The subject of early (for our purposes, pre–World War II) data on mortality is a vast one, and thus this treatment is quite broad. The emphasis is on identifying classes of data, sources of ambiguity, and general approaches to problems of interpretation. Wherever possible, citations are made to critical surveys of the literature, rather than to the literature itself. Some of the points discussed here can be extended, with appropriate caution and revision, to the equally important, but much less tractable area of early morbidity data.

Protostatistical Populations

There are rich opportunities for studying death and disease in populations for which vital statistics in the modern sense are nonexistent. Primary data sources include faunal evidence obtained by archaeological excavation, epigraphic evidence from funerary monuments, and information contained in parish records and family genealogies. In most cases, however, although these data allow inferences to be made regarding overall mortality among specific and highly localized populations, they contain little information on national-level populations and, with the exception of some faunal evidence, on causes of death. We can address the first shortcoming merely by assuming that the population studied accurately represents the total population, an assumption that is probably robust in very high mortality populations. The second difficulty – the lack of information on causes of death – is irremediable in the main. Furthermore, these data are rarely complemented by accurate population statistics, which are essential for computing rates and probabilities. Because of the dearth of early census data, genealogies and parish records, which provide a link between birth and death records, are especially important for estimating life expectancy (e.g., Henry 1956; Hollingsworth 1964; Wrigley 1968).