AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 79235-246 (1989) Consanguinity Avoidance and Mate Choice in Sottunga, Finland ELIZABETH O’BRIEN, L.B. JORDE, BJORN RONNLOF, JOHAN 0. FELLMAN, AND ALDUR W. ERIKSSON Department of Human Genetics, University of Utah School of Medicine, Salt Lake City, Utah 84132 (E.O., L.B.J.);Samfundet Folkhalsans Genetiska Institut, 00101 Helsinki 10, Finland (B.R.,J.O.F., A. W.E.); Institute of Human Genetics, Free University, 1007 MC Amsterdam, The Netherlands (A.W.E.) KEY WORDS Inbreeding, Potential mates analysis ABSTRACT Potential mates analysis is used to determine some of the social and demographic characteristics that influence mate choice in a small island population. Potential mate pools are defined for males in this population; characteristics such as population size and composition with respect to consanguinity are specified. Determinants of mate choice are examined in light of mate availability and potential mate pool characteristics for endogamous maters, exogamous maters, nonmaters, and males of various occupations. Random kinship is assessed from potential mate pools and compared to kinship between actual mates. The island community approximated a random mating population from 1700 to 1900 with some evidence for consanguinity avoidance intensifying in the period 1900-1950. Despite the island’s small population size, kinship coefficients between random mates and actual mates are not high because of relatively high immigration rates. Having considered the contributions of various factors that influence mate choice, the significance of the island mating structure for genetic variation and the distributions of certain genetic disorders is discussed. Regular mating systems have been the basis for much that falls within the scope of population genetics theory. To take one obvious example, the “fixation index,” F, was originally intended a s a measure of the effects of mating systems on heterozygosis (Wright, 1921, 1922). Many theoretical treatments of the effects of mating systems on genetic variability have been developed since the beginning of the century (Kimura and Crow, 1963; Robertson, 1964; Wright, 1965). Less easy to articulate are systems of mating in natural populations. Human populations can prove especially elusive in this regard, lacking strict regularity in mate selection. However, certain demographic and structural characteristics of human populations influence mate availability and, therefore, mate selection. These characteristics, such as the size of a population, its age and pedigree structure, the geographic distribution of indi- @ 1989 ALAN R. LISS. INC. viduals, social stratification, or migration patterns, can all impose constraints on mate choice. These constraints, acting in concert, produce mating systems that are far more complex than those studied by theoreticians. In previous studies we have analyzed inbreeding and the pedigree structure of Sottunga in order to understand the factors underlying the genetic composition of the population. A high frequency of tapetoretinal degeneration, a rare autosomal recessive disease, cannot be attributed to high inbreeding rates (O’Brien et al., 1988a). High frequencies of both tapetoretinal degeneration and von Willebrand disease, a n otherwise rare autosoma1 dominant disorder, are more likely explained by “founder effect,’’ the legacy of unusual gene frequencies originally estabReceived April 15, 1988 revision accepted August 18, 1988. 236 E. O’BRIEN ET AL. lished by a small founding group (O’Brien et al., 1988b). Previous analysis of this population leads to the prediction that any nonrandom mating in Sottunga should be in the direction of consanguinity avoidance. Here mate choice in Sottunga is analyzed in order to examine its effect on genetic variability in this small, remote island community in the &and archipelago. Potential mates analysis (PMA) is used to define Sottunga’s mating structure. Mate choice within human populations occurs within a context of social and demographic constraints. The PMA technique considers the effects of various population characteristics that influence mate availability, mate choice, and ultimately the genetic structure of the population (Dyke, 1971; Leslie, 1985). Of primary interest to this study are factors that affect variation in mate choice through time, such as the size and composition of potential mate pools. Levels of consanguinity for random pairings of mates compared to actual mates will show the extent to which mate selection causes divergence from random mating in Sottunga. Changes in age at marriage and age differences between spouses will depict changes in population composition, which might influence mate choice. Regional migration patterns were known to have had an impact on genetic variability among the islands (Mielke et al., 1976; Jorde et al., 1982); the role of mate selection with respect to migration will therefore be considered. Social stratification within the island, which h a s been studied very little, will be considered for its potential role in mate choice. BACKGROUND Sottunga is one of 16 municipalities (or Lutheran parishes) of the &and archipelago located in the Baltic Sea between Sweden and