Navigation – Plan du site
études et essais

The structural articulation of generations in Africa

L’articulation structurelle des générations en Afrique
Malcolm Ruel
p. 51-82


Cet article met au jour le contraste existant entre les sociétés bantous d’Afrique australe d’une part et les sociétés ouest-africaines du groupe Niger-Congo d’autre part. Dans les premières, les générations proches sont nettement distinguées tandis que les générations alternées sont significativement confondues. Au sein des sociétés du groupe Niger-Congo, en revanche, ce sont les générations proches qui sont confondues – un fait que Radcliffe-Brown ignorait. Il existe donc une différence absolue entre les deux régions, encore qu’une certaine continuité puisse être observée à travers le phénomène de l’«expansion bantoue». Une explication possible de l’attention portée aux différences de génération doit être recherchée en premier lieu dans les contradictions existant entre la parenté matrilinéaire d’une part et la résidence patrilinéaire d’autre part ainsi que dans l’émergence décisive du peuple nommé “Mashariki” par C. Ehret dans la région des Grands Lacs.

Haut de page

Texte intégral

1A striking but not commonly recognised feature of the cultural map of Africa concerns the differing value accorded to generational relationships in the traditional societies of West Africa as compared with those of eastern and southern Bantu. Typically amongst the latter peoples lineal kin of different generations are quite precisely distinguished and there is no question about the generational category in which any relative stands to ego. In particular, kin of adjacent generations are clearly separated, whilst those of alternate generations are commonly merged in behavioural norm and terminology, an alternation that serves to sharpen the extended sequence of generations so that ascendant or descendant lineal kin can be precisely identified in the ongoing generational cycle. Where “grandparents” are merged with generational peers, so too “great grandparents” are merged with parents, and in this way each generation at whatever remove is sharply distinguished from those adjacent to it. Separate terms for lineal kin up to seven or even nine generations are common (e.g. Kikuyu, Kuria, Chagga, Lozi, Sotho, see Leakey 1977: 838; Ruel 1957; Moore 1977: 38; Gluckman 1950 figure 6; Hammond-Tooke 1981: 29, respectively).

2In contrast to this general pattern amongst eastern and southern Bantu, amongst West African peoples there is much less precision over generational status, certainly as part of an extended sequence, the overriding issue being rather the simple one of relative generational seniority. Grandparents and parents are less frequently contrasted with each other, as members of opposed categories, than regarded as of the same category, with grandparents the more senior “parents” who, if anything, are owed greater respect and obedience. Thus Bascom (1969: 50) reports of the Yoruba:

“To use the descriptive terms of reference in addressing grandparents is to express contempt for them, showing a lack of normal respect due to them. They are usually addressed as ‘father’ or ‘mother’, and it is said that, in the olden times, when fathers were addressed by their personal names, if a child addressed his father as ‘father’, the reply would be, ‘No, I am not your father. There he is’, pointing to the father’s father.”

3Such merging of adjacent (rather than alternate) generations is common to many West African societies and is characteristically associated with the extended use of the terms for “child” or “father” to refer to all junior or senior lineal kin at whatever generational remove. So, for example, in the case of Tiv: “The phrases ‘one father’… ‘the father who begot us’… or ‘one penis’… may refer to any common ancestor in the direct line of ascent”; and the word for “child” “… may be used of any descendant of the individual in question, no matter how remote or through what line” (Bohannan & Bohannan 1953: 58, 59).

  • 1  When Radcliffe-Brown (1950: 28ff) came to discuss in an African context the structural principle o (...)

4In these West African societies the dominant model for relations between generations is the simple dyadic one of parents to children 1.

5A paper by Walter Sangree comparing the two societies in which he conducted fieldwork–the eastern Bantu Tiriki of Kenya and the variously classified but certainly Niger-Congo Irigwe of the Nigerian Plateau–specifies for these two peoples a contrast which parallels the general difference I am arguing. For the Kenya Tiriki generational kin-relationships are dominated by what he calls the “baguuga triad”, a linked series of three alternate generations consisting of grandparents, ego’s own generation, and grandchildren, who are terminologically equated and share an ongoing identity of interest that contrasts with their members’ “essentially economic and authoritarian” relations with those of the intervening (adjacent) generations who stand to them as parents or children. The world of the Irigwe in Nigeria is by contrast characterized as “profoundly dyadic”. Grandparents and grandchildren are terminologically equated (as grandparents are also with deceased elders) but the relations between members of these alternate generations are not contrasted with those between members of adjacent generations but rather assimilated to them. So, for example, grandparents have much to do with their grandchildren and may “beg for” a younger grandchild, who is “given” to the grandparent to bring up. But in doing this grandparents are acting effectively as parents: the relationship “becomes virtually indistinguishable from the usual parent-child relationship”. In interpersonal behaviour grandparents act somewhat differently from parents but it is the term for “father” rather than “grandfather” that is used to address ritual elders. In Irigwe

“generational subdivisions are of comparatively little significance in structuring authority relations… Instead formal authority is characteristically ascribed to any group simply on the basis of the known or putative agnatic seniority of the men involved; and this leader is then referred to and addressed by the group as ‘father’ (be) or ‘father of the compound’ (beri)” (Sangree 1974: 67, 69 and passim).

6My own fieldwork, like Sangree’s, was conducted amongst both a west and east African people–Banyang of the Cameroon and Kuria of Kenya and Tanzania–and their two societies exhibit a comparable contrast. For Banyang lineal kin are elided in the reckoning of relationships and the terms for “father” and “child” denote ascendant and descendant kin irrespective of generational remove (Ruel 1962a). One of my closest informants spoke of his paternal grandfather as his “real father” since this was the man who had brought him up and whom he most respected. Such usages would be unthinkable for Kuria, who always distinguish sharply between kin of the parental and grandparental generations, and beyond (Ruel 1962b, 1997 App. 1).

7The broad difference that is evidenced in this way between eastern and southern Bantu societies on the one hand and West African Niger-Congo societies on the other is clearly associated with other cultural and structural features. Notably, as Sangree’s paper makes clear, the dominant dyadic model for intergenerational relationships in West Africa owes much to its use in expressing authority relations among kin living together in a localized lineage group. For Yoruba, Ewe and many other West African societies the implicit logic behind describing someone as “the father of my father” is to intensify the respect due to the senior father: if one respects one’s father, one should respect his father (whom he respects) the more. All turns upon the parent-child relationship, senior to junior, and what gives this relationship its especial force is its setting within the corporate organization of lineage groups.

8Nowhere is this more fully demonstrated than in Meyer Fortes’s writings on the Tallensi. He makes the point quite explicitly: “The centre of gravity of the whole social system of the Tallensi is the relationship of father and son” (Fortes 1949: 235).

9This relationship is basic to and is generalized throughout the complex lineage system. The quality of piety that characterizes it and constrains the divisions that can arise between father and son is further extended to the ancestors, whose worship Fortes (1949: 234) characterizes as “a culturally standardized projection… of the tangle of attachments, reciprocities, tensions, and submerged antagonisms that bind parents and children to one another in life”.

10In his book on Tallensi kinship Fortes spends four chapters–a third of the book–specifically on the parent-child relationship. A single chapter of a mere five pages suffices for grandparents and grandchildren.

11All the other West African peoples so far cited–Ewe, Yoruba, Tiv, Irigwe, Yako, Banyang–share (as do many other West African peoples) the basic feature so well documented for Tallensi: within the localized lineage group the authority relationship between senior and junior kin is identified with the parent to child (father to child) relationship which then provides the dominant model for all intergenerational relationships.

12This does not of course mean that grandparents are unrecognised or that the features classically associated with alternate generations do not occur. For example: Tallensi grandchildren can act familiarly towards their grandparents, will joke with them, and there is a sense, as Fortes notes, whereby the life and memory of the grandparents are perpetuated in and through their grandchildren. In this respect attitudes to parents and to grandparents are in contrast: “There is … a kind of freedom and equality between grandparents and grandchildren which could not exist between parents and children. A grandparent does not normally discipline a grandchild” (Fortes 1949: 237).

