This page is hosted on AFS file server space, which is being shut down on November 13, 2018. If you are seeing this message, your service provider needs to take steps now. Visit for more information.
Robert E. Daniels
A paper presented at the 75th Annual Meeting of the American Anthropological Association, November, 1976, Washington, D.C.
The cyclical age-set system of the Kalenjin-speaking peoples in Kenya is based on seven or eight age-set names which recur in fixed order. The principal ethnographers of the closely related Nandi and Kipsigis sought to determine the number of years between age-set transitions in order to establish a chronology of precolonial events. Their analyses differed, however, and the controversy between them remained unresolved. This paper reviews their accounts and offers a new interpretation based on fieldwork among the Kipsigis.

Prior to pacification in 1906, the change of age-sets was punctuated by the saget ap eito ceremony. Hollis (1909) stated that this occurred every seven and a half years among the Nandi; Huntingford (1953a) calculated fifteen year intervals. Central to Huntingford's account were the use of a plant, setiot, which flowers every seven or eight years, as a signal to begin initiations, and a ritual expert, orgoiyot, who sanctioned, and thus synchronized, the rituals. But for the Kipsigis both Orchardson (1961, writing in 1929) and Peristiany (1939) stated that age-set intervals varied, and denied the significance of both setiot and the orgoiyot. Peristiany explained that the decision to end an age-set arose from the interplay of conflicting interests among individuals of various age-sets. This analysis, however, does not explain how such face to face negotiations can produce a coordinated transition in an acephalous society of tens of thousands of people, particularly since the loss of the saget ap eito.

A network model of Kipsigis social organization is proposed in which homesteads, likened to neurons, are connected by agnation and especially affinity. Local communities are exceptionally dense clusters of interrelationship. The impetus to change age-sets arises independently in a number of communities, and is communicated into the wider network. At some point enough such decisions sum to excite those remaining areas where local opinion is still approaching the threshold, and the whole system is 'convulsed' into a change to age-sets. Evidence from one community indicates how local decisions are first made tentatively and confirmed if similar events occur elsewhere in the network.

This model is used to clarify many seemingly inconsistent points in earlier accounts and, more generally, to suggest that many features of Kalenjin social structure, such as descent groups and traditional leadership, are, like age-sets, characterized by ambiguous boundaries and should be understood in terms of network dynamics.


East African age-set systems, into which men are inducted by elaborate initiations, are a topic of long standing in anthropological literature and a perennial favorite in introductory texts in much the same fashion as the potlatch and the kula ring. Maasai initiations and age-sets were cited by van Gennep in 1908. The following year Hollis's pioneering work on the Nandi presented the first description of the cyclical age-set system characteristic of a series of localized groups in Kenya and Uganda speaking closely related dialects: the Nandi, Kipsigis, Keyo, Tugen, Marakwet, Sabaot/Sebei, Pokot, and others, whose descendants are now known collectively as the Kalenjin.1

The age-set system of the two largest Kalenjin groups, the Kipsigis and the Nandi, is based on seven age-set names which recur in fixed order (in the Kipsigis dialect these are Kablelach, Kimnyige, Nyongi, Maina, Chumo, Sawe, and Korongoro).2 A complete cycle takes well over a century. Thus at any one time the adult male population contains members of five, or possibly six, age-sets (singular ibinda, plural ibinwek), ranked in seniority.3 As the next set is constituted it takes the name, and something of the reputation of the formerly most senior set, all of whose members have long since died. Currently the Kipsigis are initiating members of Korongoro; the next age-set will be Kablelach. The basic principle is well depicted by the familiar chart showing the age-sets following each other around in a wheel (e.g. Evans-Pritchard 1939:xxviii; Huntingford 1953a:55).

Given the amount of ethnographic attention which this system has received, one might expect that its operation and functions have been thoroughly analyzed. In fact, however, the principal ethnographers of the Kipsigis and Nandi strongly disagreed with each other's analyses of the fundamental dynamics of the system. Since then both initiation practices and the interests of anthropologists have changed, and the controversy has subsequently remained unresolved. This paper reviews the earlier material and offers a new interpretation of age-set processes in light of more recent fieldwork among the Kipsigis.4

