AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 82:61-71 (1990)
Population Structure of the Jirels: Patterns of Mate Choice S. WILLIAMS-BLANGERO Southwest Foundation for Biomedical Research, San Antonio, Texas 78284
KEY WORDS
Potential mates analysis, Population genetics,
Nepal
ABSTRACT
The influence of mating practices on genetic structure has been an area of great interest for anthropologists. In this paper, the techniques of potential mates analysis are employed to explore the mating patterns observed among the Jirels, a tribal population of eastern Nepal. Genealogical, anthropometric, dermatoglyphic, and demographic data for members of seven Jirel villages are used. Potential mate pools for a sample of 268 females are enumerated by village. Age structure and the Jirel restriction against clan endogamy are found to severely limit the number of males who are potential mates for a given female. The mating structure of the population is illuminated by statistical analysis of the characteristics of 160 actual mate pairs and all corresponding potential mate pairs. Using this approach several general mate choice practices were verified: 1) biological kin tend to be avoided as mates, 2) members of the same clan are excluded as potential mates, 3) mate exchange between clans is nonrandom, 4) individuals similar in age tend to be selected as mates, and 5 ) mates are drawn from the natal village more often than random expectation. A multivariate phenotypic distance measure between individuals did not reveal any evidence for assortative mating for either anthropometric or dermatoglyphic characters.
The genetic structure of a population depends partially on the patterns of mate choice that are present. An examination of mating structure reveals cultural and demographic factors that influence the genetic characteristics of a population. This paper explores the mating structure of the Jirels, a tribal group in eastern Nepal. High levels of inter-village female marital migration occur in this population; 61.4% of ever-married females are migrants, while only 5.3% of ever-married males are migrants. Migration between villages is nonrandom and demonstrably influences the magnitude of phenotypic microdifferentiation (Williams-Blangero, 1987, 1989; Williams-Blangero and Blangero, 1989). The movement of females to their mates’ natal villages upon marriage is widely reported in the ethnographic literature (Murdock, 1981). To understand the proximate determinants of marital migration, it is necessary to establish mate choice preferences. 0
1990 WILEY-IJSS, INC
The delineation of such marriage rules may permit identification of villages between which marital migration is most likely to occur. Mating rules, as actually practiced, can be inferred using potential mates analysis (Dyke, 1971; Leslie, 1985). Potential mates analysis examines marriage patterns by comparing the characteristics of actual mates to those of possible mates, and several populations have been studied using this approach (Brennan and Dyke, 1980; Brennan et al., 1982; Dyke, 1971; Fix, 1981). There may be little congruence between beliefs held about optimal mate choice practices and actual mating patterns (Leslie et al., 1978; Morton, 1973). Generally, all marriages cannot be of the ideal type because of demographic restrictions. In certain situations, the preferred type of mate is simply unavailable (Dyke, 1971; Kunstadter et al., Received May 3,1988; accepted April 21,1989.
62
S. WILLIAMS-BLANGERO
1963; Leslie et al., 1978). Additionally, the number of available mates from which to choose is partially determined by more subtle internal subdivisions of the population, such as the social and kin structures (Dyke, 1971; Fredlund and Dyke, 1976; Leslie, 1985). Both the characteristics of observed mating pairs and potential mate pools can be examined using potential mates analysis (Leslie, 1985). Many factors must be considered to delineate the pool of individuals from which mates ultimately will be drawn. Age difference, genealogical relationship, social status, and culture-specific rules are all important determinants of whether a pair of individuals may marry. An important feature of social structure that often restricts marriage choice in tribal societies is clan membership (Murdock, 1949). The number of an individual's potential mates is reduced if clan exogamy is an observed marriage rule within a closed population, although the effect is minimal if the number of clans is large (MacCluer and Dyke, 1976). Potential mates analysis is also a useful alternative technique for examining assorta-
\
tive mating, which traditionally has been investigated by evaluating the simple correlations between pairs of mates (Spuhler, 1968).Unlike traditional techniques, the potential mates approach to the analysis of assortative mating is not dependent on the assumption that actual mate pairs are a random sample of potential mate pairs (Leslie, 1985). MATERIALS AND METHODS
Study population This study was conducted in Jiri, Nepal (Fig. 1). The area is located approximately 190 km from the capital city of Kathmandu, in the eastern hills, and is at an average elevation of 2,000 m. The region consists of a system of small independent valleys, but is geographically well defined, being bounded on the east and west by two important rivers, the Tamba Khosi and the Likhu Khola. Administratively it is in the Janakpur Zone of Dolakha District. The Jirels form a small tribal group of approximately 3,525 individuals (1985 estimate, Blangero, 1987) and are geographically restricted to a total of nine villages, all of which are located within the Jiri region
TIBET 84"
INDIA
Fig. 1. Map ofNepal showing the study area.