Finland. The island is one of five very small municipalities situated in the more remote eastern reaches of &and. The founding population of the contemporary Rlanders is considered to be those settlers who reinhabited the islands after the Great Northern War of 1700-1721. During the war h a n d was occupied by Russian troops, and a large proportion of the population of more than 10,000 individuals evacuated the islands. Approximately 6,000 individuals, including many of the previous inhabitants and others from the Swedish mainland, resettled the islands following the war. (See Mead and Jaatinen, 1975; Mielke et al., 1976; Eriksson, 1980 for brief summaries of regional history and settlement.) Most Alanders trace their early origins to the east coast of Sweden. Genetic distance studies have demonstrated high affinity between h a n d e r s and Swedes (Jorde et al., 1982). Until the present century Alland remained isolated from all but a small amount of immigration from Sweden, Finland, and nearby islands. Estimates suggest that until about 1900, only 2.5% of all spouses in h a n d came from outside the population, and parish endogamy was as high as 86% (Workman and Jorde, 1980). Genetic studies have shown larger genetic distances and greater drift potential concentrated in the outer islands (Workman and Jorde, 1980; Jorde et al., 1982). Three generations of gene frequency data from Alanders showed Sottunga to have the largest average genetic distance from other parishes owing to unusual frequencies of ABO and Rh alleles (Eriksson et al., 1973; Carmelli and Jorde, 1982). The turn of this century marks the beginning of the breakdown of Aland’s insularity (Mielke et al., 1976). The effects of this process were demonstrated in a previous study of inbreeding in Sottunga (O’Brien et al., 1988a) where it was shown that, after 200 years of gradual increase, inbreeding declined precipitously after 1900. However, increased migration between islands and with outside populations h a s not been a uniform trend throughout Aland since 1900. Sottunga’s population size, which h a s always been small (300-400), h a s declined in this century because of emigration (Jorde et al., 1982; Mielke et al., 1976; O’Brien et al., 1988a). MATERIALS AND METHODS According to the Swedish ecclesiastical law of 1686, parish ministers were required to record all births, marriages, and deaths among parishioners. The parish registries contain vital data that are quite complete for individuals born in &and since the early 1700s (Mielke et al., 1976,1987).This study is based upon genealogies reconstructed by one of us (B.R.) from information contained in Sottunga’s parish registry. The genealogies currently consist of 3,292 individuals and over 800 nuclear families. The pedigree information spans three centuries and up to 15 generations for some individuals. Over onehalf of those included in our records (1,860 CONSANGUINITY AND MATE CHOICE individuals) were born in Sottunga between 1690 and 1986. This study is based on the first marriages of 573 males from Sottunga’s genealogies. The age distribution of males and females at the time of marriage and the distribution of age differences between spouses were evaluated for these marriages. Potential mate pools were constructed using the following criteria for mate selection. All potential mate analyses include only individuals from the data base who 1) have a known birth year and 2) lived 24 or more years (the cutoff for the lower quartile of the distribution of age at marriage). The potential mate pool for a given male was first drawn from all females in the population 520 years his age, excluding sibs, mother, grandmothers, daughters, aunts, nieces, and first cousins; 99.8%of actual marriages fall within this interval of age difference a t marriage. first-cousin marriages were prohibited without dispensation until 1872 (Norio et al., 1973) and were, therefore, excluded. (Certain other types of marriages, such as a male marrying his brother’s widow, also required dispensation but these implicate second marriages which were excluded from this analysis.) Potential mate pools were established for each male in the genealogy. Mate pools were redrawn for Sottunga-born males paired with Sottunga-born females using the 520 year age difference criterion and excluding the same relatives. Coefficients of kinship were estimated for each 1) male in the genealogy paired with his potential mates, 2) Sottunga-bornmale paired with Sottungaborn potential mates, and 3) pair of actual mates. A second set of kinship coefficients was calculated for the Sottunga-born sample weighting each coefficient by the proportional amount of time t h a t a given male and female were actually available to one another a s potential mates. An individual was considered available to another as a mate from age 16 (youngest age at marriage) until marriage or death, whichever came first. Exposure was calculated as the proportion of the male’s period of availability during which a given female was also available as a mate. This proportion h a s in the denominator the male years from age 16 to marriage or death, whichever came first, and in the numerator the number of female years from age 16 to marriage or