The Kipsigis

The homeland of the Kipsigis, now Kericho District, is the largest and one of the most fertile of the several highland zones occupied by Kalenjin-speaking groups. In the nineteenth century the Kipsigis were settled pastoralists and hoe agriculturalists living in neighborhoods of 20 to 40 small homesteads concentrated in the hills at elevations between the high forest, occupied by small groups of Okiek ("Dorobo"), and lower grasslands, where the Kipsigis maintained cattle camps. Precolonial density in the core area was relatively high, perhaps 15 homesteads or 150 people to the square mile (Pilgrim 1961:36). During this century the Kipsigis have experienced tremendous population growth. In 1907 the Kipsigis population was thought to number about 68,000.5 While this may have been an underestimate, more recent census figures suggest that the current population exceeds 600,000.6 In large part this increase has been offset, at least until recently, by expansion of the domestic settlement pattern of the core area into all corners of the district. Thus the Kipsigis have avoided the problems of overcrowding now typical of the Kenya highlands, while adapting to rapid economic development (Manners 1967; Daniels 1980) and have maintained a great deal of their traditional domestic social organization. Polygyny is common, and inheritance (primarily involving the family herds) is from father to sons according to the house-property complex. One son in each set of full brothers remains at his parental homestead while the others establish homesteads in other communities according to their opportunities. Larger descent categories, or 'clans' (ortinwek), of which there are over a hundred, are thus dispersed and intermingled in each community. These clans are exogamous and serve to define expectations of mutual support, but have no corporate functions beyond shared responsibility for bloodwealth and related matters.

Cutting across all this are the age-sets. Adolescent boys enter the system through initiations. Following circumcision, they are given secret instruction and training emphasizing pastoral and military virtues, and emerge as young men or 'warriors' (murenik) to take their place in the most junior age-set. In contrast to some other East African age-set systems, murenik are not barred from marrying and do so, usually when in their early 20s, according to family resources. A man cannot, however, marry the daughter of an age-mate (i.e. a member of his ibinda). Similarly, although an individual's social career is measured by personal and familial events rather than by formal age grades, common membership in an ibinda implies mutual loyalty, and differentials in seniority define patterns of authority and deference which do much to maintain social control.7

Age-Set Transitions

Just as one's position in the age-set hierarchy served to order social interactions, so the sequence of age-sets served as the primary temporal reference in oral history. Therefore, the principal ethnographers of the Nandi, A. C. Hollis and G. W. B. Huntingford, and of the Kipsigis, J. G. Peristiany and I. Q. Orchardson, each sought to establish the periodicity of age-set transitions in hopes of determining an approximate dating of precolonial events.8

These ethnographers presented similar reconstructions of the sequence of rituals as practiced prior to pacification. Over a span of several years a series of initiations formed a single age-set. A period of a few years followed during which the age-set was said to be closed and no further initiations were held. Then came a mass ceremony, saget ap eito (the sacrificial slaughter of a white ox), which marked the establishment of the new age-set and confirmed the assumption by its members of the responsibilities of 'warriors' (murenik). At the ceremony members of the age-set senior to them removed their markers of murenik status and assumed the dress of young elders. Not long after this initiations for the first members of another age-set commenced.

The saget ap eito apparently served to punctuate the transition between age-sets. A solution to the question of how many years were represented by a typical age-set was therefore sought through an examination of the factors which influenced the timing of this ceremony. The various authors, however, presented incompatible and contradictory information concerning the exact function of this ceremony, the events which precipitated its occurrence, and many other features of the transition process. Thus while Huntingford claimed to have worked out the answer, neither Peristiany nor Orchardson agreed with his analysis.9

The problem was compounded by the fact that, for reasons discussed below, neither the Kipsigis nor the Nandi have celebrated a saget ap eito since the start of colonial rule just after the turn of the century. Moreover, the actual intervals of age-sets formed during this century are known with some certainty, thereby yielding an approximate answer to the immediate question. In light of this information, Manners comments that the duration of Kipsigis and Nandi age-sets 'probably has a great deal more flexibility than has previously been suggested' (1967:257). However he offers no explanation of the transition process and concludes that the matter of age-set transition 'is of no overwhelming importance so far as traditional and contemporary functioning of the institution is concerned' (1967:258).

