The past, once considered the exclusive domain of historians and antiquarians, has increasingly been embraced by anthropologists. Today, it is difficult to find a major anthropological study that does not claim to offer a diachronic, processual, historical analysis. In examining 10 years of historical anthropological writing, I cover three broad topics in this essay. First, I explain the emergence of a more historical anthropology as a widespread response to a crisis in the conceptualization of culture. Second, I argue that while there are certain identifiable themes that cut across this literature, in general, it reflects long-standing topical interests within anthropology; I review this literature according to these topics rather than divide it into interpretive or cultural studies versus studies of political economy. Third, I try to assess this body of work critically. I concentrate here on anthropological history as both research and textual practice, as well as briefly examine anthropological uses of the concepts of time, colonialism, and structure and agency.

Historians generally agree that crime, especially serious violence and disorder in America’s urban areas, increased during the first half of the nineteenth century, then leveled off, and eventually declined to the end of the century (Ferdinand 1978, 1980; Lane 1968,1979; Warner 1968). The need to improve institutions of control seemed apparent to many city dwellers in the early part of the century. Their responses, which remain part of our criminal justice system, included urban policing, large-scale penitentiaries, and specialized facilities, such as houses of refuge and reform schools for juvenile offenders. While the story of these urban developments has been well told previously (for example, Barnes 1927; Lane 1967; Rothman 1971; Platt 1969; Johnson 1978; Schneider 1980; Monkkonen 1981; Harring 1983), comparatively little is known about crime and control in peripheral communities, even though less than one-tenth of the American population was urban in 1820 and only one-quarter in 1860.

There has developed recently a body of Self-reflexive literature on the methodologies and epistemological bases of social history (e.g., Zunz 1985; Lloyd 1988; Stearns 1988). Some critics have argued for a greater synthesis between social history and politics (Fox-Genovese and Genovese 1976; Eley and Nield 1980; Hochstadt 1982; Rabb 1981); others have focused more on the mode of social-historical disquisition (Stone 1979); still others have placed especial emphasis on the epistemological basis of historical explanation (Lloyd 1988; Giddens 1978; Abrams 1982). Much of the literature is programmatic. Very few of these works provide concrete examples of how their approaches can be applied. The primary purpose of this article is to furnish a case study that highlights the drawbacks to one type of social history while exemplifying the merits of another. In particular, I want to focus on the nature of the relationship between structure and human agency in social-historical explanation, using as my basis the foundling home established by the British Colonial Office on Kephallenia in 1824. Utilizing the data contained in the foundling register kept by the Greek Police Department at Argostoli over the years 1830 to 1834, I conduct statistical analyses similar to those performed by social historians working on comparable institutions elsewhere and thus place the Kephallenian case into a comparative European context.

As furstenberg et al. (1979: 497) suggested over a decade ago, “The enormous promise of quantitative history can only be fulfilled when we are intimately acquainted with the limitations of our sources of information.” This collection of articles follows that suggestion by assessing the limitations of perhaps the most important source of information for social science researchers: the U.S. census. Certainly the degree of underenumeration may seriously affect research in many disciplines. For example, our estimates of migration rates might be inflated. In the process of nominal record linkage, researchers typically assume that those who are not listed in the linked census document out-migrated, when in reality many may simply have been missed by the census taker. As a result, past populations (especially in cities) may appear to have been much more mobile than they actually were. Census underenumeration might also exaggerate voter turnout. If the census in a particular community listed 1,215 eligible voters and election returns showed that 1,150 voted in an election, the voter turnout would be estimated at 95%. However, if just 10% of the eligible voters were not counted by the census taker, the estimated level of voter participation would decline to 85%. Finally, our understanding of the rural economy in the past might be affected by underenumeration. Agricultural production might be overstated if the more successful landowning farmers were more often enumerated and less productive farm tenants were missed. In short, the accuracy of the U.S. census should be a central concern to social science historians.

Few paid-up readers of this journal will deny the importance, as a historical source, of the U.S. manuscript population censuses, particularly those since 1850, which supposedly contain the names of all individuals. Since from 1965 to 1989 my work concerned primarily the four earliest such “nominal” censuses, 1850–80,1 shall restrict my remarks and estimates to them.