death t h a t overlap the male years. If both individuals had 237 missing marriage dates and death dates, including those who did not marry and had not died, then age 50 was assigned the upper age limit of exposure. The calculated exposures, expressed as proportions, were then multiplied by a pair’s coefficient of kinship. Thus, kinship coefficients between individuals who did not overlap i n their years of availability were given weights of “0.” Those who overlapped completely on their dates of availability were given weights of “1,” and all others were given weights between 0 and 1.Mean kinship coefficients for males paired with their potential mates were standardized by dividing the sum of the weighted coefficients by the mean weight (Leslie 1985). Individuals of a cohort spanning a n interval of time do not experience life history events simultaneously; one’s potential mate pool changes constantly as members of the cohort enter and leave potential mate status (Leslie, 1983a). Weighting kinship coefficients in the fashion described above incorporates the probability of choosing for a mate each female who is selected as a potential mate by simple age restriction. This weighting scheme, therefore, more accurately represents a male’s average kinship with his pool of potential mates. Furthermore, consanguineous individuals are agecorrelated (Barrai et al., 1962; Hajnal, 1963; Cavalli-Sforza eta]., 1966; Leslie, 1983b), so that without weighting, consanguineous relationships might bias kinship with potential mates. Average kinship and average size of potential mate pools were determined for males divided into various marriage and migration categories. For each 50-year birth cohort in the interval 1700-1950, potential mate pools were defined for males who were born in Sottunga a n d 1)married in Sottunga, 2) never married, 3) emigrated at marriage. In a previous study it was shown that farmers had larger families, had more descendants through time, and made larger contributions to Sottunga’s gene pool t h a n did nonfarmers (O’Brien et al., 198813).Given t h a t farmers might have had distinct pedigree affiliations, and perhaps social stature, Sottunga’s males were divided into two groups by the broad occupational distinction, farmers vs. nonfarmers. Although individuals might have had more than one occupation in their lifetime, classifications were based on activities of greatest duration 238 E. O’BRIEN ET AL. and economic consequence. Potential mate pools were established for the two groups, and differences between them were evaluated. The average number of first through third cousins, including half-cousins, were calculated for potential mate pools. The number of actual cousin matings at each level of relationship was compared to the number of cousin matings expected under random mating using a chi-square goodness-of-fit test. Potential mates were selected again using a reduced age difference criterion of H O years, and new distributions of observed and expected cousin matings were generated. Ninety percent of actual marriages were between mates meeting this age difference criterion. The two expected distributions of cousin matings demonstrate the amount of age correlation between cousins (Leslie, 1983b). The expected proportions of cousin matings were not adjusted to account for variation in the amount of time t h a t pairs were available to one another as mates. RESULTS The distributions of age at marriage and age differences between spouses are given in Figures 1and 2, respectively. The mean ages a t marriage are 28.5 years for males and 26.3 years for females. The youngest person married was 16, and the oldest was 97. Temporal trends in the size of potential mate pools and average kinship with poten- T 250 I 150 Frequency ” “;I1 1 * \ * - Females 0 <20 tial mates are shown in Figures 3 and 4. In Figure 3 the number of potential mates is shown to increase continuously until 1900 and decline thereafter. This trend is a direct result of changes in Sottunga’s population size. Birth cohort sizes grew until 1850 and declined after 1900. The greater number of potential mates shown for the entire genealogy (“All”) as opposed to island-born males demonstrates the considerable number of migrants in the genealogy. In Figure 4 average kinship between Sottunga-born males and their potential mates is shown to increase from 1700 to 1900 and decline somewhat thereafter. The general increase in kinship until 1900 reflects population growth with better ascertainment of ancestors as pedigrees gained depth. The average coefficients for those born in Sottunga, unweighted and weighted, using a 10or 20-year selection criterion, are very similar. The average values for Sottunga-born males and their spouses (nonrandom kinship) are notably lower than for island-born males and their potential mates (random kinship). The difference between these two components of kinship demonstrates evidence of consanguinity avoidance. Mean values for the number of potential mates for Sottunga-bornmales through time are shown in Figure 5. The labels “20 yrs.” and “10 yrs.” refer to the age difference criteria used to select potential mates. The label “all” refers to all females who met the 25 30 35 40 45 + : ; X ” 50 55 Age at Marriage 60 65 70 75 80 x 85 F 90 Fig. 1. Distribution of ages at mamage for males and females. % 95 -* 100 239 CONSANGUINITY AND MATE CHOICE 140 1 Frequency <-20 -16 -12 -8 -4 0 4 8 12 >20 16 Age D i f f e r e n c e Fig. 2. Distribution of age difference (male age - female age) between males and females at marriage. who never married, and who emigrated at marriage are shown in Figure 6. These values represent no significant differences among groups except in the most recent time period when kinship with potential mates for those who never married is lower than for endogamous maters and emigrants. Kinship values for island-born males and their spouses are shown for comparison. Kinship between mates is lower at all times t h a n average kinship with potential mates. This again suggests persistent consanguinity age criterion. “Non-zero” refers to those who were available as mates, i.e., those with nonzero weights according to the weighting scheme described above. Substantial differences in the size of potential mate pools occurs with changes in selection criteria. The number of available females within 10 years of a given male’s age is less than one-half the number of all females within 20 years of his age. Average kinship with potential mates for males born in Sottunga who married there, ~otentia~ Mates 1 ,50 *0°‘ /i x * * Sottunga- 1 0 I-.+ 1700 --- 1750 1800 1850 1900 Cohort Fig. 3. Average number of potential mates for males of each birth cohort. Potential mates are females aged f20 years of male age. 240 E. O’BRIEN ET AL 0.014 0.012 -- 20 yrs. unneighted 0.010 + 20 yrs. weighted 0.008 0.006 x 10 yrs. weignted 0.004 0 Mates Kinship 0.002 / _ ~ _ _ 0.000 I 1700 ----I 1750 1850 1800 1900 Cohort Fig. 4. Average kinship coefficients with potential mates for males of each cohort and for three selection criteria. Values are for Sottunga-born males paired with potential mates and actual mates. avoidance. The large discrepancy between kinship with potential and actual mates suggests that consanguinity avoidance was strongest in the 1900-1950 period. The decline in kinship with potential mates for nonmaters during this period suggests that these individuals are less connected in a pedigree sense to the rest of the population. Males of all three marriage statuses born in Sottunga are similar in terms of the size of their potential mate pools. Variation in the number of sibs and first cousins is the only factor that might have caused differ- ences among the three groups. Any group having more sibs and/or first cousins would have comparatively smaller mate pools, while contributing more mates to others’ pools. Farmers as a group remained very stable in number through time in Sottunga because the mandated number of farms changed only once during the island’s history. The nonfarmer group followed the island‘s general pattern of growth and decline. A previous study showed that farmers had higher mean family sizes, more descendants, and 180 1201 ,A. , , \ \ 140 , , , f i 100 Po t e n t l a 1 - - - All 20 yrs. Mates * 80 ’ , ^^ 0 1700 * i - * x Nun-zero 10 y r s . I t 1800 1750 Non-zero 20 yrs. i+ 1850 1900 Cohort Fig. 5. Average number of potential mates for “all” males in the genealogy, and for Sottunga-bornmales fol- lowing 10- and 20-year selection criteria with weighting are shown. 241 CONSANGUINITY AND MATE CHOICE o'016 T 0.014 0.012 0.010 Avg Kinshio - Stga-married 0.008 x Unmarried 0.006 +- Emigrated 0.004 0.002 0.000 LI 1700 1750 1850 1800 I 1900 Cohort Fig. 6. Average kinship with potential mates for Sottunga-born males who married in Sottunga, who never mamed, and who emigrated a t marriage are plotted for each birth cohort. Kinship values are unweighted, and the age criterion is f20 years. Average kinship between actual mates is shown for comparison. made larger genetic contributions to the population on average than nonfarmers (O'Brien et al., 1988b). I n addition, it was shown that farmers were the sons and grandsons of farmers more often than expected by random assignment. The higher average kinship values among farmers compared to nonfarmers (1750-1950) demonstrated in Figure 7 reflects the bias in favor of lineal pairs (father-son and father-grandson) in the farmer category. Farmers and nonfarmers show no significant differences in the size of their potential mate pools or in average kinship with them (ANOVA results are not illustrated). No consistent trends through time suggest a distinction between farmers and nonfarmers with respect to potential mates. Farmers do show slightly higher kinship with actual mates through time than do nonfarmers, but the difference is not significant. The average kinship coefficients reported 0'06 0.05 I\\\ Avg . Kinshio 1700 1750 1 BOO 1850 1900 Cohort Fig. 7. Average kinship between farmers (F/J?), between farmers and nonfarmers (F/NF), and between non- farmers (NF/NF) are plotted for each birth cohort. Only males born in Sottunga are included. 