While it is understandable that, when considering the question of age-set transitions, ethnographers working in the first decades of the colonial era focused their attentions on the past and its major ritual (which was banned at the time of their fieldwork) rather than on contemporary, less dramatic, events, it is rather surprising that subsequent scholars have apparently taken the loss of the ceremony to be the loss of the information necessary to analyze the transition process. For as everyone is aware, the age-set cycle has continued to reproduce itself in a regular and orderly fashion. An examination of the varying lengths of age-sets in the twentieth century demonstrates that it was inappropriate to approach the analysis in terms of an absolute, calendric sense of time.10 Furthermore, the continuation of the cycle suggests that the more general question of how transitions are achieved, far from being made moot by the loss of the saget ap eito ceremony, is relevant to the understanding of contemporary Kalenjin society. It is worthwhile, therefore, to reconsider the controversy contained in the primary sources in some detail, for although they focused on the precolonial situation, these accounts contain many useful clues (as well, perhaps, as a few red herrings) concerning the dynamics of the system.

Points of Controversy

The earliest source on the Nandi, A. C. Hollis, was Secretary for Native Affairs in the East African Protectorate. He managed two visits to the Nandi area, the second within two years of the final punitive expedition of the "Nandi Wars", which ended in 1906. For the most part his material was collected while on duty elsewhere in the Protectorate from five informants: two servants, two political prisoners, and a warrior wounded during pacification. Hollis lists the seven age-sets, each with three sub-divisions, and describes the sequence of ceremonies which comprise initiation. He also gives a brief, reconstructed account of the saget ap eito, 'the ceremony of handing over the country from one age to another' (1969:12). Hollis states that the ceremony was last held about 1904 and takes place about every seven and a half years.

G. W. B. Huntingford lived in close contact with the Nandi from 1921 to 1945 (1953a:ix). His major work was written under the auspices of the Colonial Social Science Research Council. Huntingford's account is particularly valuable concerning the Nandi resistance to British rule and the circumstances under which the saget ap eito ceremony was suppressed. Central to these events were the orgoiik (singular orgoiyot), men descended from a Maasai ritual expert, or 'laibon' (the anglicized form of the Maasai title ol-oiboni [Huntingford 1953a:39]), who had been adopted into Nandi with the defeat and dissolution of the Uasin Gishu Maasai around 1860. Believed to have patrilineally inherited powers to predict, curse, and bewitch, the orgoiik imposed their uncontested will on other Nandi, who both feared and resented them. The leading orgoiik among the Nandi asserted great authority, though it would be wrong to style them chiefs or leaders of a clearly centralized power structure. Several orgoiik led resistance campaigns, and in 1905 the most prominent was killed by an expeditionary force. Others were considered agents of sedition by the colonial government and were exiled from their home district as political prisoners. According to Huntingford, the Nandi were due to hold a saget ap eito around 1908 but the newly established administration forbade it. Their fears seem justified, for in 1923, with discontent running high, the leading orgoiyot plotted to turn the next saget ap eito, with its massed gathering of warriors, into an attack on the colonial post. The British were tipped off and aborted the ceremony with last minute arrests and police action.11

Huntingford presents by far the most detailed discussion of the age-set system in general and the question of age-set duration in particular. On the basis of informants' statements, political records, genealogical data, and personal observations, he concludes that the interval is, with slight variation, 15 years, or twice that reported by Hollis.12 Of particular interest in Huntingford's account is the mechanism by which this length of time is determined. The opening of a period of circumcision 'is fixed by the flowering of a bush called Setiot (Mimulopsis sp.)', a plant found in adjacent forest zones which blossoms spectacularly every seven or eight years (Huntingford 1953a:62). All proceedings were organized in terms of 24 military/territorial units called pororosiek. When the setiot flowering had been observed, representatives from each area made offerings to the leading orgoiyot and sought his sanction to open the next round of circumcision ceremonies. Approval was announced by a further ceremony held separately in each area. Initiations were then held for four years, creating four sub-sets. Initiations were then closed for several years. "Nowadays, when warfare is no longer possible, the two junior mostinwek [sub-sets] are combined....and the number....has come to be regarded as three instead of four" (Huntingford 1953a:59). Three or four years after the next flowering of setiot, which occurred during the closed period, the saget ap eito ceremony came due, and with the second subsequent flowering the initiations were opened for the next age-set.

Huntingford uses the setiot cycle to structure a great deal of information, summarized in a series of highly regular charts. He is aware, of course, that precise 15 year intervals are an approximation of actual events. Nonetheless I am uneasy, for reasons I will explain at the end of this paper, with the implication that such variation is mere noise in the system due to nothing more than the fact that the exigencies of life interfere with the exact realization of the ideal pattern. Rather, I will argue that variation and ambiguity are essential to the age-set system, and are necessary for the transition process to work as it does.

