Population relationships by isonymy in frontier Pennsylvania

Abstract Data from the 1800 US census were used to study relationships by isonymy among 7 civil subdivisions of Bedford County, Pennsylvania. Two analyses of the data were conducted. In the first analysis heads of household served as the unit of analysis. In the second analysis the total number of individuals in each household was used to correct for varying family sizes. All measures of internal differentiation were approximately doubled when the complete population numbers were used. The head-of-household analysis produced F^sub ST^ and R^sub ST^ values of 0.0012 and 0.0007, respectively; the complete population analysis yielded 0.0021 for F^sub ST^ and 0.0015 for R^sub ST^. Interpopulation a priori kinship estimations were similar using both methods. Conditional kinship estimations varied more, with almost all values negative, but the head-of-- household estimates were less negative. Multidimensional scaling of isonymy values coincided fairly well with actual geographic relationships, but a Mantel test revealed no significant relationship between geographic distances and isonymy, and isolation by distance values indicated a low relationship between the 2 measures. The population of the county was heterogeneous, with low kinship between its constituent communities. This appears to be a result of kin-structured long-distance migration rather than of local processes. The head-of-household values are more comparable with other studies and more representative of population relationships; complete population values exaggerate heterogeneity because of random fluctuations in household size.

KEY WORDS: SURNAMES, 1800 US CENSUS, BEDFORD COUNTY, PENNSYLVANIA, ISONYMY

Surnames are a valuable source of information on the biological structure of historical populations. They are inherited quasi-genetically and, unlike any standard genetic markers, can be determined from historical records. Previous studies of population isonymy have used many different sources of surnames. These sources can be lumped into 2 broad categories. The first is marriage registers, which have proven useful for studies of historical population structure. In addition to using the names for calculation of community isonymy, inbreeding can be estimated from patterns of marital isonymy (Crow 1980) and migration patterns can be derived from the birthplaces of spouses (Swedlund et al. 1984). However, marital data provide a restricted sample of a population, because they include only those individuals who were married within the time span in question. Census data, on the other hand, provide a synchronic sample of the entire population. As such, they have the potential to be more informative about population structure than marital data do. Census-type data are also more easily available (in the form of formal censuses, tax rolls, or even phone books) from many populations that were not subject to any rigorous form of civil registration that would produce accurate marriage records.

Census data generally take the form of a list of heads of households. With telephone directories usually only the individual to whom the bill is sent appears. With tax rolls the person who pays the taxes is listed. Similarly, before 1850 the US census included only the name of the head of household, generally the senior male. This introduces a degree of bias into calculations because not all households contain the same number of people and a household that contains 10 individuals of the same surname represents a larger share of the gene pool than a household that contains 1 or 2. In contemporary England the inclusion of multiple co-resident males with the same surname has been found to increase within-village isonymy by approximately 50% while exerting a minimal effect on the relationships between villages (Lasker 1997). With large enough sample sizes random differences in household size should roughly cancel each other out. But in small populations, especially those that are undergoing demographic changes, differences in household size might dramatically affect isonymy calculations.

Another problem with any form of census data is that maiden names of women are not generally available. As a result, calculations of population relationships reflect only marital migration by males, not females. If 2 adjoining communities were strictly patrilocal, they might appear unrelated despite a high level of bride exchange.

Early studies of isonymy used indexes of isonymy, usually in the form of I^sub ii^ and I^sub ij^, as measures of surname similarity within and between population samples, respectively. Relethford (1988) provided equations for converting isonymy values to measures of genetic kinship, both a priori (phi) and conditional (r). A priori kinship estimates the decline in heterozygosity relative to a founding population. For surnames this founding population is a fictitious population consisting of the first bearer of each surname represented in the population. Conditional, or relative, kinship estimates the loss of heterozygosity relative to the contemporary population. These values are more directly comparable to estimates of r derived from allele frequencies, quantitative traits, and migration matrices.