242 E. O’BRIEN ET AL. in Figures 4, 6, and 7 reflect a substantial contribution from remote consanguinity. Since mating couples would not usually be aware of consanguinity at a remote level, consanguinity avoidance was further evaluated in terms of close cousin (first through third) relationships. Table 1reports the proportion of potential mates, together with the observed and expected number of matings, for each cousin relationship. The results are subdivided by birth cohort and are given for the 20-year age-restricted potential mate sample. First through third cousin categories are designated in the column headings; “other” consists of potential mates less related than the third cousin level. “N” is the number of marriages for the cohort. The chi-square goodness-of-fit test was used to evaluate the differences between observed and expected numbers of cousin matings. The chi-square values are statistically significant (P 5 .05) only for the 1900-1950 cohort. Combining all categories of cousin marriage also shows no significant difference between the observed and expected number of cousin marriages over the five cohorts (x2 = 4.28, df = 4) in Table 1. However, the trend toward avoidance of cousin matings through time is clear. The same test was applied to observed and expected cousin matings using the smaller 10-year age criterion for potential TABLE 1. Chi-square test for differences between observed and expected numbers of cousin matings by degree of relationship (first through third cousins)’ Cohort N 1 2 1% 70 ,025 ,014 ,030 0 0 2 1.77 1.00 2.09 98 ,024 ,018 ,041 1 2 4 2.42 1.82 4.03 18004 135 .021 ,018 .038 Observed 0 1 7 Expected 2.80 2.43 5.16 18505 138 ,028 ,026 ,052 Observed 0 5 8 Expected 3.84 3.65 7.29 19006 132 ,017 ,016 ,048 Observed 1 3 0 2.32 2.17 6.34 Expected 17002 Observed Expected 17503 Observed Expected 2% ,012 0 86 ,033 1 3.29 ,035 3 4.79 ,047 6 6.51 .044 3 5.83 3 Other ,007 ,904 0 68 .53 63.75 ,049 ,828 3 87 4.82 81.62 .083 ,801 8 116 11.20 108.61 ,071 ,770 6 113 9.86 106.83 ,074 .796 4 121 9.88 105.45 ~ ‘The proportions are average proportions of potentid mates in each cousin category for the cohort. The age criterion for potential mate selection is i 2 0 years. ‘x2 = 1.61. P 5 .51. 3x2 = 2.09. 5 .64. 4x2 = 4.02. P 5 28. = 4.56. P 5 29. 6x2 = 8.86. P 5 .04. 5iL mate selection. These results are reported in Table 2 where the same general pattern of consanguinity avoidance in nearly every cousin category through time, significant only in the recent period, is repeated. DISCUSSION Figure 5 demonstrated how changes in the inclusion criteria for potential mate selection caused large reductions in the size of potential mate pools. Under the 20-year age criterion, weighting the coefficients by availability caused a 35% reduction in the size of potential mate pools. Restricting the age selection criterion more, to +lo years, decreased the size of the potential mate pools by >55% (except in the 1850-1900 cohort where the reduction is 53%) compared to the unweighted 20-year sample. Despite the large reductions in the number of mates caused by varying the age criterion, the average kinship values change very little. This is because remote consanguinity, which is not strongly affected by changes in age restriction, contributes much more to total kinship in this population than does close consanguinity (O’Brien et al., 1988a). In the island population of St. Thomas in the Virgin Islands, larger than Sottunga, yet a small isolate, similarly large reductions in mate pool size resulted from differences in selection criteria (Dyke, 1971). Again, despite these differences in mate pool size, the general effect on kinship with potential mates was minimal. TABLE 2. Chi-square test for differences between observed and expected numbers of cousin matings by degree of relationship (first through third cousins)‘ Cohort 17002 Observed Expected 17503 Observed Expected 18004 Observed Expected 18505 Observed Expected 19006 Observed Expected N 1 ___ 1% __ 2 __ 2% __ 3 ___ Other 70 .033 .007 ,030 .011 .009 .908 0 0 0 2 0 68 2.32 .54 2.10 .so .65 63.59 98 ,029 ,012 ,050 ,028 ,053 331 1 2 4 1 3 87 2.87 1.19 4.57 2.74 5.16 81.47 135 ,027 ,012 ,046 ,028 ,091 ,797 0 1 8 7 3 116 3.61 1.58 6.22 3.79 12.23 107.57 138 ,034 .021 ,060 ,039 ,079 .765 0 5 8 6 6 113 4.76 2.94 8.36 5.38 10.93 105.63 132 ,020 ,012 .053 .032 ,087 ,796 1 4 0 3 3 121 11.48 105.04 4.26 __ 2.68 1.55 6.98 - ‘The age criterion for potential mate selection is +lo years 2x2 = 1.89. P 5 .46. 3x2 = 2.14. P 5 .52. 4x2 = 4.44. P 5 29. 5x2 = 6.89. P 5 .1L 6x2 = 9.62. P 5 .03. CONSANGUINITY AND MATE CHOICE The redistribution of cousin proportions caused by changing Sottunga’s age criterion for potential mates from +20 years to + l o years (Tables 1, 2) is as expected (Hajnal, 1963): cousins in even “steps” (first, second, and third) increase, while cousins of “half-step” decrease. The difference in the distribution of cousin relationships, however, is very small given that the age interval was halved. Leslie (1983a,b) has shown theoretically and empirically the conditions under which age restrictions on potential mates cause negligible differences in consanguinity among married individuals. When pedigrees are deep, the age correlation between relatives declines a s relationships become more distant. Furthermore, in a growing population as Sottunga was until this century, a n individual accumulates a larger number of potential mates among distant relatives through time t h a n among relatives of