13Similar patterns are found elsewhere (e.g. Mamprusi, Drucker-Brown 1982; Ewe, Nukunya 1969: 42; Gonja, Goody 1973: 179). But as Fortes makes clear for the Tallensi, the grandparent-grandchild relationship is situationally restricted to immediate kin, and is not generalized to the wider context in the way that the parent-child relationship is. Moreover, the difference gives rise to a conflict in role when the “grandfather” is also a lineage elder and is expected in that capacity to act authoritatively (i.e. as a “father”): “As head of the family the grandfather holds jural, economic and ritual authority over all its members, including those whom he calls his grandchildren, and there is a contrast between his authority and his familiarity in personal relations with his grandchildren” (Fortes 1949: 236).

14One way in which this conflict is resolved is for a lineage elder who needs to discipline his grandchildren to call upon those who stand as “father” to the latter to take action against them. Thus, in the case of Nyaang’zum, who “stormed and raged” against the miscreant grandson of his late half-brother: “But he could do nothing directly. ‘Why don’t you control your sons?’ he shouted at his ‘sons’ (i.e. the sons of his half-brother)… the culprit’s ‘fathers’” (id.: 237). The example illustrates the possible logic, already noted, of the way in which a “father’s father” relationship can operate as an extension and not merely a reversal of the parent-child relation.

15In Fortes’s own terminology, the parent-child relationship acquires jural significance for Tallensi society in a way that the grandparent-grandchild relationship does not. It is consistent with the interpersonal, limited character of the grandparent-grandchild relationship that it operates bilaterally, being common both to maternal and paternal parents. On the other hand, the parental relationship is extended to all senior lineage kinsmen, who stand to ego as his “fathers” (id.: 154).

  • 2  On this point Nyakyusa rules are particularly explicit and are cited as the reason for the residen (...)

16The situation is very different for eastern and southern Bantu, where larger lineage groupings certainly exist but where the emphasis in the corporate organization of kin-groupings lies rather at the lower level of the unitary homestead associated with the compound or extended family. In this situation adjacent generations tend not to be merged but are sharply distinguished, a differentiation that is often associated with a rule of sexual avoidance between adjacent generations, and notably between those who are close affines: it is as though the procreative continuity of society is seen as dependent upon preventing any possible confusion in the sexual activity of kin in immediately succeeding generations 2. Such a sharp differentiation of adjacent generations is in turn associated with a stronger emphasis on the equality or merging of alternate generations. The logic of the “father’s father” now becomes that familiar to us from Radcliffe-Brown’s analysis, where the duplicated relationship cancels out its characterizing authority or social distance and leaves the grandparent equated in status with the grandchild.

17An eastern Bantu example that can be used to match that of Tallensi are the Gusii of Kenya. Like Tallensi, they are a people with an important and elaborate lineage organization (P. Mayer 1949). The corporate homestead or family grouping remains however basic to Gusii lineage structure and the sharp differentiation of generations within the family is replicated upwards, as it were, to the ascendant kin within the lineage, as it is to all lineage kin: “Each lineage up to and including the clan is stratified into generations whose relationships are explicitly patterned after, though not identical with, customary relationships in the family” (Levine & Levine 1966: 13).

18All Gusii kinship terms specify generation relative to ego. Although lineage and descent are fundamental to the social and political structure, the kinship terminology merges descent lines freely but generations never (I. Mayer 1965: 13). Rules of respectful constraint (nsoni) govern the behaviour of those who stand, classificatorily, in adjacent generations, and its obverse, familiarity, characterizes relations between people of alternate generations. The same norms are consistently extended, in alternating sequence, to persons of greater generational remove (I. Mayer 1965 passim; Matsuzono 1981).

19Not all eastern Bantu societies exhibit the classificatory consistency of Gusii with regard to the differentiated articulation of generations, but all give some recognition to it. One major variation occurs with the generational “skewing” of those extra-descent group kin whose cognatic kinship with ego is traced through a grandparent. Shona of Zimbabwe provide an example. In this traditionally patrilineal society the sons of a daughter who has married out of her natal lineage (as she must do) identify all the male members of that lineage as “grandfathers” (i.e. of the same relationship to them as their mother’s father) and are themselves spoken of as “grandsons”. The “grandparent-grandchild” relationship comes here to be associated with the personal link of a man to his mother’s agnatic family group, and in particular to his mother’s brothers. Being linked to the group but at the same time outside it gives such a “grandchild” (muzukuru) a privileged position that enables him to act as arbiter in issues of potential conflict within the family group, to represent the group in its ritual dealings with the family’s spirit elders and, in the case of chiefs, to “look after” the chief’s people in a way that others cannot (Bourdillon 1982: 31-33). In this case the grandparent-grandchild relationship does not give way in the face of organized lineages to the dominant parent-child relationship but is, as it were, moved sideways, serving to define a particular relationship of “identity at a remove” not unlike that between alternate generations.

20Amongst eastern and southern Bantu there are other variants within the general pattern of sharply distinguishing adjacent from alternate generations. My concern, however, is not to trace out kinship structures in detail but to point to the one overriding difference between these societies and those of West Africa. To summarize: for Bantu societies occupying the eastern half of Africa (a grouping to be identified below on linguistic grounds) kinship categories distinguish sharply between generations and all relatives are placed in a distinctive generational category. “Grandparents” are always distinguished from “parents”, “grandchildren” from “children”, and, whilst cognatic kin of adjacent generations are sometimes merged, lineal kin of adjacent generations never are. The separation of adjacent generations is commonly associated with the identification or merging of kin in alternate generations, an alternating pattern that commonly extends both upwards and downwards to lineal kin of greater than three generations’ remove. By contrast, for the West African Niger-Congo societies generational distinctions are less elaborately formalized and less consistently applied. At an interpersonal level, grandparents are likely to be distinguished from parents (although even interpersonally both can be regarded as kinds of parents), but in any wider context the dominant generational distinction is the broad one of relative senior to relative junior generation, that is of parent (commonly “father”) to child. The importance of the parent-child relationship is such that it is extended, both terminologically and behaviourially, to relevant kin at any generational remove, so that a “father”, to whom respect is due, may be a senior relative of ego’s own generation, of his father’s, grandfather’s or great-grandfather’s, without concern for generational standing as such. For eastern Bantu such disregard for relative generational status would be quite anomalous.

How Significant Is this Difference and Can it Be Explained?

21It could be said that in itself the presence or absence of generational differentiation is a trivial matter, requiring no special explanation. Surely, both grandparents and parents are recognised in both areas and the behaviourial norms applying to the two relationships have also much in common. The difference is basically a matter of how, contingently, the two relationships are taken up and elaborated in different aspects of the social organization of the two areas: arbitrarily, more has been made of generational distinctions for eastern Bantu than for West African societies.

22The matter might be left there were it not for two other facts. First, if the difference were merely arbitrary why is it so general to each of the two areas? A random variation should occur randomly, and this one does not. Second, whilst the difference in itself is small and may be counted as trivial, it is associated in at least some of the societies where it is found with other, larger and structurally central institutional features. The very widespread prohibition on any sexual or quasi-sexual connection between kin, especially affines, of adjacent generations has already been noted, as well as the relevance of this prohibition for the creation of the Nyakyusa age villages (see above). For Kikuyu, Embu, Meru, Kuria, Zanaki and others generational sequence was institutionally formalized in a cyclical series of named classes or sets that were basic to the age-grading structure of the societies (Lambert 1956; Bischofberger 1972; Ruel 1962b). Further south, rules of generational sequence appear critically at times of initiation (e.g. Zaramo, Swantz 1970), in the passing on of names between generations (Stefaniszyn 1954) and in the transmission of elders’ “spirits” to younger kinsmen (Holleman 1953).

23Granted then the significance of the difference between the two areas, can we look for an explanation in terms of its correlation with other social structural features of the two areas? For example, we might seek to correlate the merging of adjacent generations with the presence of corporate, localized descent groups (Tallensi, Yoruba, Irigwe, Banyang, etc.) and the differentiation between generations with dispersed lineages composed of scattered, unitary compounds or households (Gusii, Kikuyu, Shona, etc.). Yet once we embark upon such a general correlation, exceptions spring readily to mind: except for the higher levels of lineage organization, Tiv lineages are not localized and Tiv live in scattered, unitary compounds very similar to those characteristic of many eastern Bantu. “Many” but by no means all: the settlement patterns and descent organization of Nyakyusa and Ndembu, to name only two examples, do not fit the general “eastern Bantu” model, but are two of the clearest examples of generational differentiation (Wilson 1951, Turner 1955). These exceptions (or internal variations, for one might wish to preserve the correlation by special argument and analysis focussing on individual cases) lead one to the conclusion that the major difference lies at a deeper, more general level of cultural norm than a simple tying to the specifics of social organization allows.