63
MATE CHOICE IN THE JIRELS
(Fig. 2). The nine villages are 1)Upper Sikri, 2) Lower Sikri, 3) Kharayoban, 4) Ratomate, 5) Dhunge, 6) Bhandar, 7) Kot, 8) Chetrapa, and 9) Jungu. There is dearth of information about the Jirel culture. The existing ethnographic literature (Bista, 1980; Fournier, 1974, 1978; Valeix et al., 1972) consists of general descriptions of the population by researchers primarily concerned with a related population, the Sunwars. Jirels speculate that their population originated as a hybrid group formed by Sherpas and Sunwars seven to eight generations ago. This hypothesis has received support from a recent quantitative genetic study (Blangero, 1987). The Jirel culture shares many of the characteristics of the Sunwar culture, including a system of exogamous patrilineal clans. The Jirel language has elements similar to those found in both Sunar and Sherpa, but is thought to be closer to Sherpa (Strahm and Maibaum, 1971). Marriage choices within the endogamous population are determined in part by clan
membership. Members of a given clan share a putative common descent from a generally unknown ancestor. There are currently a total of 22 patrilineal clans, but over 90% of Jirels belong to one of ten numerically dominant clans. Table 1shows the distribution of clan membership by village. Clan exogamy is prescribed (Bista, 19801, but when endogamous marriages occasionally occur they result in the formation of subclans. These subclans are designated as “high and “low” forms of the original clan. For example, if two Sherbas married, one would assume the subclan Tallo (high) Sherba and the other would take on the Mathilo (low) Sherba designation. Only a few marriages of this type were observed. The only other documented culturally prescribed rule is the prohibition against cousin marriage in the Jirels (Bista, 1980;Fournier, 1974). Upon marriage, residence is either patrilocal or virilocal (Fournier, 1974) and often entails migration by the female from her natal village to the husband‘s village. The
-=
~~
lkm
~
Fig. 2. Map of the study area showing the Jirel villages.
S. WILLIAMS-BLANGERO
majority of Jirel marriages are monogamous, but when polygynous marriages occur the wives and their offspring maintain separate households (Fournier, 1974). Other forms of marriage, such as fraternal polyandry, have not been reported for the Jirels (Bista, 1980).
cuooommr-oo
9"qooyqaqoo
a
rT:
m o ~ o m w m o o Lr:
*
Data collection Household surveys were conducted in each of the nine villages in which Jirels reside as described by Williams-Blangero (1987). Visits to randomly selected households involved the gathering of demographic and genealogical information on all members of the household and were accomplished with the aid of a trained interpreter. Demographic data gathered included age, sex, clan membership, birthplace, and marital history. Unfortunately, unambiguous assignment of subclan status is not possible in the current data set; clans described in this paper could potentially contain unidentified subclans. Marital histories were collected from all individuals over 15 years of age who were present during the household visit. The histories consisted of the number of previous marriages, reason for dissolution, and current marital status. Genealogical relationships between members of the household were recorded. Each individual in the household was asked for his or her name, clan, birthplace, and the names, clans, and birthplaces of his or her mother, father, and offspring. In an attempt t o improve assessment of the genealogical relationships between family members, each available individual was questioned separately when possible. The information was cross checked by asking all adult members of the household present for verification. A pedigree of all members of a given household was drawn during the visit, and the relationships were reverified. Genealogical links between households were determined and verified with the heads of the households involved. The final linking of the pedigrees, which varied from two to four generations in depth, was done at the conclusion of the field project. Anthropometric and dermatoglyphic information was obtained from all individuals in the households who were willing to be measured. Anthropometric data were not collected from Jungu or Chetrapa. However, the available demographic and genealogical data indicated that these two villages exchanged mates primarily with each other.
65
MATE CHOICE IN THE JIRELS
For these reasons, Jungu and Chetrapa were not included in the potential mates analyses. Anthropometric measurements were taken according to the methodology of Weiner and Lourie (1981) and were all recorded to the nearest millimeter. Height, head length, head breadth, minimal frontal diameter, bizygomatic diameter, and head circumference were assessed. These measures have been shown to exhibit moderate t o high heritabilities in several populations (Black, 1983; Devor, 1987; Poosha et al., 1984). Anthropometric variables were ageand sex-corrected using regression techniques and z-score standardization. A complete description of these data is given by Williams-Blangero (1987). Bilateral finger dermatoglyphics were collected using an inkless method (Walker, 1957). Radial and ulnar counts for each finger for each individual were determined by Blangero (1987) as outlined by Holt (1968). This yielded a total of 20 dermatoglyphic traits per subject. The ridge counts were sex-corrected using the z-score method and then reduced to six standardized orthogonal variables using principal components analysis. ANALYSIS
Mating structure is determined by the personal preferences of the individuals involved, mating rules, and the social, kin, and demographic structures of the population within which the process is occurring. A number of techniques related to potential mates analysis were used to consider the relationships between a sample of 268 females of reproductive age (15-49 years) and their actual and potential mates. The kinship, clan, age, spatial, and phenotypic relationships between individuals were described using various pairwise statistics. All individuals were assigned to their natal villages for the analyses.