close consanguineous relationship (Leslie, 1983b). This is because close relatives are the offspring of a smaller range of ancestors removed by a limited range of generations from one’s own. Temporal changes in mate choice patterns in Sottunga occur in the last two birth cohorts, 1850-1900 and 1900-1950. Changes in the demographic characteristics of the population in these two time periods help to explain the differences in mate choice patterns. I n 1850-1900 average kinship coefficients between actual mates and between potential mates reached their maximum values, as did the number of potential mates for island-born males. These trends can be attributed to gains in pedigree depth and population growth, respectively. Emigration among males a t marriage declined somewhat, and female immigration was down to 9.5% from 15.5%in the previous cohort. I n the 1900-1950 cohort, kinship with potential mates leveled off, and kinship between actual mates fell to its lowest level since 1750. Sottunga’s effective size dropped in 1900-1950 because emigration rose to >16%, immigration fell almost to lo%, and the proportion of island-born males who never married reached its maximum of >45%. The decline in population size is also reflected by reduced numbers of potential mates and cousins. Given Sottunga’s reduced size, one might have expected kinship between mates to increase as a result of higher random kinship, but this did not occur. 243 Similar changes in kinship between actual and potential mates accompanied the breakup of isolation beginning at the turn of this century in the Sanday population of the Orkney Islands (Brennan, 1981; Brennan and Boyce, 1980). I n both Sanday and Sottunga lower levels of consanguinity between mates reflects the changed character of potential mate pools as migration patterns, exposure to off-island individuals, and perhaps attitudes about marriage preferences changed. Other studies have shown differences in kinship with potential mates among endogamous maters, exogamous maters, and nonmaters. I n the Virgin Island populations of St. Bart and St. Thomas, those who emigrated or never married were shown to be more closely related to their available mates than were endogamous maters (Dyke, 1971; Leslie, 1980; Leslie et al., 1981). In Sottunga differences in mate availability are not found among exogamous, endogamous, and nonmater groups. Sottunga-born males who emigrated at marriage are not more closely related to their potential mates than others, even if groups are compared in terms of weighted kinship values. Nonmaters are slightly less related to their potential mates than endogamous or exogamous maters and show a slight lack of close cousins among potential mates. Sottunga-born males who never married possess a distinguishing characteristic t h a t might account for their marital status. Sixtythree percent of nonmaters have no recorded occupation. (Nonmaters account for 73% of all males without occupations.) Given t h a t these individuals were born in Sottunga and have pedigree information no less complete than males with recorded occupations, their lack does not appear to signify mere gaps in the data. Furthermore, 30% of those lacking a n occupation were the offspring of fathers born elsewhere as opposed to 12% of those with occupations. These data suggest t h a t individuals with limited economic prospects are at a mating disadvantage. Although a n individual who lacks a recorded occupation does not necessarily lack a means of subsistence, the implication is one of a marginal economic situation aggravated by “outsider” status. It h a s been remarked that farmers (i.e., landowners) constituted an “upper class” in h a n d (Eriksson, 1980), but mate choice does not show marked variation according to this 244 E. O’BRIEN ET AL. occupational distinction. Farmers are slightly more closely related to their mates than are nonfarmers and somewhat more closely related among themselves than nonfarmers. These patterns might reflect some effect of land inheritance on mating structure; alternatively, farmers’ pedigrees tend to have greater depth owing to the fact that the farming occupation is older than any other on the island. As a class, farmers are not significantly different from nonfarmers in mate availability or choice. Because Sottunga’s population size was never much larger than approximately 300 individuals, one might have expected potential mate pools to be small, contain large proportions of related individuals, and give relatively high kinship coefficients. However, kinship with potential mates in Sottunga in this century (<.012) is not much higher t h a n in the larger (661) Northside, Virgin Islands population (.007) for the period 1890 to 1966 (Dyke, 1971). Furthermore, it is lower in Sottunga than in the St. Bart population, which is larger still (2,000). St. Bart contains 12 geographic subdivisions (Leslie, 1980, 1983b), each of which is roughly the size of Sottunga. The average kinship coefficient for random pairs within subdivisions was .026 for the period 19451969. Leslie (1983a) h a s shown that, in a growing isolated population, the frequency of remote consanguineous matings increases a s proportions of potential mates in that category