24Two other considerations point in a similar direction. The first is methodological. The difference we have noted between West African and eastern Bantu societies is less one of kinship arrangements as such (both areas distinguish grandparents from parents and apply similar norms to each category) than in the way kinship relations have become entailed in other social arrangements and have in turn been affected by them, whether the arrangements have been ones of authority relations within descent groups, settlement patterns, the cycle of domestic groups or the extension of kin-norms to the age-grade structure of the wider community. In this situation it would be hazardous to follow a methodology that assumes an “ideological” expression (kin-categories) for a “deeper” social structural cause (actual lineage and settlement patterns). The broad difference that has been noted is one that includes both ideal and actuality, norm and use. It is simply not appropriate to look for an explanation that reduces one level of social life (the ideal or culturally normative) to another (practical arrangements) since each, in this case, is entailed in the other. The second consideration is more straightforward. Least contentiously expressed, the two variants do not have equal standing as possible models of social organization: the West African pattern is less specific, more ambiguous and situationally fluctuating, less formally elaborated in terms of cultural norms, than the eastern Bantu pattern. The eastern Bantu pattern, on the other hand, is the more specialized, the more elaborated. If there be continuity between the two patterns it would seem more likely that the eastern Bantu has developed from some such base as that represented by the West African pattern than that a development should have occurred in the reverse direction.

25The answer to the question that heads this section is then that the difference between the two areas is significant, since it can be tied to other features, and that an explanation for it can usefully be sought in terms of a broad development focussing around this one structurally “entailed” principle, the eastern Bantu pattern being more specific and thus the one requiring particular explanation.

Matriliny, Generational Differentiation and the “Bantu Expansion”

26It is now generally accepted that the Bantu-speaking peoples of the eastern half of Africa, roughly from the western Rift Valley to the Cape of Good Hope, can be considered as a single linguistic region that has greater internal homogeneity and whose differentiation has occurred over a shorter time scale than the remaining areas of Bantu Africa (Guthrie 1967; Dalby 1975; Heine 1979; Ehret & Posnansky 1982; Vansina 1990; Ehret 1998). Work both by linguists and archaeologists over the last two to three decades has been sketching the processes by which the expansion of Bantu occurred. It is accepted that the initial expansion of Bantu languages took place in the northwestern corner of the present Bantu-speaking area, in the borderland between Cameroon and southeastern Nigeria. From there the western Bantu languages spread in a series of waves that first established Bantu speakers in the immediate area of Cameroon, from where some moved firstly into the forest and expanded there before finally penetrating into savannah country south of the rain forest (Heine 1979; Ehret 1982; Phillipson 1985; Vansina 1984, 1990). Turning to the eastern Bantu languages, it is generally agreed, also, that they can be traced in their origins to a nucleus of Bantu speakers who lived in the southern interlacustrine area, somewhere is the region of presentday southwestern Uganda. What is not agreed is how this eastern Bantu nucleus is related to the wider Bantu grouping –whether via the western Bantu (who abut the eastern Bantu nucleus) or directly with their own separate origins in the northwestern Bantu borderland. One group of scholars have argued that eastern Bantu languages have their common and distinctive origin in the Bantu cradleland and reached the interlacustrine area probably by a route north of the rain forest whose traces have been submerged in the later waves of western Bantu (Bennett & Sterk 1977; Bastin, Coupez & de Halleux 1983; Vansina 1984, 1990). A second group of scholars see the eastern Bantu as a further extension of a southerly branch of western Bantu–a twig (or twigs) of the western Bantu tree that have grown to a size commensurate with the parent tree itself (Heine 1979; Ehret 1982, 1998). It is agreed by both groups that the languages and cultures of peoples living in the savannah belt south of the rain forest share many cognate features with eastern Bantu, but the origin of these features is traced back from the east to the west and they do not form part of the earlier eastwards expansion.

27Two grand overarching accounts have been published in the last decade, both by leading scholars in the field, Jan Vansina’s Paths in the Rainforest: Toward a History of Political Tradition in Equatorial Africa (1990) and Christopher Ehret’s An African Classical Age: Eastern and Southern Africa in World History, 1000BC to AD400 (1998). Both are enormously impressive books, synthesizing a vast range of material and opening out a canvas on which events and processes can be viewed on a scale that has not before been possible. Both books rely heavily on a time scale and a series of phased ethnic relationships derived from historical ethnolinguistics, the comparative study of cognate languages which draws out and traces their genetic (“upstream”) relationships–and in Ehret’s case especially traces also the phased borrowings and spread of words within eastern Bantu from non-Bantu (Sudanic and Cushitic) languages. The parts of these books most open to criticism demonstrate the difficulty of extrapolating backwards from language or from the simple present existence of certain social institutions the social structural features of earlier populations. Yet surely social and cultural continuities must have accompanied the historical linguistic continuities that they are at pains to unravel. Here anthropologists have not been a great help, with their attention on particular societies and cultures and their distrust of historical explanations, especially long-term conjectural ones. Nevertheless, it is evident that important historical continuities do exist for African societies and cultures, even if the spelling out of these continuities has so far been only partial and piecemeal.

  • 3  This, of course, is one aspect of the “matrilineal puzzle” first identified by Audrey Richards (19 (...)

28Where an anthropologist can contribute is by looking at the patterns of internal structural change the western Bantu societies may have undergone in their spread through and from the rain forest. One of the most remarkable features of the contemporary distribution of these societies is the continuous belt of matrilineal societies that are largely coterminous with the savannah country south of the forest. I do not wish to comment on the possible conditions that generated this development (assuming that it was in fact a sequential process) except to caution against too radical a reading of the change, for in these matrilineal societies men still hold the positions of authority and command of resources and their successors are other men. It happens that the line of succession places particular emphasis upon matrifocal kinship, succession firstly from full-brother to full-brother and only then to full-sister’s son. But matrifocal kinship is well recognised in most “patrilineal” societies (either in the closer ties of descent from an ancestral mother of a polygynous marriage or in the alternative support to be accorded by a mother’s own kinsfolk), the shift for “matrilineal” societies being the assertion of a claim by the mother’s kinsfolk–her full brothers–for a successor to their status through her. Yet with this claim comes a tension between kinship and co-residence which has intergenerational effects 3: if wives are to live with their husbands, the husbands’ successors can only be found from elsewhere. The lineally extended family that is characteristic of patrilineal societies is then impossible in matrilineal societies. What becomes a submerged patriline in matrilineal societies is now residentially and generationally demarcated: “children” and “grandchildren” in the patriline acquire a special significance.

29The point can be demonstrated by tracing through some of the ways in which lineal kin relationships are taken up in the wider organisation of Bantu societies. In the northwestern Bantu-speaking corner–that closest to the Niger-Congo linguistic family from where Bantu derives–generational alignments accord very closely with the pattern already described for West Africa. The dominant and critical category of relationship remains that of parents to children. As Fernandez (1982: 324) puts it for Fang:

“[I]n Fang patrilineal kinship […] practically all relatives are either parents or children. Older siblings of one’s own sex are either ‘little fathers’ or ‘little mothers’, younger siblings are ‘children’. Collateral relatives are, depending on whether they are traced through the father’s or mother’s side, either ‘little fathers’ or ‘my child’. The parent-child axis is central and crucial.”

30Grandparents are distinguished from parents and are commonly attributed more personal, protective attitudes to their younger kin, but it is the parental role that is extensively generalized to the elders of lineages or to the leaders of kin-constituted settlements (Fernandez 1982; Alexandre & Binet 1958). Amongst Ngombe, further east, one finds a similar pattern of localized corporate lineages, headed by elders whose role is identified as that of “our father” (sangwasu). The leader of the wider, agnatically based community (etuka) “is conceived as the oldest living ‘father’, recognised as having considerable influence with the ancestors” (Wolfe 1961: 28, 30).

