The striking changes in marriage and family in many societies of the Third World, as well as changes in aspects of social behavior commonly associated with them, have been well documented in many excellent ethnographic accounts (e.g., Richards 1940; Barnes 1951), and analyzed more generally in various works of synthesis (e.g., Goode 1964; Phillips 1953; Mair 1972). Despite a wealth of scholarship, much about the processes at work requires to be clarified. Study of the topic is indeed beset with difficulties. Some of these are methodological. Thus much of anthropological analysis has focussed on rules or norms. But in a changing situation the rules tend to remain (in that people continue to use the customary terms of discourse), whereas practice no longer conforms to the expectations that the rules or norms once embodied. On the other hand, if analysis were to focus on practice, on how people actually conducted themselves, we would confront another difficulty. The study of change necessarily requires that we compare data at different points on a temporal dimension, but if the resulting measures are to be valid, the data recorded at each point must refer to the same phenomena to be comparable. In societies where accurate records are maintained, some degree of comparability is possible, and the way is thus opened to the use of quantitative measures. By such means changes can be precisely identified, and the consequences of these changes can more reliably be examined. But good records are not generally available in the kinds of societies that anthropologists have conventionally studied. Moreover, even where an area is richly documented in the ethnographic accounts of earlier observers, as is notably the case on the Gazelle Peninsula of New Britain, the data may not be germane for present purposes. So, for example, while the earlier literature on the Gazelle, stemming from many scholarly missionaries and government officials, yields a good deal of information about customary marriage among the Tolai, this tends to be presented in terms of custom and offers little sociological analysis (see e.g., Danks 1889; Winthuis 1927). Some attempt to overcome this difficulty has been to pursue the experience of different generations using the evidence to be gleaned from genealogical records, but the further back in time one goes, the wider the margin of error to which the method is exposed. The best hope for avoiding some of these difficulties is for anthropologists to make return visits to the area in which they had previously conducted research (Foster 1979). The material I present in this paper results indeed from one such revisit.1 Although it was not my intention on this occasion to focus on marriage, in the course of checking the census I had conducted on Matupit in 1960-61,1 found that I was collecting information that seemed to point to important changes in marital patterns. The material I present here on post-