Some interesting research has been done, but much more cries out yet to be done, on the manuscript population census as a source, especially on its accuracy and comprehensiveness. Trudging along the road of improvement, we begin to learn that there are potholes aplenty in it, but how many, where, and, most vital of all, how deep? It is sloppy technique for historians to be less critical of the census than of other sources they employ.

We have been studying the migrations of the descendants of nine men who came to Massachusetts before 1650 and have compiled a computerized database that includes all the people born before 1860 in the patrilines. Thus we have what the nine genealogists who studied these families thought was close to a complete list of family members alive in 1850. Here we focus on our attempts to find these individuals on the 1850 federal census.

To facilitate our task, we made up a search list that contained all males alive in 1850, but we omitted females known to have married by 1850. The search list included both the last known place in each individual’s record before 1850 and the next known place after 1850. The list was deliberately over-inclusive. In the 30% or so of the cases in which we did not have death dates for individuals, we carried them on the list until they were 100 years old. Again, the census is supposed to have been taken as of 1 June, but we included anyone who was born or died in 1850, whether it was before or after that date. All told, we searched for 7,627 individuals, about two-thirds men and one-third unmarried women.

The problem of census undercounts, a familiar political issue for modern groups or instrumentalities that consider themselves underrepresented in the Census Bureau statistics, has only recently attracted attention from historians. While the modern “miss rate” is potentially high among some groups (the reason for the emphasis on the homeless in the 1990 census), the general rate of underenumeration appears to have diminished in recent censuses. The bureau acknowledges a net undercount of 5.6% of the population in 1940; the error declined gradually to an estimated 1.4% in 1980 (Burnham 1986; Anderson 1988; Edmondson 1988).

Nineteenth-century censuses no doubt contained more serious errors. Although he did not have underenumeration specifically in mind, the administrator for the 1870 census said that “the censuses of 1850, 1860, and of 1870 are loaded with bad statistics. There are statistics in the census of 1870,I am sorry to say, where some of the results are false to the extent of one-half. They had to be published then, because the law called for it; but I took the liberty of branding them as untrustworthy and in some cases giving the reasons therefore at some length” (quoted in Sharpless and Shortridge 1975: 411). Strikingly modern quarrels surrounded the accuracy of the 1840 Boston and New Orleans censuses, while the errors in the 1870 enumeration of New York City and Philadelphia were sufficient to cause recounts of both cities (ibid. ; Knights 1971: 145).

Political historians have long depended on aggregate records of political behavior—“election returns”—as a crucial resource in the historical study of American politics. During the past 30 years, however, the “new” political historians have increasingly turned to the U.S. census to reveal relationships between political and social behavior in the American past. The veritable revolution in American political history in recent decades, in fact, has largely depended on the political analysis of census records. Political historians now routinely mine the U.S. census for social, demographic, and economic information about voters and officeholders, states and regions, specific decades, and even “political eras.” Census data are most crucial to the political history of the nineteenth century and have benefited, in particular, the debate over levels of voter turnout, the study of political leadership through collective biography, and the ethnocultural model of political behavior.

Quantitative historians who mine diverse sources of evidence usually fret over the quality of their raw materials. Most papers that introduce new sources of data attempt to explain who originally collected the evidence, why it was collected, and how the actual use of the data may differ from the original intentions. These discussions often include estimates of omissions or biases in the records and may explain the sensitivity of results to underlying assumptions.

With a few notable exceptions, scholars have given surprisingly little attention to the quality of census data (see Ginsberg 1988; Furstenberg et al. 1979; Sharpless and Shortridge 1975). This neglect is remarkable in view of the public debate over the accuracy of the modern censuses, the conventions of scholarly scrutiny that apply to other types of records, and the widespread use made of census documents by nearly all disciplinary approaches to study of the past. Stimulated in part by the declining costs of entering and managing large databases, historians have intensively used the manuscript schedules of population for the nineteenth century, and numerous projects were recently completed or are planned for those of the early twentieth century (e.g., Ruggles and Menard 1990). Researchers have also made considerable use of additional schedules collected by the U.S. census beginning in the mid-nineteenth century, including those for slaves, manufacturing, and agriculture.