31When invoking the dead Ngombe distinguish between the immediate “fathers” of the lineage (“our fathers”, basangwasu) and the distantly placed ancestors or “grandfathers” (batatasu) but the distinction is a broad one and generationally imprecise. The two terms, basangwasu and batatasu, overlap in their reference to the dead “fathers” or ancestors (“grandfathers”) of a lineage, and as a kin-term tatambi (sing of batatasu) applies to “an agnatic kinsmen of two, three or more generations remove, either ascending or descending” (Wolfe 1961: 42, 73). Essentially then, the pattern remains that of generalized dyadic relationships, conceived in terms of the parent-child relationship, only partially and situationally qualified with reference to the alternate generation, “grandparents”. “Fathers” are those who dominate in lineage and community affairs. “Grandfathers” are either personally related individuals or the more distant “fathers” or ancestors of lineage or community. In its classificatory use neither term is generationally exact. In this regard Fang and Ngombe closely match Tallensi and Irigwe.

32This is in accord with a grouping Vansina (1990: 74-77) identifies as the basic level of the “ancient and common” social organisation of Western Bantu and which he terms a “House”. Consisting of 10 to 40 junior kinsfolk gathered round a “big man”, it was the unit of food production and provided an optimally organized labour force for agriculture, trapping and hunting. Vansina (id.: 75) is careful to allow for its recruitment through either male or female parental ties, but the man who leads them is still seen as “father” to its members. Characteristically a number of such kin-based Houses were aggregated to form residentially defined Villages. Accepting this as the traditional, fundamental pattern of social organization, we need to examine the consequent changes that occur when the balance of kin-ties shifts from male to female.

33South of the northwestern Bantu, in an area established by one of the last phases of expansion, such a change occurs which affects both the reckoning of “descent” and the significance accorded to generational categories. As Richards (1950: 221) has shown, the nuclear kin for the local residential group amongst the matrilineal western Bantu (her Type A, the Mayombe-Kongo group) are firstly brothers and secondly their sisters’ sons. Recruitment to the localized descent groups (Vansina’s “Houses”) through sisters is accompanied by wives predominantly coming to live with their husbands. The combination of the two gives rise to a complex mesh of lineage and parental ties, some internal to the local group (sons who continue to live with their fathers), others crosscutting the local group (daughters’ children who are born outside). Men could marry polygynously, but as Richards (id.: 218) points out, such a polygynous family “can never become a grandfamily”. Descent and co-residence cannot run in simple parallel. The localized descent groups have constantly to look for a way of re-establishing their numbers through the attraction to them of sisters’ sons who have been born away. The sons who continue to reside with their fathers stand outside, although in patrilateral relationship to, their fathers’ matrikin groups. The constant crosscutting of residential and filiation ties–matrilineal descent recruitment and patrifiliation–gives a particular significance to generational categories, notably on the father’s side (kitaata). Thus each member of a Kongo matrilineal descent group (futa or kanda) stands also as Child to his or her father’s descent group and Grandchild to the mother’s father’s or father’s father’s descent group (Laman 1953; MacGaffey 1970). This is not a matter of precise generational reckoning per se (in a number of contexts generations are elided in much the same way as for West African societies) but of the structural importance of individual patrilateral links to established matrilineal descent groups–an importance that has come to be denoted in ethnographic usage by capitalizing the “Children” and “Grandchildren” who stand in this special relationship. Thus, MacGaffey (1970: 93): “One is either a member of a descent group occupying a particular place, or a Child, or Grandchild, or affine (nkwezi) to such a group.”

34Patrifiliation, kitaata, acquires in this context a special quality, partly as an individuating tie linking persons to established groups but also as a relationship that depends for its maintainance upon active renewal. Kitaata unsustained will simply lapse: to be maintained it has to be positively re-established. One of the ways by which this is achieved is through marriage. The two forms of preferred marriage that Kongo practise both re-activate former affinal ties and thus help to perpetuate the chain of patrifilial links: patrilateral cross-cousin marriage (Kongo speak of it as “we marry our fathers” i.e. a son marries a daughter of his father’s descent group) and “granddaughter marriage” (whereby a man marries a second generation descendant–“Granddaughter”–from a man of his own descent group). Such marriages have the effect, when repeated in successive generations, of establishing a patrifilial line that weaves in alternating generations across matrilineal descent groups (see Figure 1). One would not expect actual marriages to follow the preferred forms with this degree of regularity but the figure illustrates how the effect of the preferred marriage forms is to create a locally endogamous patriline that is sharply differentiated by generation, with the two sets of alternate generations forming an interlocked, continuous chain.

35Further to the east, Lele society shares the basic organizational feature of Kongo–localized clusters of matrikin interrelated by a dense and complex mesh of patrifilial ties–but in other respects differs markedly from it. A notable feature of Lele society is the association by interest and identity of persons in alternate generations, which has been discussed by Douglas in an early paper (1952) and commented on by Radcliffe-Brown (1953). Douglas presents the Lele case, interestingly, as a counter-example to Radcliffe-Brown’s argument that the equality and friendly relations he claimed as a general principle between alternate generations develop as a counterpoise to the authority and respect that necessarily define the parent-child relationship. As Douglas points out, parental authority is little pronounced for Lele and there is a close personal tie between fathers and sons that cannot be characterized in terms of the respect due to the father. Yet there remains a clear identification of grandparents and grandchildren that is consistently expressed in the kinship terminology. How is this to be explained?

Figure 1. – Kongo Forms of Preferred Mariage

Figure 1. – Kongo Forms of Preferred Mariage

36Douglas does so by relating the merging of alternate generations with a preferred form of marriage, whereby a father gains the right to give in marriage the firstborn daughter of each of his daughters. The “granddaughters” are commonly given in marriage to young men of the man’s own matrilineage, serving to attract them to live with him in his (and their) matrikin village. Thus Douglas correlates the merging of alternate generations with the situation of weak authority plus preferred marriage between alternate generations. The preferred form of marriage, she argues, meets a structural, integrative need: “The custom helps to build up a permanent association between clan and locality, counteracting the dispersive effects of patrilocal marriage” (Douglas 1952: 64). Its effect in this respect is represented in Figure 2, which is developed from Douglas’s own diagram (1963: 92) and shows how matrilines are dispersed between villages by the shift in residence expected at marriage both of bride (who comes to live with her husband) and groom (who is expected to take up residence in the village associated with his matriline) but how the preferred marriage form serves to reinforce in alternate generations the local connection between village and matriline.


37Radcliffe-Brown, in his comment on the Lele institution, takes issue with Douglas’s analysis. He points out that the “preferred marriage” is not strictly one between alternate generations, since the patrilateral grandfather does not himself marry his granddaughter but rather gives her in marriage to a younger matrikinsman. Radcliffe-Brown wishes to interpret this right as a delayed reciprocal claim deriving from the grandfather’s earlier marriage and the children he has produced for his wife’s kin. This analysis ignores the local dimension of Lele descent grouping and advances an argument based purely upon the formal logic of kin and lineage relationships. For Radcliffe-Brown the form of preferred marriage is seen as a kind of debt between lineages, a product of the principle of lineage unity: the marriage of the daughter’s daughter to a matrikinsman of her grandfather’s lineage being a “return marriage” due from one matrilineage to another.

38Shorn of their general structural-functional claims, both analyses can be seen as valid, and they do not conflict. Different in form from the Kongo “granddaughter marriage”, the Lele father’s right to give away his granddaughter in marriage has a comparable effect in counteracting the dispersive consequences of virilocal marriage in a matrilineal society and replicating in alternate generations a pattern of patrifilial links (here through daughters) that maintains a correspondence between men, their matriline and locality. Lele society is more fluid, its members more mobile than Kongo society, but in both cases the crosscutting of kin and local relations is picked up and expressed in generational differences: the important status accorded to Children and Grandchildren in a Kongo village has its parallel in the intergenerational tensions and alliances of a Lele village (Douglas 1963: 80ff).