Kinship relationships Kinship relationships between individuals were obtained from the genealogies. The metric used for assessing kinship was the coefficient of kinship as described by Malecot (1948). This was calculated by the StevensBoyce path searching method (Boyce, 1983). Remote kinship relationships are underestimated since the pedigree depth is relatively shallow. However, most important intragenerational relationships (e.g., first- and second-degree cousins) are identified.
Clan relationships The clan relationships between individuals were assessed using an index of concordance. If the members of the pair were of the same clan, the relationship was scored as one. If the pair was discordant, the pairwise relationship was assigned a score of zero. This measure is used to assess the similarity between pairs or within sets of individuals with regard to clan membership. Age relationships Age differences between all male-female pairs in the reproductive segment of the population were determined by subtracting the female’s age from the male’s age. Spatial relationships Relationships between the spatial locations of the natal villages of the mate pairs were measured in two ways. First, pairs of individuals from the same village were given village endemicity scores of one, while nonendemic pairings were scored as zero. Second, the geographic distance (in kilometers) between a pair of individuals’ natal villages was obtained based on footpath distances. Phenotypic relationships Phenotypic relationships between pairs of individuals were assessed by a multivariate distance measure using the age- and sexcorrected anthropometrics and the sex-corrected dermatoglyphics. Distance measures are commonly applied in studies at the population level but can also be applied at the individual level (Cannings and Thompson, 1981; Defrise-Gussenhoven, 1967) if distance is determined simply between specific points rather than between the centroids of discrete groups. Defrise-Gussenhoven (1967)extended the Mahalanobis (1936) distance for use at the individual level as
d: = (xi - xj)’ (P-l) (xi -
xj),
where Ppl is the sample phenotypic covariance matrix, and xi and xj are vectors of the phenotypic measurements for the i-th and j-th individuals, respectively. This metric divided by two is used in this paper and is distributed as a chi-square, with degrees of freedom equal to the number of traits under consideration. Since the population is subdivided into distinct villages, the means of these pairwise statistics were calculated separately for each
66
S. WILLLAMS-BLANGERO
village, making it possible to summarize each female’s potential mate information in a seven by six matrix whose i-th row consisted of the number of potential mates from the i-th village and the five associated mean indices (i.e., kinship, clan concordance, anthropometric distance, dermatoglyphic distance, and age difference) between the female and potential mates of that village. Summary descriptive statistics were calculated as pooled averages (pooled variances) over villages of potential mates. Potential mates analyses To establish which structures are important in the mate choice process, four constraints were successively placed on the maximal mating pool, following a strategy used by Dyke (1971). These constraints were 1) exclusion of first-degree relatives from the potential mate pool; 2) exclusion of males outside the observed range of relative ages for actual mates (i.e., more than 5 years younger or more than 20 years older than the female) from the potential mate pool for that female; 3) exclusion of all known relatives (+ij > 0) from the potential mate pool; and 4) exclusion of all members of the female’s clan from the potential mate pool for that female. After each constraint was imposed, the pairwise characteristics noted above were recalculated and the means for each village determined. With increasing restrictions, it was expected that the fit between observed patterns in spouse pair characteristics and those resulting from the potential mates analysis should improve. Several specific null hypotheses about marriage patterns were tested by comparing the indices between actual mate pairs and potential mate pairs: 1) marriage is random with respect to kinship, 2) marriage is random with respect to clan membership, 3) mate pairs are random with respect to age, 4) mate pairs are random with regard to phenotype, and 5 ) marriage is random with regard to location. The hy otheses concerning clan concordance an village endemicity were tested by comparing the mean indices between spouse pairs with those between potential mate pairs using a x2 statistic. The remaining hypothesis were tested using the nonparametric Mann-Whitney U statistic (Zar, 1974). The test of whether mate pairs are random with respect to phenotype is a test of multivariate assortative mating. The phenotypic distances between mates were compared