increase. Under these conditions, and with close consanguinity avoidance, remote consanguinity becomes the much larger component of random kinship and has a greater effect t h a n close consanguinity on inbreeding levels. This effect has been shown in the Northside and St. Bart populations (Leslie et al., 1981), as well a s in Sottunga (O’Brien et al., 1988a). Migration, however, limits the effects of remote consanguinity (Leslie, 1983a). Kinship between spouses was lower in Sottunga in all time periods than kinship with potential mates for island-born individuals. I n part this was due to close consanguinity avoidance and in part to marriage with individuals from outside Sottunga. The effects of migration were also demonstrated in a previous study, which showed a striking build-up of remote consanguinity within particular pedigrees, yet the proportion of such pedigrees diminished rapidly in this century as migration patterns changed (O’Brien et al., 1988a). Finally, the effects of Sottunga’s pedigree structure a n d mate choice patterns are considered for their effects on homozygosity levels in the population. Random mating is, of course, expected to produce genotypes in Hardy-Weinberg proportions among offspring. Departures from this expectation can occur in either direction, depending upon such factors as population structure, size, and mating preferences (Workman, 1969). I n the small island population of Tristan da Cunha, for example, incest avoidance produced a n excess of heterozygotes (nonsignificant) at some loci despite a comparatively high inbreeding rate (.040) (Thompson and Roberts, 1980; Jenkins et al., 1985). Consanguinity avoidance slowed the rate of increasing homozygosity resulting from the accumulation of random inbreeding in the Northside population (Leslie et al., 1978). In Sottunga, genotype frequencies for six codominant systems show no departure from Hardy-Weinberg proportions. This result is not inconsistent with the opposing effects of 1) patterns of mate choice including close consanguinity avoidance and marriage with individuals from outside the population and 2) the build-up of remote consanguinity over many generations in a small island community. Two cautionary notes concerning the data used in this analysis should be considered. First, these estimates of kinship, like all such estimates from genealogical data, are somewhat undervalued. Pedigree information cannot be complete in a n absolute sense for a given individual or to the same extent for all individuals. High migration rates in particular cause gaps in pedigree information. For this reason, the bulk of this analysis was restricted to individuals born in Sottunga who, it appears, have pedigree information of equal quality. Second, Sottunga is a small island of approximately 16 km2 and might therefore constitute a homogeneous population with respect to mate choice. Nonetheless, geographic characteristics important to group structure might have been important for mate choice but are unmeasurable with the data at hand. CONCLUSIONS An analysis of mate choice in Sottunga demonstrates consanguinity avoidance through- CONSANGUINITY AND MATE CHOICE out the period of analysis, 1700-1950. A nearly consistent lack of close cousin (first through third) marriages compared to the number expected under random mating was observed in all time periods, although the discrepancy is not significant until the most recent period, 1900-1950. (Incestuous matings were not considered in this analysis; these were certainly avoided and would have elevated the avoidance effect for relationships of close consanguinity.) Mates chosen from outside Sottunga contributed to both close and remote consanguinity avoidance. These marriage trends in Sottunga produce consistently higher random kinship compared to nonrandom kinship through time. Occupational data for this population proved useful to characterize further the factors that influenced mate choice on the island. Sottunga’s emigrants and nonmaters do not appear distinct in the number of potential mates available to them nor in their relatedness to potential mates. However, evidence for economic status a s a condition of marriageability is apparent from the very high prevalence (73%) of nonmaters among those with no designated occupation. The marginal economic status of these individuals is also associated with a n immigrant background. Whereas landholders in Sottunga constitute a group slightly more related among themselves and slightly more related to their spouses than nonfarmers, their potential mate pools and mate choices are not characteristically distinct. This study supports results from a previous investigation of inbreeding, where it was determined that the random and nonrandom components of inbreeding together did not account for the high frequencies of recessive genetic diseases in this population (O’Brien et al., 1988a). Although the population remained very small through time, comparatively high migration rates and population growth into the late 1800s limited the potentially more dramatic effects of the build-up of consanguinity that might have developed were the island more isolated. Here we show evidence for persistent consanguinity avoidance, which should