Figure 2. – Lele Preferential Marriage and Village Continuity

Figure 2. – Lele Preferential Marriage and Village Continuity

39Another very well described society, the Plateau Tonga of presentday Zambia, provide a third example of the way amongst the societies of the matrilineal belt generational grouping acquires a special significance. With the Tonga we move to the eastern border of these societies, where there is evidence of changes in the mode of descent reckoning. In certain respects Tonga kinship organization closely recalls that of Kongo, notably in the way that each Tonga stands in a defined relationship to four distinct matrilineal groups: his own of which as sister’s son he is a full member (basimukowa), that of his father to which he stands as child (basyanausi), and the two groups of his mother’s father and father’s father to which he stands as grandchild (each basikulu). The same four groups (although known, of course, by different terms) are said to form the “four corners” of a Kongo’s kinship universe (MacGaffey1970), and much the same can be said of them for Tonga. A Tonga’s claim as “child” on the support of his “fathers” (basyanausi) is however stronger than his claim as “grandchild” on the support of his “grandfathers” (basikulu). For example, the “fathers” take a direct interest in the marriage of a “child” and share in providing or receiving bridewealth–a situation that gives a distinctly bilateral character to Tonga kinship. “Grandfathers”, although represented and participating in a marriage, are less directly responsible for the affairs of a “grandchild” (Colson 1951: 144-145, 1958: 48-60). Even so, a Tonga’s field of kin-relationships is made up in very much the same way as it is for Kongo of a network of patrilateral personal ties leading to a series of established matrikin groups. Where Tonga differ from Kongo is, firstly, in the non-localized character of their matrilineal groups, and, secondly, in the strongly solidary, non-segmentary unity of these groups. Strictly, Tonga matrilineal groups are not descent groups at all but groups of uterine kin who maintain a generalized loyalty and responsibility to co-members and their children, without internal differentiation. This feature is stressed by Colson (1958: 18):

“When a man or woman dies all members of his or her group have a right to share in the inheritance. Uterine siblings and their descendants have no greater rights than any other… The bridewealth of the woman is not necessarily taken by his own mother’s brother to the exclusion of classificatory mothers’ brothers… Men accumulate property for their bridewealth by approaching any matrilineal relative whom they think likely to help…”

  • 4  Colson speaks of such alternate generations as the “senior” and “junior” branches of–as it were–a (...)

40Undifferentiated by particular lines of descent, these matrilineal groups are however sharply differentiated by generation. This applies not only to the groups patrilaterally related in successive generations but to all uterine kin. A man stands to the fellow-members of his own matrilineal group either as a “brother” to those of his own generation or as a “mother’s brother” or “sister’s son” to men of the two generations adjacent to his own. Moreover, alternate generations are terminologically and normatively merged, so that a man’s sisters’ daughters’ sons (classificatory or real) are identified with his brothers and count formally as belonging to his own generation 4. The rights and obligations within a matrilineal group closely follow this internal structuring by generation: the “mother’s brothers” of the group have clear authority over their “sisters’sons”, their junior adjacent generation, who should respect their seniors while looking to them for support; “brothers” of the same (and alternate) generation stand as equals to each other, and also to some extent as rivals (Colson 1958: 51ff). Such generational structuring of relations within a matrilineal group is consistent with the normative patterning of behaviour between a patrifilial “child” or “grandchild” and his father’s or grandfather’s matrilineal groups, the “fathers” (basyanausi) having, as we have seen above, a formal responsibility and authority in the affairs of their “children”, that the “grandfathers” lack. All these relationships are between men, but a closely comparable pattern exists in the relations of women, with the exception that within the matrilineal group “grandmothers” and “granddaughters” are separately recognised, although they are equated with each other as “sisters” and occasionally addressed reciprocally as such (Colson 1958: 52-53).

41The pervasive and precise articulation of generations that Tonga society evidences, with adjacent generations sharply distinguished and alternate generations merged, corresponds closely to the general pattern of generational separation and alignment previously described for eastern Bantu, including now both patrilineal and matrilineal societies. It is worth noting that for both Tonga and a number of other of these eastern Bantu societies, matrilineal and patrilineal, a preferred form of marriage was often with crosscousins (Colson 1958: 325-327; Bruwer & Van 1958; H. Kuper 1947: 95). The common effect of such marriages is to re-activate what may be called “extra-descent group” ties that might otherwise be lost.

42In examining briefly a few selected societies from the range of those situated between eastern Bantu and those in the northwestern corner of Bantu Africa, where Bantu-speakers border with the remaining West African Niger-Congo-speaking peoples, my aim has not been to trace through in any detail the variably interrelated patterns of descent, localized and generational grouping, but rather to suggest how, on the wider canvas of broad cultural patterns and historical developments, an apparent contrast and disjunction –the striking difference between West African and eastern Bantu societies in the institutional recognition of generational alignments–can be mediated and possibly explained in terms of a set of continuously modulating principles. The critical step in this sequence is that whereby the dual affiliative ties that develop within a matrilineal society practising virilocal marriage come to be identified with generational differences and alternations–the crosscutting patrifilial relationship becoming identified with generational differences, with the possibility of re-activating the affinal (and thus patrifilial) relationship in alternate generations. I am arguing that this step could have occurred in the linguistic and cultural expansion of the Bantu, from a branch or branches of western Bantu where generational differentiation already existed and could be further developed.

The Historical Evidence

43Relevant here is Christopher Ehret’s recent important book, An African Classical Age, which treats essentially the same problem, the emergence of eastern and southern Bantu from those of the west. This is not the place to attempt a review of the book, which draws upon data from historically oriented ethnolinguistics. This is different from the approach of the present article, which is based rather upon comparative structural sociology, but the book does provide a useful framework for discussion.

44Focusing upon the last millenium BC, the story that Ehret tells is of a slow and gradual migration of a people from the rain forest of the eastern Congo, whom he terms the “Mashariki”, first to the uplands west of the Western Rift Valley, across the narrow Rift Valley, and then eventually to the savannah plains of the interlacustrine region. The Mashariki were not a single group (in the way that tribal traditions often tell of migrations) but a dispersed congeries of linguistically related migratory groups exploring new territory. The Mashariki emerge in the area of the Western Rift Valley between Lakes Tanganyika and Ruiru around the beginning of the first millenium BC and separate initially into two branches, the Kasakazi and Kusi, who further diverge and separate, the Kasakazi to the north and the Kusi to the south, from whom can be traced all or most of the Bantu linguistic groupings of the present.

45This history is by no means solely linguistic: Ehret is concerned with actual people occupying particular environments, with their modes of livelihood, social organization and religion that he seeks to establish through the analysis of common root vocabularies. In a chapter on Aspects of Social History he is happy to treat kin and authority among early Mashariki and on the basis of shared words to make statements, for example, that “early Mashariki communities recognised two levels of the preternatural, God and the ancestors” (pp. 146f, 158). It must be said however that the historical, genetic framework that he first establishes is far more secure than the adduced concomitant environmental conditions, social institutions and religious ideas, that he adds to it. Some of my sociological doubts will be apparent later.

46Nevertheless, Ehret’s broad account is helpful in the present context and provides a number of suggestive leads. The fundamental datum of the book, the emergence of the Mashariki, can be used as the axis of my own enquiry. This somewhat shifts our attention from the matrilineal peoples discussed in the last section to the area of eastern Congo adjacent to but north of the matrilineal belt, whose peoples nevertheless share a number of common features. In terms of existing societies, this is a region of uncertain and varying principles of social grouping: matrilineal changing to patrilineal, no extended lineal grouping or variant lineal groupings playing against each other and/or residential grouping in a constant shifting mix. And, as one would expect, categorical generational differences also play a part. It is an area where the rain forest changes to savannah woodland in the eastern uplands towards the Western Rift Valley. Biebuyck, quoted by Vansina, stresses the extraordinary mix of groups and types and writes of “cultural mixtures, transitional among larger ethnic identities. That kind of interlinkage [which] makes an immense region of eastern Zaire a cultural continuum” (Vansina 1990: 178; Biebuyck 1986: 266). Vansina emphasises the role of the Mwami society in organizing and integrating this mixed population, where affinal and cognatic links are critical, especially via the “maternal uncles” stretching back through successive generations, where intergenerational seniority is significant (Biebuyck 1973). A comparably mixed population exists southeast of Lake Tanganyika, where both matrilineal and patrilineal ties are recognised but where there is a movement towards patrilinearity which incorporates the notion of generational succession (Willis 1966: xi). Amongst the Pimpwe, for example, a man belongs to the kin-group (uluko) of his father’s father or son’s son, but not to that of his father or son, resulting in a perpetual generational alternation of the two uluko names of his patrilineage.

47It would be wrong to extrapolate backwards from such near contemporary accounts to look for what may have occurred in the distant past, but it serves to remind us of the continuities in social structure which as they exist in the present must surely have existed in the past.