cp
with the distances between potential mates using the Mann-Whitney U test. Smaller phenotypic distances between actual mates (as opposed to potential mates) are indicative of positive assortative mating, while larger mean distances indicate negative assortative mating. A second set of analyses comparing actual and potential mates was performed with the added restriction that a woman’s potential mates must come from her actual mate’s village. This allowed a control for spatial factors that can influence potential mate pool characteristics. Exchange of mates between clans The randomness of marital exchange between clans was tested with information on individuals for whom father’s and mother’s clan were known. A marital exchange matrix between the seven largest clans was tabulated. The ij-th elements contained the number of individuals whose father was a member of clan i and whose mother was a member of clan j. To allow for social proscriptions against clan endogamous marriages, a loglinear contingency table analysis (Bishop et al., 1975) with structural zeros corresponding to the diagonal elements of the exchange matrix was performed. This procedure circumvents problems associated with the unknown subclan heterogeneity biasing the estimates of endogamous mating. A likelihood ratio x2 was calculated to test the hypothesis that marital exchange between clans is random, given clan exogamy. Standardized residuals of each element of the marital exchange matrix were examined for statistical significance. Residuals whose absolute value is greater than or equal to 1.96 are considered evidence for significant deviation from randomness. In this way constellations of clans which referentially exchange mates . i were identifie? A three-dimensional incomplete contingency table based on birthplace, mother’s clan, and father’s clan was used to verify that there is evidence of nonrandom mate exchange between clans, given the heterogeneity among villages in terms of clan composition. RESULTS AND DISCUSSION
The mating structure of a population can be considered in terms of the availability of potential mates. Table 2 presents estimates of the percentage of potential mates available in each village for females between the
67
MATE CHOICE IN THE JIRELS TABLE 2. Percentage of males who are potential mates for females, b y uillage' Female's
Villages Rt
Cumulative restrictions'
US
LS
Kh
All
0
N=60
N=125
N=61
US (N = 30)
1 2 3 4
99.5 54.5 54.5 44.0
100.0 46.4 46.4 42.2
1
100.0 58.3 58.3 51.2
Kh (N = 21)
Rt (N = 57)
Village
LS (N = 38)
2 3 4
Dh (N = 38)
1
2 3 4
Bh (N = 35)
1
2 3 4
Kt (N = 49)
1 2 3 4
Dh
Bh
Kt
Total
N=120
N=101
N=113
N=143
N=723
99.8 44.9 44.9 44.9
100.0 42.9 42.9 37.3
100.0 46.3 46.3 42.1
100.0 45.6 45.6 43.6
99.9 46.7 46.7 45.0
99.9 46.3 46.3 42.5
98.5 56.3 54.8 27.6
100.0 55.6 55.2 51.0
100.0 56.5 56.5 48.8
100.0 56.4 56.4 48.1
100.0 57.8 57.8 55.3
100.0 56.5 56.4 55.4
99.7 56.7 56.4 47.7
100.0 63.0 63.0 62.2
99.9 68.2 68.2 53.7
97.2 63.3 62.6 17.0
100.0 67.0 67.0 62.9
100.0 66.8 66.8 65.7
100.0 68.8 68.8 65.5
100.0 67.2 67.2 64.1
99.8 66.9 66.8 58.4
100.0 59.0 59.0 47.7
100.0 52.6 52.6 45.0
99.8 51.1 51.1 49.2
98.5 48.7 48.4 38.9
100.0 53.1 53.0 44.1
99.9 52.7 52.6 51.3
99.9 53.3 53.2 49.9
99.7 52.6 52.5 46.4
100.0 57.0 57.0 43.7
99.9 51.9 51.9 44.8
100.0 49.7 49.7 49.7
99.9 49.1 49.1 37.3
98.4 51.4 51.4 37.6
99.9 51.7 51.5 48.9
99.9 52.1 52.0 45.9
99.7 51.6 51.6 43.7
100.0 57.3 57.3 55.8
99.8 53.8 53.6 48.3
100.0 52.0 52.0 51.6
100.0 50.7 50.7 48.9
99.7 53.1 53.0 51.0
97.6 52.0 51.7 35.8
99.9 53.3 52.7 45.7
99.4 52.9 52.7 47.2
100.0 60.8 60.8 57.2
100.0 60.2 60.0 56.6
99.8 57.9 57.9 54.6
100.0 58.1 58.1 53.3
99.8 59.4 58.9 54.5
99.7 59.8 58.7 51.1
98.3 58.4 57.2 43.1
99.6 59.2 58.6 52.1
'US, Upper Sikri; 123, Lower Sikri; Kh, Kharayoban; Rt, Ratomate; Dh, Dhunge; Bh, Bhandar; Kt, Kot; J u , Jungu; Ch, Chetrapa. ' 0 , Considering all males as potential mates; I , excludingfirst-degree kin as potential mates; 2, specifying that potential mates be no more than 5 years younger or 20 years older; 3, excluding all relatives as potential mates; 4, excluding clan members as potential mates.