More problematical are the contradictions between Huntingford's explanation and the information in the other sources. For example, he contradicts Hollis's statement that since 1905 the Nandi have held male initiations more or less annually (Hollis 1969:12fn; original 1909), and presents evidence that open and closed periods were observed as late as 1944 (Huntingford 1953a:65ff). He also discounts Hollis's claim to have seen Nandi male initiates in 1908, which would have been a closed period by Huntingford's calculations.13 Although I can offer no further information on the Nandi, the Kipsigis data would support Hollis on both counts. This raises the question of whether Kipsigis and Nandi age-sets are coterminous. Unfortunately, none of the sources directly compare either the timing of Nandi and Kipsigis transitions or the changes each group has made in initiation practices during the 20th century. It appears, however, to have been a tacit understanding of all parties to the controversy that they were discussing the same system, and although there are minor discrepancies between the age-set dates given by the several authors, there is no overall pattern to suggest that Kipsigis and Nandi age-sets are not, at least approximately, synchronized (a possible exception is discussed in footnote 18). Moreover, some Kipsigis children have always been initiated among the Nandi, and vice versa, just as some Kipsigis and Nandi individuals have freely settled and married in the other group. In terms of traditional law the two dialect groups continue to form one jural community, and my informants (including a 'Kipsigisized' Nandi who was active in initiations) insisted that there were no significant differences between the two. But if the age-set system does not operate independently in the two groups, then Huntingford's analysis is open to doubt, for two of its main features, the use of setiot as a biological clock and the coordination of key events by a central orgoiyot, are not corroborated by the Kipsigis data.14

The Maasai-derived orgoiik spread from Nandi into Kipsigis around 1899. But by all accounts their influence among the Kipsigis was limited to individual selfish acts and did not become a dominant force in group activities. They did not serve as foci of resistance, and pacification of the Kipsigis was achieved with much less difficulty. Indeed, in the 1930s the Kipsigis cooperated with the British authorities in identifying and deporting all members of the descent lines involved (Manners 1967:265). Moreover, among the Kipsigis the pororosiek (or boriosiek) categories had become geographically dispersed, and for other reasons had ceased to serve any significant social functions by this time. The evidence is clear that the orgoiik did not have a role in coordinating Kipsigis initiations.

The standard source on the Kipsigis is Peristiany's monograph (1939) based on research conducted in 1937. Whereas the Nandi use the same four names for the sub-sets of each ibinda, Kipsigis sub-sets are given historically specific nicknames which do not recur. Thus some refinements in dating are possible, and while Peristiany accepts Huntingford's estimate of 15 year intervals as a generalization, he indicates that among the Kipsigis the period could be as much as 20 years (Peristiany 1939:31). According to Peristiany, precolonial age-sets contained three sub-sets separated by closed periods, but starting with Kimnyige (whose initiations began under unusual conditions shortly before pacification), initiations have been held yearly, and age-sets in the historic period contain from three to seven sub-sets lasting a couple or a few years each. While he lists six or seven nicknames for some age-sets, I think he is wrong to count them all separately; as he himself comments 'exceptionally, there are discrepancies between the nicknames of young men initiated during the same year at spatially distant kokwotinwek [communities]' (Peristiany 1939:45).

Peristiany's description of the events leading up to the saget ap eito takes as its example the last formal transfer of senior warrior status, from Kipkoimet (i.e. Korongoro15 to Kablelach, late in the 19th century. Compiled from statements made by a large number of men in these age-sets, late in their lives, the account is rich in emotional detail, though unfortunately clouded by a confusion over the status of Kablelach16 and an overly rigid identification of an age-set as the sons of the age-set next but one to them in seniority (discussed below, pp. 18-19). Nonetheless, the basic description rings true, and revolves around the conflicting reluctance of the senior warriors, in this case Kipkoimet, to divest themselves of this position, and the resentment of the junior warriors of the age-set recently formed, i.e. Kablelach, who were eager to hold the saget ap eito and be publicly recognized as a fully constituted age-set and the holders of senior warrior status.17 To this was added the growing impatience of older youths waiting for initiations to be reopened -- initiations that would make them the first members of the next age-set, i.e. Kimnyige. They, in turn, appealed to their fathers, the majority of whom would have been members of age-sets senior to Kipkoimet (e.g. Sawe), for help in their cause. Among these elders those individuals who had been recognized as outstanding leaders of their age-mates when warriors were prevailed upon to host beer parties for the senior warriors of Kipkoimet, entreating them through personal diplomacy and long speeches to accept the transition to eldership. To Peristiany's account one might add that a significant proportion of the fathers of potential initiates, particularly the younger fathers whose first sons were involved, were themselves members of Kipkoimet. Since initiating a son is a major achievement in one's life (and the clearest indications that one has earned the respected role of elder), some of the senior warriors would have had strong personal motives for accepting the 'retirement' of their age-set.