48To return to the axial Mashariki: two major factors underlie Ehret’s account. Both figure large in his analysis. The first concerns the very different environment the savannah Maskariki came to inhabit in comparison to the wet forest lands of the western Bantu. Both Ehret and Schoenbrun, in a book that focuses upon the long-term history of the interlacustrine area (Schoenbrun 1998), document this well. With the different grasslands environment came changes in patterns of residence, type of agriculture and general mode of livelihood. Notably, the ecologically defined basic kin-group that formerly shared with others a composite village-settlement was now more widely dispersed and more sharply defined. An entirely new style of house emerged, denoted by a entirely new root-term *-yumba (Guthrie 1967-1971 (4): 207, CS 2168) and we shall see that a term borrowed from peoples already living in the area was used to describe the “home” or “homestead”, *-ka (Schœnbrun 1997: 91), so constituted. Schœnbrun (1998: 92-93) puts the matter well:

“Great Lakes Bantu speakers created wholly new architectural forms for their houses and for the enclosures that surrounded them. The distinctive square house built by Bantu in the Congo Basin, with its panelled and gabled roof, gave way in the east to a round style of house, with a thatched roof… As their settlements expanded into the more open woodland plains of the Kivu Rift Valley they began to use the new building materials of these areas… The most distinctive attributes of these new homes were the fences surrounding them… These tall fences, with their main gateways… separated domestic space from the rest of the landscape of fields, pastures, and wilderness. They protected people and domestic animals from predators and they provided privacy for individual families. By encoding gendered and aged spaces within the homesteads, their internal layout choreographed interactions between family members and guided contacts with outsiders.”

49In discussing the social organization of the Mashariki, Ehret (1998: 151) points to the matrifocal connotation of a house in African societies, where a “house” is typically a woman’s domain, and from this infers that the early Mashariki could have been matrilineal. But this is to miss the context of African descent idiom, where matrifocal descent can operate alone (“matrilineally”) but commonly distinguishes different mother-derived groupings within a wider patrifocal and usually patrilineal grouping. The “wives” of a common husband provide the model for segmentary differentiation in much the same way as the different sons of a common father often do. The basic conceptual model is then very much that of a unitary homestead owned by a man but with different wives, each with their own house, and this reappears throughout eastern Bantu, and stands in contrast to Vansina’s “House” model of proto-western Bantu in which the big man “father” gathers to himself variably related junior kin.

50Can we assume that with the more sharply delimited homestead/family a clearer distinction was needed between kin of different generations within it? It is reasonable to do so. The solidarity of dispersed homesteads within the wider locality was important (although more contingent and less immediate than that of Houses within a Congo nucleated village) but the homestead/family needed to replicate itself and could do so primarily by the cyclical process of splitting and reforming at each generation. For this the procreative potential of parents had to be sharply distinguished from that of their children: the “respect” and “avoidance” relationships came to play an important part in maintaining the generational cycle of the eastern Bantu family and homestead. I discuss in the next section the distribution of the root-term for “grandchild” that gives some support to the emergence of generational differentiation within the homestead.

51Apart from the ecological changes Mashariki needed to adapt to, there is a second major factor, evident from Ehret’s account, that impinged upon their development: this is their progressive encounter in their newly utilized savannah territory with other non-Bantu peoples. These encounters are treated by Ehret in sequence, involving the Central Sudanians, Eastern Sahelians (each different branches of the wider Nilo-Saharan linguistic group), Southern Cushites and finally Southern Nilotes (again, a distantly related Nilo-Saharan grouping). We have already noted one significant adoption from the Nilo-Saharans, which is the use of the root-term *-ka, to denote the homestead or “home” of a family. This term, which belongs to a wider set of related words (Schœnbrun 1997: 89-91), can be traced from the Great Lakes area through to the southern Bantu, but is found only in this eastern half of Africa. It should be noted that it co-exists for eastern Bantu with another root-term for village, *-gi or -ji (Guthrie 1967-1971 (3): 217, CS 818, (3): 247, CS 936) which is found through all Bantu. Ehret’s own map of “the mosaic of cultural interaction” offers no evidence of generational differentiation deriving for Mashariki from non-Bantu contacts.

52Even so, further east and at a later period than the Mashariki, there has very evidently been a process of interethnic layering and assimilation that involves the recognition of formal generational differences and sequencing. The “gada-system” of the contemporary eastern Cushites, the age and generational set systems of some southern Nilotes, and the generation classes of certain East Nyanza, Luyia and Thagiicu (Kikuyu, Meru) Bantu, clearly share historical links, not least in the reappearance of sequential names in the appropriate order amongst distant and different peoples (Ruel 1962b, 1997: 250-1; Ehret 1971: 46). Unfortunately there has been a tendency in the literature to assimilate such named generation groups to the category of age set system (e.g. Sutton 1990: 30). As a study of contemporary East Nyanza societies shows (Bischofberger 1972; Ruel 1997) the principles of recruitment by generation at birth and to an age-set at initiation are distinct and have a different socio-logic. They can combine, as they do in a number of Nilotic societies, but they also exist separately (as they do for all the Bantu societies listed above), would seem to have different historical links and certainly have different implications for the larger structuring of their society. For this reason I find it difficult to accept Ehret’s discussion of age-sets and rites of passage among the early Mashariki (Ehret 1998: 155-158). I would question the gloss of “age set” given to the root-term *-kula (admittedly derived from Guthrie) and challenge the suggested existence of age sets and adolescent initiation by circumcision among Mashariki, for which the evidence is very thin. But if age-sets did not exist among the early Mashariki was there a form of institutionalised generation grouping?

53The evident later assimilation of named generation groups by certain eastern Bantu here sets a puzzle. The puzzle is this. For the Bantu there is a clear pattern of adoption: generation classes are found among some peoples (Kuria and their southern congeners, also Kikuyu, Embu and Meru) and not amongst others otherwise related (Gusii in the case of Kuria, Kamba in the case of Kikuyu). But those who do not have generation classes still share the eastern Bantu characteristic that is the subject of this article, the alternating differentiation of generations. It is as though the institutional system of classes is a mere dressing: the actual, alternating separation of generations already exists, e.g. for Gusii (as cited above), for Kamba (Middleton 1953), as well as for the many eastern and southern Bantu societies already cited. Is it possible that some older, more widespread influence from Nilo-Saharan societies impinged upon the Mashariki? There is at present no answer.

The Emergence of the “Grandchild” among Eastern Bantu

54Arguments from root-terms are hazardous, but drawing upon Guthrie’s Comparative Bantu, I would suggest that there is one root-term that can be put alongside the two terms already adduced, which enables us to trace something of the way kin-generations became distinguished among eastern Bantu.

55Guthrie lists three comparative series of words with the meaning of “grandchild” (Guthrie 1967-1971 (3): 255, CS 963, (4): 178, CS 2049, 2050) which are closely related in form and which he derives from a common root, *-yijokodo. He records no other comparative series with the same meaning. The interest lies in the distribution of the series. Each of the three series (combining to form an “osculant cluster”) has a few reflexes in western Bantu but by far the greatest number of reflexes–forty against six–are found in eastern Bantu languages, from Nyoro in the north to Sotho and Zulu in the south. Among eastern Bantu almost without exception the word for grandchild is some easily recognised version of omwijukuru (Ganda), nzukulu (Kamba), ng’wizukulu (Sukuma), ndzukulu (Nyanja), nsukulu (Yao) or mzikulu (Xhosa).

56All this is in accord with the empirical contrast sketched at the beginning of this article. But unlike the two root-terms already adduced, *-yumba and *-ka, there are these few cases of instances appearing in western Bantu. It becomes then critically important where these are located. They are:

“One instance (Ki-Mbundu) on the western coast amongst a people close to and cognate with the Kongo cluster considered above (p. xx);

three instances (Lwena and dual Luimbe) living east of the above and forming part of the matrilineal belt of peoples extending from the west;

two instances (Lunda and Luba-Lulua) from one of the most easterly of the western Bantu linguistic groups where patrilineal and matrilineal principles co-exist and can be related to an alternating sequence of generations in kinship terminology, grandparents being distinguished from parents, and great grandparents being terminologically identified with parents” (McCulloch 1951: 20).

57This evidence cannot be pressed too hard but the indication is clear and is summarized in Guthrie’s own comments (1967-1971 (4): 178, CS 2049) that the origin of the term would seem to lie in western Bantu but its great extension took place in the east.