ages of 15 and 49 years by natal village. Cumulative restrictions placed on the pool of potential mates permit inferences about the relative importance of factors that limit mate choice. When all males are considered as potential mates for a female, the size of the pool in each village is the same for any woman regardless of her natal village. However, restrictions placed on mate choice lead to significant differences between females in number of potential mates. Table 2 includes the results of the first four restrictions: 1) exclusion of first-degree male relatives as potential mates, 2) exclusion of males more than 5 years younger or more than 20 years older as potential mates, 3) exclusion of all known male biological relatives as potential mates, and 4) exclusion of male members of the same clan as potential mates. The exclusion of first-degree relatives has minimal effect on the size of the potential
mate pool. For all seven villages, this restriction reduces the potential mate pool by less than 1%.Adding the restriction on age differences between mates markedly reduces the size of the potential mate pool by an amount rangingfrom 40.4% in Kot to 53.6%in Upper Sikri. The exclusion of all biologicallyrelated males (as measured by a nonzero kinship coefficient)from the potential mate pool of a female has a small effect (
Population Structure of the Jirels: Patterns of Mate Choice S. WILLIAMS-BLANGERO Southwest Foundation for Biomedical Research, San Antonio, Texas 78284
KEY WORDS
Potential mates analysis, Population genetics,
Nepal
ABSTRACT
The influence of mating practices on genetic structure has been an area of great interest for anthropologists. In this paper, the techniques of potential mates analysis are employed to explore the mating patterns observed among the Jirels, a tribal population of eastern Nepal. Genealogical, anthropometric, dermatoglyphic, and demographic data for members of seven Jirel villages are used. Potential mate pools for a sample of 268 females are enumerated by village. Age structure and the Jirel restriction against clan endogamy are found to severely limit the number of males who are potential mates for a given female. The mating structure of the population is illuminated by statistical analysis of the characteristics of 160 actual mate pairs and all corresponding potential mate pairs. Using this approach several general mate choice practices were verified: 1) biological kin tend to be avoided as mates, 2) members of the same clan are excluded as potential mates, 3) mate exchange between clans is nonrandom, 4) individuals similar in age tend to be selected as mates, and 5 ) mates are drawn from the natal village more often than random expectation. A multivariate phenotypic distance measure between individuals did not reveal any evidence for assortative mating for either anthropometric or dermatoglyphic characters.
The genetic structure of a population depends partially on the patterns of mate choice that are present. An examination of mating structure reveals cultural and demographic factors that influence the genetic characteristics of a population. This paper explores the mating structure of the Jirels, a tribal group in eastern Nepal. High levels of inter-village female marital migration occur in this population; 61.4% of ever-married females are migrants, while only 5.3% of ever-married males are migrants. Migration between villages is nonrandom and demonstrably influences the magnitude of phenotypic microdifferentiation (Williams-Blangero, 1987, 1989; Williams-Blangero and Blangero, 1989). The movement of females to their mates’ natal villages upon marriage is widely reported in the ethnographic literature (Murdock, 1981). To understand the proximate determinants of marital migration, it is necessary to establish mate choice preferences. 0
1990 WILEY-IJSS, INC
The delineation of such marriage rules may permit identification of villages between which marital migration is most likely to occur. Mating rules, as actually practiced, can be inferred using potential mates analysis (Dyke, 1971; Leslie, 1985). Potential mates analysis examines marriage patterns by comparing the characteristics of actual mates to those of possible mates, and several populations have been studied using this approach (Brennan and Dyke, 1980; Brennan et al., 1982; Dyke, 1971; Fix, 1981). There may be little congruence between beliefs held about optimal mate choice practices and actual mating patterns (Leslie et al., 1978; Morton, 1973). Generally, all marriages cannot be of the ideal type because of demographic restrictions. In certain situations, the preferred type of mate is simply unavailable (Dyke, 1971; Kunstadter et al., Received May 3,1988; accepted April 21,1989.
62
S. WILLIAMS-BLANGERO
1963; Leslie et al., 1978). Additionally, the number of available mates from which to choose is partially determined by more subtle internal subdivisions of the population, such as the social and kin structures (Dyke, 1971; Fredlund and Dyke, 1976; Leslie, 1985). Both the characteristics of observed mating pairs and potential mate pools can be examined using potential mates analysis (Leslie, 1985). Many factors must be considered to delineate the pool of individuals from which mates ultimately will be drawn. Age difference, genealogical relationship, social status, and culture-specific rules are all important determinants of whether a pair of individuals may marry. An important feature of social structure that often restricts marriage choice in tribal societies is clan membership (Murdock, 1949). The number of an individual's potential mates is reduced if clan exogamy is an observed marriage rule within a closed population, although the effect is minimal if the number of clans is large (MacCluer and Dyke, 1976). Potential mates analysis is also a useful alternative technique for examining assorta-
\
tive mating, which traditionally has been investigated by evaluating the simple correlations between pairs of mates (Spuhler, 1968).Unlike traditional techniques, the potential mates approach to the analysis of assortative mating is not dependent on the assumption that actual mate pairs are a random sample of potential mate pairs (Leslie, 1985). MATERIALS AND METHODS
Study population This study was conducted in Jiri, Nepal (Fig. 1). The area is located approximately 190 km from the capital city of Kathmandu, in the eastern hills, and is at an average elevation of 2,000 m. The region consists of a system of small independent valleys, but is geographically well defined, being bounded on the east and west by two important rivers, the Tamba Khosi and the Likhu Khola. Administratively it is in the Janakpur Zone of Dolakha District. The Jirels form a small tribal group of approximately 3,525 individuals (1985 estimate, Blangero, 1987) and are geographically restricted to a total of nine villages, all of which are located within the Jiri region
TIBET 84"
INDIA
Fig. 1. Map ofNepal showing the study area.
