechanisms on the island. Previous ethnohistorical and migration investigations have shown that three population groups can be differentiated, which corroborates those discriminated morphologically. The western and northern groups correspond to the Croatian substratum that settled on the island in the sixth to eighth centuries. The eastern group corresponds to a superstratum that immigrated from the mainland to this biotope during the sixteenth and seventeenth centuries, most probably introducing a new gene pool to the preexisting one. SujoldziC et al.'s (1987) 13-village migration analysis (N = 1967) for Pag, which relates to the last 100 years, revealed that the eastern villages are the most open to immigration and that the northern villages are the most closed to immigration. This analysis yielded an estimated endogamy value of 64.14% for the entire island. Using the data from Sujoldzic et al.'s (1987) report, we deleted four villages that are not included in the present anthropometric analyses [i.e., two major tourism centers on the island (Novalja and Pag) and two small samples from Gorica and Miskovici] and obtained an endogamy estimate of 65.82% (N = 1198) for the entire island. From the three-region analysis we obtained a considerably higher value of 76.51% for the entire island. For the western, northern, and eastern groups the endogamy estimates were 76.75%, 91.20%, and 46.26%, respectively. The contemporary morphological differences between the three population groups appear to reflect the reproductive isolation that has existed among them throughout the migration history of the island. The northern villages remained in their local biotope, obviously most closed to immigration; with 80% endogamy the northern region can still be regarded as a genetic isolate. The northern villages have retained a local Chakavian dialect of the Croatian language that is different from the dialect of the Chakavian substratum existing today in the western villages of the island (Sujoldzic 1990). According to the 77% endogamy, the reproductive isolation of the western group is also considerable. The amount of overall anthropometric variation between the northern and western populations that belong to the same linguistic group (Chakavian) is much lower compared with the amount of north-east and west-east variation (as seen from the minimum F sub ST values reported in Table 6). The eastern villages have been most open to population migrations from the Zadar coastal region and its hinterland. The endogamy estimate of 46% for the eastern villages suggests considerable gene flow, which has taken place through migrations from the Zadar coastal region and its hinterland. Linguistically, the eastern villages form a distinct Stokavian group. Although anthropometric data could not be collected from all subjects in the migration analysis, so that not all subjects are the same in the two analyses, the correspondence of anthropometric and migrational variation on the island supports the hypothesis of genetically different groups living in the same biotope. The estimates of the minimum F sub ST, [i.e., the minimum levels of genetic divergence (Williams-Blangero and Blangero 1989)] gave us more explicit insight into the underlying genetic structure of the Pag population. The analysis was performed at the village, regional, and within-region levels. The minimum F sub ST value for the regional level is lower than the values for the village and the within-region levels, which implies that genetic divergence occurred primarily at the village rather than at the regional level. However, the minimum F sub ST for all nine villages is considerably higher than that for the villages within each region, indicating that the village divergence includes the interregional divergence, which is an important determinant of the island's population genetic structure. According to the endogamy estimate of 77%, the regional population groups are mutually considerably isolated. Their subdivisions, the village populations, are even more reproductively closed. Because the village populations are themselves small units of mating, a large amount of their divergence possibly arose from genetic drift that has operated for the past 13 centuries. This is not the case with the village populations in the eastern region of the island, whose lowest degree of divergence reflects a continuous influx of immigrants from the mainland in the past four centuries. To corroborate the interpretation of intrapopulational morphological relationships within the context of village settlement history, we analyzed the mutual isolation, social relationships, and anthropometric-geographic-linguistic correspondences of the villages. Our initial hypotheses were (1) that anthropometry and linguistics would covary (Friedlaender et al. 1971; Sokal and Winkler 1987) and (2) that because geography has been shown to influence genetic structure (Malcot 1948; Jorde 1980; Zegura et al. 1990) and because the Mahalanobis distance matrix derived from phenotypic anthropometric traits represents a matrix of minimum genetic distances [as demonstrated by Williams-Blangero and Blangero (1989)], anthropometric distances would correlate positively with geographic distance. Both anthropometry-linguistic and anthropometry-geography relationships turned out as predicted. The results show that both geographic distance and dialect differences act as barriers to intrapopulational contacts, effecting morphological variation on the island. Strong isolation among the village populations (seen also through high endogamy in the western and northern regions) prevented their genetic homogenization through gene flow. According to the 13 village migration predictions, if the conditions remain stable, 26 more generations would be necessary for the homogenization of the entire island population (Sujoldzic et al. 1987). If the reproductively open Pag and Novalja populations are excluded from the analysis, the prediction might be even greater. Thus the current phenotypic morphological variation on the island--characterized by the population's division into three regional groups that are further subdivided into small village populations--reflects genetic variation, originated from ancestral differences, and has been retained because of marked geographic and social isolation. The significance of isolation as a factor of both biological and sociocultural variability has been demonstrated in previous studies in the Dalmatian area [e.g., Rudan et al. (1992)]. Malecot's isolation by distance model adequately describes the anthropometric variation on the middle Dalmatian islands of Hvar and Korcula, where, according to the endogamy estimates of 75% and 64%, respectively, the village populations remained rather closed to migration (Rudan, Angel et al. 1986, 1987; Rudan, Roberts et al. 1986; Simic and Rudan 1990). The village populations of the neighboring island of Brac and the Peljesac peninsula, with endogamy estimates of 38% and 11.5%, respectively, are much more open to immigration and exhibit less anthropometric differentiation, and Malecot's model proved inadequate to evaluate the micoevolutionary processes (Rudan, Angel et al. 1987; Simie et al. 1990). The population of Pag, which shows close morphology-geography relationships, appears to have been more closed to migration than any other middle Dalmatian island and the northern Dalmatian island of Silba (35% endogamy) but still was more open than the northern Dalmatian island of Olib (83% endogamy). Therefore it is hardly surprising that in a previous study based on 10 head and body variables the Pag population took a distant morphological position from the neighboring populations of Silba and Olib and from the middle Dalmatian island populations (Smolej-Narancic and Simic 1989). The fact that the pattern of anthropometric relationships on the island of Pag and in the investigated areas of the Dalmatian region corresponds to the known ethnohistorical, linguistic, and migrational data says much for the conservatism of population morphology. At the same time it indicates the strength of the cultural and geographic isolating factors that cause or maintain anthropometric variation. ACKNOWLEDGMENTS--This research was supported by the Ministry for Science, Technology, and Informatics of the Republic of Croatia (Project 3-01-133 entitled "Anthropological Investigations of Population Structure of Croatia") and by the Institut National d'Etudes Demographiques (project entitled "Anthropological Studies of the Small Isolated Populations of the Zadar Archipelago"). The paper was completed at the Institute for Anthropology, Johannes Gutenberg University, Mainz, Germany, during the tenure of an Alexander von Humboldt Fellowship granted to N. Smolej-NaranciC. We thank Jasna Milicie for gathering the data used in this study. The help of Dubravko Horvat and Zoran Narancie in writing the computer programs is also gratefully acknowledged. FOOTNOTES
LITERATURE CITED
Copyright Wayne State University Press Apr 1994 Source: |
|
This page compliments of Marisa Ciceran Created:
Friday, May 20, 2005; Last Updated: Friday, August 31, 2007 |