  • 5  The Norman-French “grand” added to the Anglo-Saxon “fader” provides a parallel. Note too that the (...)

58One final connecting point. It will be noted that the element *-koro or *-kulu that appears in the root-term for grandchild (written as *-kodo by Guthrie) can be associated with a number of other terms for senior person, important man, grandfather, elder and ancestor. There would seem to be general agreement (Guthrie, Ehret, Schœnbrun) that this element can be traced back to the very widespread proto-Bantu root verb *-kora, “to grow up”, “to mature”. In noun or adjectival form, it very frequently appears in words denoting a senior or important person, elder or ancestor. It also appears in compound kinship terms where one of the root terms for “father”, *-se or *-so, is qualified by the same element to denote “grandfather”, as in the kin-terms sekoro or sokoro. Although it still leaves unexplained the middle element, the final -koro in *-yijokodo would seem to have the same function in distinguishing a kin-category by generation 5. It also makes for a rather subtler reading of the Bantu notion of “growth” than the simple one of “growing up”, “becoming an adult” that is sometimes accorded to it (Ehret 1998: 156). Certainly among East Nyanza Bantu-speakers (Zanaki, Ngruimi, Kuria, Gusii) the word erikura or irikora refers to a “generation” (Bischofberger, Ruel) and rests upon elaborate notions of achieving a social maturity not through one’s age, or one’s children, but through one’s children’s children and their descendants.


59This paper has attempted to highlight a broad contrast and to provide a tentative answer to the puzzle that it poses. Its approach has been to treat the puzzle as one of discontinuity in cultural forms that otherwise can be considered as continuous. Cultural continuity does not mean uniformity, but rather variability of an interrelated, systematic kind whereby a series of correlated changes occur in linked sequence generating forms apparently different, but developmentally and historically derivable from those they do not resemble. Anthropologists, with their commitment to intensive local studies, are too frequently prone to see African cultures as a kind of patchwork quilt, each individually studied culture with a local integrity, to be compared in its own right with others, against the background of a deduced, universalist structural logic. In taking this course we fail to recognise wider cultural themes and preoccupations, and we ignore altogether the complex and varied patterns that those wider themes weave (MacGaffey 1983; A. Kuper 1982; Kuper & Van Leynseele 1978).

60To establish such variability in continuity it is useful to focus upon single elements, although these obviously need to be looked at in context. Generational alignments provide such a focus, and in considering them this paper has throughout been concerned with the particular social structural context that gives them the significance that they have. In itself, the contrast between West African societies and the eastern Bantu is slight and perhaps trivial, but the puzzle it poses leads to significant issues, both of a methodological and substantive kind.

61Methodologically, the paper has been concerned to establish a continuity that is both sociological and historical. It has not been argued that the crosscutting effect of matrilineal descent and patrilateral affiliation necessarily produces a sharpened awareness of generational standing, but simply that this could historically have happened and that there are good structural reasons why such a development might have occurred in such a way. Again, methodologically the paper has had to move between considerations of formal norm and actual practice, the life of people in society entailing always both. Thus, the simple, categorical differentiation of “grandparents” from “parents” does not as a cultural fact tell us much: we need to know the situational usage of each category, how it interacts with other forms of grouping and the consequent force it acquires for patterns of behaviour. This is not to oppose “cultural norm” with “actual behaviour” but to be constantly aware of their interaction. The Bantu term for “grandchild” emerges as a cultural category with a special significance, but it acquires this general significance since grandchildren (it has been argued) do in fact by virtue of their structural position within the society acquire qualities that elsewhere, in other Bantu cultures and at other times, they lack.

62Substantively, the pattern of generational differentiation and alignment in the eastern Bantu cultures can be connected with other central themes in their social organization and culture. The due ordering of generations in their sequence and growth to maturity features in much ritual, not least that of initiation (Droogers 1980; Richards 1956; Ruel 1997; Swantz 1970, amongst many others). It is critical to the domestic organization of homesteads and some settlements (Huber 1973; H. Kuper 1947; A. Kuper 1982; Stefaniszyn 1964; Turner 1955; Wilson 1951). It relates closely to ideas concerning ancestors, who in this region are represented less commonly as parents, authority figures, than as grandparents, seeking to be remembered in a way that gives them a continuing identity in the life of the living (see especially Stefaniszyn 1954; Holleman 1953; Krige & Krige 1943; Lancaster 1977; Tanner 1958; Colson 1962). There remains much work to be done to explore how these interrelated themes and features of social organization may be seen to be part of a common eastern Bantu or indeed Bantu culture or ideology.

Haut de page


Alexandre, P. & Binet, J.
1958   Le groupe dit “pahouin” (Fang, Boulou, Beti) (Paris: Presses Universitaires de France).

Bascom, W. R.
1969   The Yoruba of Southwestern Nigeria (New-York: Holt, Rinehart & Winston).

Bastin, Y., Coupez, A. & Halleux, B. de
1983   “Lexicostatique des langues Bantoues”, Bulletin de l’Academie royale des Sciences d’Outre-Mer, 27: 173-99.

Baumann, O.
1926   “Vaterrecht und Mutterrecht in Afrika”, Zeitscrift fur Ethnologie, 58: 62-161.

Bennett, P. B. & Sterk, J. P.
1977   “South Central Niger-Congo: a Reclassification”, Studies in African Linguistics, 8: 241-273.

Biebuyck, D.
1973   Lega Culture: Art, Initiation, and Moral Philosophy among a Central African People (Berkeley: University of California Press).
1986   The Arts of Zaire (Berkeley: University of California Press).

Bischofberger, O.
1972   The Generation Classes of Zanaki (Fribourg: The University Press).

Bohannan, L. & Bohannan, P.
1953   The Tiv of Central Nigeria (London: International African Institute).

Bourdillon, M.
1982[1976] The Shona Peoples. Revised edition (Gweru: Mambo Press).

Bruwer, J. P. & Van, S.
1958   “Matrilineal Kinship among the Kunda”, Africa, 28: 207-224.

Colson, E.
1951   “The Plateau Tonga of Northern Rhodesia”, in E. Colson & M. Gluckman, eds, Seven Tribes of British Central Africa (London-New York: Oxford University Press for the Rhodes-Livingstone Institute): 94-162.
1953   “Social Control and Vengeance in Plateau Tonga Society”, Africa, 23: 199-212 (reprinted in Colson 1962).
1958   Marriage and the Family among the Plateau Tonga of Northern Rhodesia (Manchester: Manchester University Press).
1962   The Plateau Tonga of Northern Rhodesia: Social and Religious Studies (Manchester: Manchester University Press).

Dalby, D.
1975-1976 “The Prehistorical Implications of Guthrie’s Comparative Bantu”, Part 1 “Problems of Internal Relationship”, Part 2 “Interpretation of Cultural Vocabulary”, Journal of African History, 16: 481-501 & 17: 1-27.

Douglas, M.
1952   “Alternate Generations among the Lele of the Kasai”, Africa, 22: 59-65.
1963   The Lele of the Kasai (London: Oxford University Press for the International African Institute).

Droogers, A.
1980   The Dangerous Journey: Symbolic Aspects of Boys’ Initiation among the Wagenia of Kinsangani, Zaire (Paris: Mouton).

Drucker-Brown, S.
1982   “Joking at Death: the Mamprusi Grandparent-Grandchild Relationship”, Man, 17: 714-727.

Ehret, C.
1971   Southern Nilotic History (Evanston, ILL: Northwestern University Press).
1982   “Linguistic Inferences about Early Bantu History”, in C. Ehret & M. Posnansky, eds, The Archaeological and Linguistic Reconstruction of African History (Berkeley: University of California Press): 57-65.
1998   An African Classical Age Eastern and Southern Africa in World History, 1000BC to AD 400 (Charlottesville: University Press of Virginia; Oxford: James Currey).

Ehret, C. & Posnansky, M., eds
1982   The Archaeological and Linguistic Reconstruction of African History (Berkeley : University of California Press).

Fernandez, J. W.
1982   Bwiti: an Ethnography of the Religious Imagination in Africa (Princeton, NJ: Princeton University Press).

Forde, D.
1950   “Double Descent among the Yako,” in A. R. Radcliffe-Brown & D. Forde, eds, African Systems of Kiship and Marriage (London-New York: Oxford University Press for the International African Institute): 285-332.