63
MATE CHOICE IN THE JIRELS
(Fig. 2). The nine villages are 1)Upper Sikri, 2) Lower Sikri, 3) Kharayoban, 4) Ratomate, 5) Dhunge, 6) Bhandar, 7) Kot, 8) Chetrapa, and 9) Jungu. There is dearth of information about the Jirel culture. The existing ethnographic literature (Bista, 1980; Fournier, 1974, 1978; Valeix et al., 1972) consists of general descriptions of the population by researchers primarily concerned with a related population, the Sunwars. Jirels speculate that their population originated as a hybrid group formed by Sherpas and Sunwars seven to eight generations ago. This hypothesis has received support from a recent quantitative genetic study (Blangero, 1987). The Jirel culture shares many of the characteristics of the Sunwar culture, including a system of exogamous patrilineal clans. The Jirel language has elements similar to those found in both Sunar and Sherpa, but is thought to be closer to Sherpa (Strahm and Maibaum, 1971). Marriage choices within the endogamous population are determined in part by clan
membership. Members of a given clan share a putative common descent from a generally unknown ancestor. There are currently a total of 22 patrilineal clans, but over 90% of Jirels belong to one of ten numerically dominant clans. Table 1shows the distribution of clan membership by village. Clan exogamy is prescribed (Bista, 19801, but when endogamous marriages occasionally occur they result in the formation of subclans. These subclans are designated as “high and “low” forms of the original clan. For example, if two Sherbas married, one would assume the subclan Tallo (high) Sherba and the other would take on the Mathilo (low) Sherba designation. Only a few marriages of this type were observed. The only other documented culturally prescribed rule is the prohibition against cousin marriage in the Jirels (Bista, 1980;Fournier, 1974). Upon marriage, residence is either patrilocal or virilocal (Fournier, 1974) and often entails migration by the female from her natal village to the husband‘s village. The
-=
~~
lkm
~
Fig. 2. Map of the study area showing the Jirel villages.
S. WILLIAMS-BLANGERO
majority of Jirel marriages are monogamous, but when polygynous marriages occur the wives and their offspring maintain separate households (Fournier, 1974). Other forms of marriage, such as fraternal polyandry, have not been reported for the Jirels (Bista, 1980).
cuooommr-oo
9"qooyqaqoo
a
rT:
m o ~ o m w m o o Lr:
*
Data collection Household surveys were conducted in each of the nine villages in which Jirels reside as described by Williams-Blangero (1987). Visits to randomly selected households involved the gathering of demographic and genealogical information on all members of the household and were accomplished with the aid of a trained interpreter. Demographic data gathered included age, sex, clan membership, birthplace, and marital history. Unfortunately, unambiguous assignment of subclan status is not possible in the current data set; clans described in this paper could potentially contain unidentified subclans. Marital histories were collected from all individuals over 15 years of age who were present during the household visit. The histories consisted of the number of previous marriages, reason for dissolution, and current marital status. Genealogical relationships between members of the household were recorded. Each individual in the household was asked for his or her name, clan, birthplace, and the names, clans, and birthplaces of his or her mother, father, and offspring. In an attempt t o improve assessment of the genealogical relationships between family members, each available individual was questioned separately when possible. The information was cross checked by asking all adult members of the household present for verification. A pedigree of all members of a given household was drawn during the visit, and the relationships were reverified. Genealogical links between households were determined and verified with the heads of the households involved. The final linking of the pedigrees, which varied from two to four generations in depth, was done at the conclusion of the field project. Anthropometric and dermatoglyphic information was obtained from all individuals in the households who were willing to be measured. Anthropometric data were not collected from Jungu or Chetrapa. However, the available demographic and genealogical data indicated that these two villages exchanged mates primarily with each other.
65
MATE CHOICE IN THE JIRELS
For these reasons, Jungu and Chetrapa were not included in the potential mates analyses. Anthropometric measurements were taken according to the methodology of Weiner and Lourie (1981) and were all recorded to the nearest millimeter. Height, head length, head breadth, minimal frontal diameter, bizygomatic diameter, and head circumference were assessed. These measures have been shown to exhibit moderate t o high heritabilities in several populations (Black, 1983; Devor, 1987; Poosha et al., 1984). Anthropometric variables were ageand sex-corrected using regression techniques and z-score standardization. A complete description of these data is given by Williams-Blangero (1987). Bilateral finger dermatoglyphics were collected using an inkless method (Walker, 1957). Radial and ulnar counts for each finger for each individual were determined by Blangero (1987) as outlined by Holt (1968). This yielded a total of 20 dermatoglyphic traits per subject. The ridge counts were sex-corrected using the z-score method and then reduced to six standardized orthogonal variables using principal components analysis. ANALYSIS
Mating structure is determined by the personal preferences of the individuals involved, mating rules, and the social, kin, and demographic structures of the population within which the process is occurring. A number of techniques related to potential mates analysis were used to consider the relationships between a sample of 268 females of reproductive age (15-49 years) and their actual and potential mates. The kinship, clan, age, spatial, and phenotypic relationships between individuals were described using various pairwise statistics. All individuals were assigned to their natal villages for the analyses.