Prior to the actual transition from Kipkoimet to Kablelach, the former physically prevented the latter from assembling to perform the saget ap eito, until finally an armed confrontation occurred between members of the two age-sets and elders had to intervene. The use of organized force between Kipsigis was so unique that this incident was remembered as '"the war of the Ipinwek"' (Peristiany 1939:32). Presumably earlier transitions occurred when the conflict of interests reached a turning point short of crisis.

Huntingford states that such tensions between age-sets do not exist among the Nandi (1953a:68), but accounts of other East African age-set systems, such as Gulliver's material on the Arusha (1963), lend credence to Peristiany. The suggestion that age-set transitions are determined by complex social negotiations (in which members of various age-sets agree, happily or reluctantly according to their personal and group interests, that a change is overdue) is also supported by the fact that among the Kipsigis Maina age-set, opened in 1922, was closed early because of widespread anger among the elders at the disrespect of those already initiated into this group, and because the previous Maina group over a century earlier had a similar reputation as a troublesome lot (Peristiany 1939:43-44; Orchardson 1961:12).18

The final figure in the controversy was in many ways the most unusual. Ian Q. Orchardson suffered lung damage as an analytical chemist in England, and in 1910, following medical advice, settled near the Kipsigis district headquarters, eventually becoming a coffee and tea planter (Matson 1961). Apparently undeterred by the colonial "colour bar", he is fondly remembered by the Kipsigis as Chemosusu, the man who does anything and everything, which in Orchardson's case included taking two Kipsigis wives and learning the language fluently.

Orchardson denies flatly that setiot flowerings regulate Kipsigis initiations. Instead he states that the association is that 'ceremonies must not take place in the second year after the flowering' (1961:12) when it is feared that initiates might share the frailty of the new seedlings. One of my informants, born before 1880, gave substantially similar information, as does Peristiany (1939:7). Orchardson does not explain the factors that do lead to a change of age-sets except to say that it is at the discretion of the elders. He does stress, however, that the saget ap eito ceremony 'was not for the purpose of handing over the rule of the country from one set of warriors to another' as Hollis and Huntingford state, but was 'rather for determining publicly exogamy of generation' since a man cannot marry the daughter of an age-mate (1961:14). Orchardson also notes that each age-set contains, in addition to three major sub-sets, one minor sub-set in a marginal position between sets, and that 'difficulties [concerning age-set exogamy] arise at times regarding contiguous groups of two age sets, but special rules are made in these cases' (1961:13). My own fieldwork confirms both points.

Regarding the timing of transitions, Orchardson reports

the saget ap eito used to take place at any time during the latter part of a generation. It was not necessary for the age set to be completed, so that men who had been through initiation and boys who had yet to go through attended (1961:13-14).
Although he does not offer further details, this indicates that the scheduling of the regional ceremony was partially independent of various local decisions about when to hold initiations, and that age-sets thus could not be closed all at once but only over a period of time.

Whatever the situation is among the Nandi, we can summarize the Kipsigis material as follows: the duration of an age-set is variable, the end of an age-set being determined, not by any outside, naturally occurring event but by internal social pressures which periodically build up to threshold values provoking the transition. The form of the process is regular, but the timing of any given round depends on a variety of social circumstances.

The processes which lead to a transition are on at least two levels. The desire to instigate the change can arise in any number of local contexts, and at somewhat different times, as individual and group interests are negotiated through face to face interaction. Peristiany's discussion of the elder's beer party for senior warriors is comprehensible in terms of Kalenjin social organization, however, only if it is recognized, not as an account of a single key event, but as an ideal type description of the sorts of events which occurred concurrently in many places. It remains to be explained how, in the absence of any specific institution for centralizing decisions (such as the Nandi orgoiik as described by Huntingford), the transition process is synchronized across a society of tens or hundreds of thousands of people. i.e. how coordinated changes can occur on a societal level in an acephalous situation.