Fortes, M.
1949   The Web of Kinship among the Tallensi (London-New York: Oxford University Press for the International African Institute).
1970   Time and Social Structure and Other Essays (London: The Athlone Press; New York: Humanities Press).

Gluckman, M.
1950   “Kinship and Marriage among the Lozi of Northern Rhodesia and the Zulu of Natal”, in A. R. Radcliffe-Brown & D. Forde, eds, op. cit.: 166-206.

Goody, E.
1973   Contexts of Kinship (Cambridge: Cambridge University Press).

Guthrie, M.
1967-1971 Comparative Bantu: an Introduction to the Comparative Linguistics and Prehistory of the Bantu Languages. 4 vols (Farnborough: Gregg Press).

Hammond-Tooke, W. D.
1981   Boundaries and Belief: the Structure of the Sotho Worldview (Johannesburg: Witwatersrand University Press).

Heine, B.
1979   “Some Linguistic Observations on the Early History of Africa”, Sprache und Geschichte in Afrika, 1: 37-59.

Hilton, A.
1983   “Family and Kinship among the Kongo South of the Zaire River from the Sixteenth to the Nineteenth Centuries”, Journal of African History, 24: 189-206.

eondonm tp6d7bcla1sle Siversity Presilpan stynm c Tt3oI s-varbr />1ss): 57mall-cfont-v;">oody

Seven Tribes of British Central Africa (London-New York: Oxfories”, Journal of African History, 24: 189-206.

Gigebliographie Je">EigebliographJn>-T,ae EQu So
1949   The Web of Kinship among the Tallensi (London-New York: Oxford UnClasses of Zanaki (FriboKRle E
1949  Bon-Ne: Rbr le1973>E (FriboKRruwer, J. P Lersrand University Press).

A<">Goy2 (reprin‘ion toExphis/>197  “Alternate Gene48f A3ers5ty Press for the International AKRG1949   The Web of Kinship among the Tallensi (London-New York: Oxford UnClasses of Zanaki (FriboLersrand University Press).

ernand  Upsali, M.a UnClasses of Zanaki (FriboLersrand University Press).

(BerkeKikuyur />1962  Polr El-cap: Oxforine<
1949   The Web of Kinship among the Tallensi (London-New York: Oxford UnClasses of Zanaki (FriboLersrand University Press).

-Bof I  “Alternate Gene47Uni2Cen4 Afrika”, Zeitscrift fur ELersrand University Press).

-B199Kikuyurbe Tae 1903.font-v3Linguis949   The icostatcneration Classes of Zanaki (FriboLersrand University Press).

-B (BerkeNyhisrk19mun amoe EKeny R.
1em>The s an thleyan stylene (FriboMersrand University Press).

, W. D, : 199-212 (reprinted in ColsooI br />caps;">he Prehist”, Part 2 “Interpretation of Culteenth73-187an People (Berke, J. WnInternatieamoe E-caps;">luckmarn and Sokma (FriboMersrand University Press).

Bou  , A.< AtOxfolang;">ilton

& D..pan stylemuth Cenatio-of InternId1ss-caps;"haviParg;">iltinsps;">tcnGiantsoe EKeny R.
(FriboMersrand University Press).

-B (Manchester: MaMersrand University Press).

FGor /r />1950   />19lgi of rk, O.1953   The Tiv of Central Nigeria (London: InternMersrand University Press).

font-variathe Kr />-1ss): 57he Pr;tonieclateiltEmbu">Meru">MbibD, Chuk/>Mwimbi, arak/> (reprinKbb: Symbeny R.
1953   The Tiv of Central Nigeria (London: InternMersrand University Press).

Seondo">Ehretspan style="fle="fontariant:sChaggs2  Meru SymTanzheraR.
1953   The Tiv of Central Nigeria (London: InternMersrand University Press).

or (reprin:small-e Archaeologica The McGrawe All UnClasses of Zanaki (FriboNersrand University Press).

luckman, eds:e and Social S UnClasses of Zanaki (Fribon style="font-variant:small-caps;">hretThe nll-caæ;">oy;">oody,  The
1973   Contexts of Kinship (Cambridge: Cof Northern Rhodesia and the Zulu of Natal”, in A. R. Radcliffe,among the Lozi of Northern Rhodesia and the Zulu of Natal”, in A. R. Radcliffe-Brown & D. F19mcapa9 theariant:small-caps;">ouglasadcliffe-Brinceton University Press).

orde, eds, African Systems of Kiship and Marriage (London-New York: Oxford Kinship among the Kunda”, he Prehist the Lozi of Northern Rhodesia and the Zulu of Natal”, in A. R. Radcliffe-Brown & D. FGosnans rkcepnt:salCateg79 W/sps? Ar lat amonhe yasonariant:sTerminolgonship”, Part 2 “pan>,e deaAoutrop;">oEl-cap: Oxfornd Mutte92nth57-176n>-Bischofbe  “Alternate Gene32nth4-37an People9e="fontaria;">ookm>Thtu962  pan securison“Lifdeeides:eie Je"BrAll Un Peop200le="font-vriansuria H-vaonta TongSpasm2  the nship”, Part 2 “, J. WnI Bwite Gene30echt 85Bennett-He">G, both (Londo2  Dta “Social Control and44ech5-70Bennettdeu of Internal RelatioR.
, M.
fapiszynbliographBn (LonrR=-exed ne “Social Contn Plateauistory oVeng31-146n>-B1962  htu962Lifd “pan>A6] Marriage and the F>, eds, African Systems of Kiship and Marriage (London-New York: Oxford Kinship among the Kunda”, ennettennettG-Lnniversity Press for Riru962  ers19Trhis/(LondonZarColrnatieamd the F>und: Gleerup Kinship among the Kunda”, HThe DCibDolsouth CeSukut:s ,mTange y>, O.
1lternate Gener5Uni25-23 Afrika”, Zeitscrift fur Ee">Hr bliographVo thNdebbu VAllngpan>, W. DO.ruwersinaliographiJBoe Eon, eds:e (Charlottesville: University Press of VirgiW in Afrika, 1: 37-59.

oraps;">ascom1ss)3 bbian, eds:ee (London-New York: Oxford Kinship among the Kunda”, 1: 37-59.

B-He">G1967nyspan lat amonNyakyusia P-vAllngp<
1949   The Web of Kinship among the Tallensi (London-New York: Oxford UnClasses of Zanaki
(FriboW in Afrika, 1: 37-59.

, C.
1971   Southern Nilotic HistCla

Haut de page


1  When Radcliffe-Brown (1950: 28ff) came to discuss in an African context the structural principle of the merging of alternate generations (which he originally derived from his study of Australian kinship systems) the examples that he cited in illustration of the principle were in fact all drawn from eastern and southern Bantu. Yet in the same book Forde’s account of Yako kinship shows a significant merging of adjacent generations–grandparents with parents, grandchildren with children (Forde 1950: 319-320, also Figure 15).

2  On this point Nyakyusa rules are particularly explicit and are cited as the reason for the residential separation of adjacent generations; see Wilson (1951: 159).

3  This, of course, is one aspect of the “matrilineal puzzle” first identified by Audrey Richards (1950). Richards’ article focuses upon the conflicting claims of father and mother’s brother and the varying rules of residence that resolve these claims. But she is clear about the structural generational consequences: see later references in the text and pp. 218, 241, 249 of her article.

4  Colson speaks of such alternate generations as the “senior” and “junior” branches of–as it were–a single generation.

5  The Norman-French “grand” added to the Anglo-Saxon “fader” provides a parallel. Note too that the subsequent English grandchild follows a socio-logic that contrasts with the more accurate French petit-fils or petits-enfants.

Haut de page

Table des illustrations

Titre Figure 1. – Kongo Forms of Preferred Mariage
Fichier image/png, 39k
Titre Figure 2. – Lele Preferential Marriage and Village Continuity
Fichier image/png, 38k
Haut de page

Pour citer cet article

Référence électronique

Malcolm Ruel, « The structural articulation of generations in Africa », Cahiers d’études africaines [En ligne], 165 | 2002, mis en ligne le 22 novembre 2013, consulté le 14 décembre 2017. URL :

Haut de page


Malcolm Ruel

Clare College, University of Cambridge.

Haut de page

Droits d'auteur

© Cahiers d’Études africaines

Haut de page