Kinship relationships Kinship relationships between individuals were obtained from the genealogies. The metric used for assessing kinship was the coefficient of kinship as described by Malecot (1948). This was calculated by the StevensBoyce path searching method (Boyce, 1983). Remote kinship relationships are underestimated since the pedigree depth is relatively shallow. However, most important intragenerational relationships (e.g., first- and second-degree cousins) are identified.
Clan relationships The clan relationships between individuals were assessed using an index of concordance. If the members of the pair were of the same clan, the relationship was scored as one. If the pair was discordant, the pairwise relationship was assigned a score of zero. This measure is used to assess the similarity between pairs or within sets of individuals with regard to clan membership. Age relationships Age differences between all male-female pairs in the reproductive segment of the population were determined by subtracting the female’s age from the male’s age. Spatial relationships Relationships between the spatial locations of the natal villages of the mate pairs were measured in two ways. First, pairs of individuals from the same village were given village endemicity scores of one, while nonendemic pairings were scored as zero. Second, the geographic distance (in kilometers) between a pair of individuals’ natal villages was obtained based on footpath distances. Phenotypic relationships Phenotypic relationships between pairs of individuals were assessed by a multivariate distance measure using the age- and sexcorrected anthropometrics and the sex-corrected dermatoglyphics. Distance measures are commonly applied in studies at the population level but can also be applied at the individual level (Cannings and Thompson, 1981; Defrise-Gussenhoven, 1967) if distance is determined simply between specific points rather than between the centroids of discrete groups. Defrise-Gussenhoven (1967)extended the Mahalanobis (1936) distance for use at the individual level as
d: = (xi - xj)’ (P-l) (xi -
xj),
where Ppl is the sample phenotypic covariance matrix, and xi and xj are vectors of the phenotypic measurements for the i-th and j-th individuals, respectively. This metric divided by two is used in this paper and is distributed as a chi-square, with degrees of freedom equal to the number of traits under consideration. Since the population is subdivided into distinct villages, the means of these pairwise statistics were calculated separately for each
66
S. WILLLAMS-BLANGERO
village, making it possible to summarize each female’s potential mate information in a seven by six matrix whose i-th row consisted of the number of potential mates from the i-th village and the five associated mean indices (i.e., kinship, clan concordance, anthropometric distance, dermatoglyphic distance, and age difference) between the female and potential mates of that village. Summary descriptive statistics were calculated as pooled averages (pooled variances) over villages of potential mates. Potential mates analyses To establish which structures are important in the mate choice process, four constraints were successively placed on the maximal mating pool, following a strategy used by Dyke (1971). These constraints were 1) exclusion of first-degree relatives from the potential mate pool; 2) exclusion of males outside the observed range of relative ages for actual mates (i.e., more than 5 years younger or more than 20 years older than the female) from the potential mate pool for that female; 3) exclusion of all known relatives (+ij > 0) from the potential mate pool; and 4) exclusion of all members of the female’s clan from the potential mate pool for that female. After each constraint was imposed, the pairwise characteristics noted above were recalculated and the means for each village determined. With increasing restrictions, it was expected that the fit between observed patterns in spouse pair characteristics and those resulting from the potential mates analysis should improve. Several specific null hypotheses about marriage patterns were tested by comparing the indices between actual mate pairs and potential mate pairs: 1) marriage is random with respect to kinship, 2) marriage is random with respect to clan membership, 3) mate pairs are random with respect to age, 4) mate pairs are random with regard to phenotype, and 5 ) marriage is random with regard to location. The hy otheses concerning clan concordance an village endemicity were tested by comparing the mean indices between spouse pairs with those between potential mate pairs using a x2 statistic. The remaining hypothesis were tested using the nonparametric Mann-Whitney U statistic (Zar, 1974). The test of whether mate pairs are random with respect to phenotype is a test of multivariate assortative mating. The phenotypic distances between mates were compared
cp