increase heterozygosity above that expected by purely random mating in a closed population. Nevertheless, the number of heterozygotes in Sottunga does not diverge significantly from that expected under random mating. This is not surprising in that goodness-of-fit tests for such differences are insensitive, particularly when in- 245 breeding levels are low and opposing forces have influenced genotypic proportions. ACKNOWLEDGMENTS This research was supported by NSF grant BNS-8319448 and by grants from the Sigrid Jusblius Foundation, Helsinki. We thank Kari Pitkanen, Sarah Williams-Blangero, James Mielke, Margaret Gradie, and two anonymous reviewers for their helpful suggestions during the preparation of this manuscript. LITERATURE CITED Barrai I, Cavalli-Sforza LL, and Moroni A (1962) Frequencies of pedigrees of consanguineousmarriages and mating structure of the population. Ann. Hum. Genet. 25:347-377. Brennan E (1981) Kinship, demographic, social and geographic characteristics of mate choice in Sanday, Orkney Islands, Scotland. Am. J . Phys. Anthropol. 55: 129-138. Brennan E, and Boyce AJ (1980)Mate choice and marriage on Sanday, Orkney Islands. In B Dyke and WT Mom11 (ed.): Genealogical Demography. New York: Academic Press, pp. 197-207. Carmelli D, and Jorde LB (1982) A nonparametric distance analysis of biochemical genetic data from the &and Islands, Finland. Am. J . Phys. Anthropol. 57:331-340. Cavalli-Sforza LL, Kimura M, and Barrai I (1966) The probability of consanguineous marriages. Genetics 54:37-60. Dyke B (1971) Potential mates in a small human population. Sac. Biol. 18:28-39. Eriksson AW (1980) Genetic studies on h a n d geographical, historical and archival data and some other potentialities. In AW Eriksson, HR Forsius, HR Nevanlinna, PL Workman and RK Norio (eds.):Population Structure and Genetic Disorders. New York Academic Press, pp. 459-470. Eriksson AW, Eskola M-R, Workman PL, and Morton NE (1973) Population studies on the h a n d Islands 11. Historical population structure: Inference from bioassay of kinship and migration. Hum. Hered. 23:511-534. Hajnal J (1963) Concepts of random mating and the frequency of consanguineous marriages. Proc. R. Sac. Br. 159:125-177. Jenkins T, Beighton P, and Steinberg AG (1985) Serogenetic studies on the inhabitants of Tristan da Cunha. Ann. Hum. Biol. 12~363-371. Jorde LB, Workman PL, and Eriksson AW (1982) Genetic microevolution in the &and Islands, Finland. In MH Crawford and J H Mielke (eds.): Current Developments in Anthropological Genetics, vol. 2. New York: Plenum Press, pp. 333-365. Kimura M, and Crow JF (1963) On the maximum avoidance of inbreeding. Genet. Res. 4:399-415. Leslie PW (1980) Internal migration and genetic differentiation in St. Barthelemy, French West Indies. In B Dyke and WT Morrill (eds.):Genealogical Demography. New York: Academic Press, pp. 167-177. Leslie P W (1983a) Age correlation between mates and average consanguinity in age-structured human populations. Am. J. Hum. Genet. 35962-977. 246 E. O’BRIEN ET AL. Leslie PW (1983b) Cohorts, overlapping generations and consanguinity estimates. Ann. Hum. Biol. 10257-265. Leslie PW (1985)Potential mates analysis and the study of human population structure. Yrbk. Phys. Anthropol. 28~53-78. Leslie PW, MacCluer JW, and Dyke B (1978) Consanguinity avoidance and genotype frequencies in human populations. Hum. Biol. 50281-299. Leslie PW, Mom11 WT, and Dyke B (1981) Genetic implications of mating structure in a Caribbean isolate. Am. J. Hum. Genet. 33:90-104. Mead WR, and Jaatinen SH (1975) The &and Islands. London: David and Charles. Mielke JH, Workman PL, Fellman J and Eriksson AW (1976) Population structure of the h a n d Islands, Finland. In H H a m s and K Hirschhorn (eds.):Advances in Human Genetics, vol. 6. New York: Plenum Press, pp. 241-321. Mielke H, Pitkanen K, Jorde LB, Fellman JO, and Eriksson AW (1987) Demographic patterns in the ,&land Islands, Finland, 1750-1900. Yrbk. Pop. Res. Finland 2557-74. Norio R, Nevanlinna HR, and Perheentupa J (1973) Hereditary diseases in Finland; rare flora in rare soil. Ann. Clin. Res. 5:109-141. O’Brien E, Jorde LB, Riinnlof B, Fellman JO, and Eriksson AW (1988a) Inbreeding and genetic disease in Sottunga, Finland. Am. J . Phys. Anthropol. 75:477-486. O’Brien E, Jorde LB, and Riinnlof B (198813) Founder effect and genetic disease in Sottunga, Finland. Am. J . Phys. Anthropol. (in press.) Robertson A (1964) The effect of non-random mating within inbred lines on the rate of inbreeding. Genet. Res. 5164-167. Thompson EA, and Roberts DF (1980) Kinship structure and heterozygosity on Tristan da Cunha. Am. J. Hum. Genet. 32445-452. Workman PL (1969) The analysis of simple genetic polymorphisms. Hum. Biol. 41:97-114. Workman PL, and Jorde LB (1980) The genetic structure of the h a n d Islands. In AW Eriksson, HR Forsius, HR Nevanlinna, PL Workman and RK Norio (eds.):Population Structure and Genetic Disorders. New York: Academic Press, pp. 471-486. Wright S (1921) Systems of mating. Genetics 61111-178. Wright S (1922) Coefficients of inbreeding and relationship. Am. Nat. 56:330-338. Wright S (1965)The interpretation of population structureby f-statistics with special regard to systems of mating. Evolution 19:395-420.