The way in which the class structure is reproduced, and the implications and significance of these various rules for the two classes, can be seen in their marital strategies.

Although the tawsatin (descent groups) of both classes exhibit a high rate of endogamy, there is no prescriptive rule of endogamy as such. The rules are defined in terms of incest categories, beyond which the pattern and effect of endogamy is merely the result of the various marital strategies adopted. This ‘incest’ category is very small: the only people a man is forbidden from marrying are his mother, his father's and mother's uterine sisters, and other classificatory mothers (which includes those kinswomen of ascending generations), his uterine sister, his daughter, the daughters of his uterine brothers and sisters, and his mother-in-law, sister-in-law and daughter-in-law. In other words, there is no prohibition on marriage with any type of cousin.

Comorian marriage must strike anyone with even the most passing knowledge of the Islands as surprising. All men strive for the lavish, expensive and lengthy marriage form known as ndola nkuu (great marriage) or harusi ya ada (customary marriage) even though a perfectly respectable simple form, mna daho (little house), is all that moslem customary law requires. Mna daho marriages cost almost nothing: ada marriages may cost as much as £5000. Why invest such a large sum in this way? What, in plain anthropological language, are the functions of ada marriage?

There is a general belief among the Ewe of Ghana and Togo that when a person dies the soul of the deceased goes to live in a nether world which is beyond a river, a common belief along the Guinea Coast of West Africa. This spirit-world, known as Tsiefe to the Ewe, is believed to be divided into two units, one for those who have led exemplary lives and the other for ‘criminals’ such as witches, sorcerers and murderers. Some Ewe believe that even the ‘bad part’ of the underworld will not accept the spirit of a criminal unless a special type of ritual has been performed following death. In this case the spirit of the deceased criminal lingers on in this world, plaguing and punishing members of his family or lineage until the appropriate ceremony is performed.

In October 1974 I returned to the Kahama District of Tanzania for a further period of research in northern Unyamwezi where I had previously worked between late 1957 and early 1960. The present paper arises from a consideration of the implications of two facts which impressed me strongly on this second visit. The first of these was that a substantial number of the homestead heads who had been my neighbours in the village of Butumwa for the larger part of my first fieldwork were still alive though some of them had moved to other parts of the District and beyond. The second was that those who had remained, along with many others from surrounding villages, had been moved as part of the Tanzanian Government's national resettlement programme into a new large nucleated village shortly before my return there. These two facts have led me to pay further attention to the nature and functions of pre-1974 settlement patterns and to examine the relation between these and the form of the new scheme. One of the main points which will emerge from my discussion is the need, in trying to understand these settlement patterns, to take careful account of how villages change and develop over time as part of a complex combination of social and ecological processes. This processual aspect of village organization in the area has, I may add, not previously received sufficient attention in my own and other accounts of the situation there.

A number of writers have noted that for East Africa, age organization and lineage organization are structural alternatives. More precisely, Southall (1970) has argued that the full and dominant development of age organization is incompatible with that of a segmentary lineage structure. Forde (1971) has similarly concluded that the two may be alternatives, and that selection between them may be partly dependent on ecological factors. However, I would like to contend that to write of the elaboration, or non-elaboration, of age structures and descent groups or of a conflict of allegiances, may well be a sterile exercise if it is unrelated to an analysis of the mode of livelihood of the people concerned. Furthermore, in light of the modern view of social structure as the outcome, at any point in time, of continuous processes of destructuring and restructuring, the ceaseless turnover of the membership of society, which age structures attempt to reduce to order, can be seen as one of the major sources of these processes, as Bottomore has recently argued (1975). Thus a source of structural stress, and hence ultimately of change, more important, I would argue, than the conflict between age and descent, is that between the generations.

Eminent Indologists have stated that it is inappropriate to use the word ‘caste’ in non-Hindu contexts. Despite such warnings numerous Africanists have used the term, with varying degrees of imprecision. In this paper I first outline different emphases which have appeared in the literature concerning caste in India; then discuss several examples of hierarchical systems in Africa which have been described as ‘caste’. Finally, I focus upon my own fieldwork data concerning the Dime of South-West Ethiopia, amongst whom there operates a system which can, I believe, be unequivocally labelled caste.

The Dyula of northern Ivory Coast consider that joking behaviour (tlon, a word that applies more generally to ‘play’, such as children's games) is an appropriate mode of social interaction between a variety of categories of persons. Such behaviour is by no means peculiar to the Dyula and to related societies. For the most part, the kinds of role relationships characterized by joking among the Dyula have long been familiar to anthropologists: ‘grandparents’ and ‘grandchildren’, cross-cousins, certain categories of affines (cf. Radcliffe-Brown 1952: 90–116). But there is indeed one form of joking which, if not entirely specific to the Dyula, is something of an ethnographic rarity, or else has gone largely unrecorded in the literature: second and subsequent generation slaves (worosso) can and frequently do joke at the expense of free men (horon) in general. The forms which this joking may take are, in certain respects, atypical of other joking relationships among the Dyula. Consequently, it may be of some interest to compare joking between slaves and free men with other types of joking relationships.