with the distances between potential mates using the Mann-Whitney U test. Smaller phenotypic distances between actual mates (as opposed to potential mates) are indicative of positive assortative mating, while larger mean distances indicate negative assortative mating. A second set of analyses comparing actual and potential mates was performed with the added restriction that a woman’s potential mates must come from her actual mate’s village. This allowed a control for spatial factors that can influence potential mate pool characteristics. Exchange of mates between clans The randomness of marital exchange between clans was tested with information on individuals for whom father’s and mother’s clan were known. A marital exchange matrix between the seven largest clans was tabulated. The ij-th elements contained the number of individuals whose father was a member of clan i and whose mother was a member of clan j. To allow for social proscriptions against clan endogamous marriages, a loglinear contingency table analysis (Bishop et al., 1975) with structural zeros corresponding to the diagonal elements of the exchange matrix was performed. This procedure circumvents problems associated with the unknown subclan heterogeneity biasing the estimates of endogamous mating. A likelihood ratio x2 was calculated to test the hypothesis that marital exchange between clans is random, given clan exogamy. Standardized residuals of each element of the marital exchange matrix were examined for statistical significance. Residuals whose absolute value is greater than or equal to 1.96 are considered evidence for significant deviation from randomness. In this way constellations of clans which referentially exchange mates . i were identifie? A three-dimensional incomplete contingency table based on birthplace, mother’s clan, and father’s clan was used to verify that there is evidence of nonrandom mate exchange between clans, given the heterogeneity among villages in terms of clan composition. RESULTS AND DISCUSSION
The mating structure of a population can be considered in terms of the availability of potential mates. Table 2 presents estimates of the percentage of potential mates available in each village for females between the
67
MATE CHOICE IN THE JIRELS TABLE 2. Percentage of males who are potential mates for females, b y uillage' Female's
Villages Rt
Cumulative restrictions'
US
LS
Kh
All
0
N=60
N=125
N=61
US (N = 30)
1 2 3 4
99.5 54.5 54.5 44.0
100.0 46.4 46.4 42.2
1
100.0 58.3 58.3 51.2
Kh (N = 21)
Rt (N = 57)
Village
LS (N = 38)
2 3 4
Dh (N = 38)
1
2 3 4
Bh (N = 35)
1
2 3 4
Kt (N = 49)
1 2 3 4
Dh
Bh
Kt
Total
N=120
N=101
N=113
N=143
N=723
99.8 44.9 44.9 44.9
100.0 42.9 42.9 37.3
100.0 46.3 46.3 42.1
100.0 45.6 45.6 43.6
99.9 46.7 46.7 45.0
99.9 46.3 46.3 42.5
98.5 56.3 54.8 27.6
100.0 55.6 55.2 51.0
100.0 56.5 56.5 48.8
100.0 56.4 56.4 48.1
100.0 57.8 57.8 55.3
100.0 56.5 56.4 55.4
99.7 56.7 56.4 47.7
100.0 63.0 63.0 62.2
99.9 68.2 68.2 53.7
97.2 63.3 62.6 17.0
100.0 67.0 67.0 62.9
100.0 66.8 66.8 65.7
100.0 68.8 68.8 65.5
100.0 67.2 67.2 64.1
99.8 66.9 66.8 58.4
100.0 59.0 59.0 47.7
100.0 52.6 52.6 45.0
99.8 51.1 51.1 49.2
98.5 48.7 48.4 38.9
100.0 53.1 53.0 44.1
99.9 52.7 52.6 51.3
99.9 53.3 53.2 49.9
99.7 52.6 52.5 46.4
100.0 57.0 57.0 43.7
99.9 51.9 51.9 44.8
100.0 49.7 49.7 49.7
99.9 49.1 49.1 37.3
98.4 51.4 51.4 37.6
99.9 51.7 51.5 48.9
99.9 52.1 52.0 45.9
99.7 51.6 51.6 43.7
100.0 57.3 57.3 55.8
99.8 53.8 53.6 48.3
100.0 52.0 52.0 51.6
100.0 50.7 50.7 48.9
99.7 53.1 53.0 51.0
97.6 52.0 51.7 35.8
99.9 53.3 52.7 45.7
99.4 52.9 52.7 47.2
100.0 60.8 60.8 57.2
100.0 60.2 60.0 56.6
99.8 57.9 57.9 54.6
100.0 58.1 58.1 53.3
99.8 59.4 58.9 54.5
99.7 59.8 58.7 51.1
98.3 58.4 57.2 43.1
99.6 59.2 58.6 52.1
'US, Upper Sikri; 123, Lower Sikri; Kh, Kharayoban; Rt, Ratomate; Dh, Dhunge; Bh, Bhandar; Kt, Kot; J u , Jungu; Ch, Chetrapa. ' 0 , Considering all males as potential mates; I , excludingfirst-degree kin as potential mates; 2, specifying that potential mates be no more than 5 years younger or 20 years older; 3, excluding all relatives as potential mates; 4, excluding clan members as potential mates.
ages of 15 and 49 years by natal village. Cumulative restrictions placed on the pool of potential mates permit inferences about the relative importance of factors that limit mate choice. When all males are considered as potential mates for a female, the size of the pool in each village is the same for any woman regardless of her natal village. However, restrictions placed on mate choice lead to significant differences between females in number of potential mates. Table 2 includes the results of the first four restrictions: 1) exclusion of first-degree male relatives as potential mates, 2) exclusion of males more than 5 years younger or more than 20 years older as potential mates, 3) exclusion of all known male biological relatives as potential mates, and 4) exclusion of male members of the same clan as potential mates. The exclusion of first-degree relatives has minimal effect on the size of the potential
mate pool. For all seven villages, this restriction reduces the potential mate pool by less than 1%.Adding the restriction on age differences between mates markedly reduces the size of the potential mate pool by an amount rangingfrom 40.4% in Kot to 53.6%in Upper Sikri. The exclusion of all biologicallyrelated males (as measured by a nonzero kinship coefficient)from the potential mate pool of a female has a small effect (
