This paper discusses the achievements and promise of cross-cultural studies of interpersonal violence (mainly focusing on homicide). After comparing the methods of cross-cultural and cross-national research, we review some of the special problems of definition and measurement in cross-cultural studies. Then we review the results of such studies and suggest how cross-cultural studies of violence might be improved, both methodologically and theoretically.
There are relatively few cross-cultural studies of interpersonal violence. In those studies, relatively few variables have been studied; and very few of those variables have been measured similarly. In addition, the methodology of these studies differs considerably from the methodology of cross-national studies. With all these limitations, the results of the crosscultural studies are remarkably consistent with each other and with cross-national findings. We have two main purposes here - to compare the methods of cross-cultural and crossnational research on interpersonal violence (mainly focusing on homicide) and to discuss the achievements and promise of cross-cultural studies.
CROSS-CULTURAL VS. CROSS-NATIONAL RESEARCH
Because the reader of this journal is probably not that familiar with cross-cultural research, we begin our discussion by comparing it in some detail with cross-national research.
Cross-cultural research is more broadly comparative than cross-national research. The cross-national study compares only relatively complex societies. The cross-cultural study compares all types of society, from small hunter-gatherer societies with bands of fewer than 75 people and total populations in the hundreds, to large societies dependent on intensive agriculture with cities and populations in the millions. To be sure, the typical cross-cultural sample contains few or no modern industrial societies; so cross-cultural research is typically not as broadly comparative as it could be. But, other things being equal, the cross-cultural type of study (which compares many if not all types of society) has abetter chance than other kinds of comparison of coming close to the goal of knowing that an observed relationship has more or less universal validity, which is consistent with the general scientific effort to seek more and more comprehensive explanations.
Cross-national research typically uses data that are quantitative; they may not originate or be found in quantitative form, but they can be counted and measured using interval scales. Cross-cultural research typically uses data that are qualitative to begin with; the information in ethnographic texts1 is therefore more likely to be measured ordinally. Data that are quantitative to begin with may or may not be freer of error than qualitative data. It is important to remember that court or police records do not provide direct measures of violence; all measures are indirect, whether they use raw data that are sentences or numbers. S till, it is often presumed that the coding of qualitative information is likely to introduce a good deal of error, because coding rules do not always fit ethnographic descriptions that are often ambiguous and rarely quantitative. In addition, acts of violence may occur infrequently, and therefore the ethnographer (who is there for only a relatively brief time) may not collect information on cases over a longer period of time. But error in a cross-cultural measure can be reduced by using various controls on data quality, as we discuss below. Another way to reduce error is to omit ethnographic cases that are not rated reliably (Ember & Ember, 1992b).
The two types of comparative research (cross-cultural and cross-national) also differ in their typical units of analysis. The type of unit usually compared in the cross-cultural study is the society, a population that more or less contiguously inhabits a particular area and speaks a language not normally …