It would seem, from today's perspective, that the necessity of this second level of analysis is obvious. The regular, orderly continuation of age-set transitions unmarked by public rituals indicates that an understanding of the transition process should be sought in terms of the basic organization of Kalenjin societies, rather than in the specific context of, or beliefs associated with, these rituals. But to the earlier ethnographers the saget ap eito and the historical circumstances surrounding its demise were more recognizable subjects for research than all the small scale discussions associated with the abbreviated initiations of the colonial era, and the broader issue of social dynamics went unattended. The structural-functional concepts with which they were working, and the closely related nature of the data available to them, primarily institutional descriptions and explanations of normative rules collected from key informants, were ill-suited to the processual analysis of the age-set system.

For example, Peristiany's point that the local decision-making process regarding transitions can be understood in terms of the relative positions of fathers and sons in the age-set sequence is well taken. Problems arise, however, from his treatment of the number of age-set intervals between fathers and sons. Several sources report that a boy should not be initiated into the age-set immediately following his father's (Steward 1977:103), and both Peristiany and Huntingford refer to the members of the second subsequent age-set as the sons of the first. In a manner of speaking (used by the Kalenjin), they are. But while Huntingford cites data demonstrating that this is the most frequent, but not the only situation, Peristiany only mentions the possibility of such variation in passing (1939:31n.), and bases his analytical argument on the folk description of a standard two set interval rather than on the empirical distribution of age-sets. As a result his discussion of the interpersonal dynamics involved in these negotiations is both incomplete and inconsistent. For 124 pairs of fathers and sons in my genealogical tables, the number of age-set intervals are distributed as follows: one interval (i.e. adjacent sets) 13 cases, two intervals 54 cases, three intervals 40 cases, four intervals 16 cases, and five intervals 1 case. The 'typical' two interval span is thus found in less than half the cases. The explanation of the longer intervals between fathers and sons is obvious in a polygynous system. Concerning fathers and sons in adjacent age-sets, Peristiany comments in a latter publication that 'the Kipsigis now breaking down and the elders are at a loss to rationalize the numerous infractions of this rule' (1951:292). The 13 'infractions' in my data, however, are distributed fairly evenly throughout this century. While I do not know the details of these cases, there are a number of ways in which traditional family patterns, quite apart from recent changes, could have produced these outcomes. Such irregularities are not moral offenses and, in the absence of centralized authority requiring their rationalization, neighbors and relatives condone such events according to their individual circumstances. As I shall argue more fully below, to understand the age-set system one must consider not only how decisions are negotiated, but also how the rules which constrain those decisions are themselves subject to local interpretation and redefinition.

Normative descriptions are a vital part of ethnography, and structural models at this level of generality are an important stage in analysis. But they are not sufficient to generate accurate predictions of the patterns of social organization that exist 'on the ground'; these must be discovered inductively. Normative descriptions divorced from a flow of events are inadequate for an understanding of the dynamics of a system no matter now much functionalist logic is applied. While the point is general, and lies behind such diverse approaches as the extended case method and computer simulation, I believe it is particularly applicable to the analysis of social systems in which information is uncentralized, and coordination of large numbers of individuals is typically temporary, periodic, or episodic.

Kipsigis Social Organization

The basic corporate unit among the Kipsigis is comprised of the people dependent upon the cattle of an individual elder. Over a third of all married men are polygynists; of those over 40 about half are.19 The majority of married women thus have co-wives. De facto division of the family herd starts with the separate capitalization of each wife's 'house', increases further if co-wives are given separate homesteads, and becomes sharper still as grown sons marry and establish their own homesteads. After a man's death his sons may have to share rights in his property jointly until the younger of them reaches maturity, but when it becomes practical the property is divided among the houses and then between the full brothers within each house. The terms of the division cannot be contested by the next generation. Similarly, disputes between brothers-in-law over bridewealth cannot be continued by their sons. While one may expect support from one's cousins, even agnatic first cousins are no more obligated to one than mere clansmen, whose assistance depends upon maintaining good relations. Thus there are no lineages and even cattle sharing family groups are likely to be dispersed over a wide area.

Marriages, on the other hand, are highly localized. In some areas at least, as many as half of the women remain in their natal communities after marriage, in many cases marrying men coming from their natal homes elsewhere. Of 41 marriages involving people in one community over a 16 year span, 14 were between two member homesteads and, significantly, the remaining 27 were all to partners living within an hour's walking distance. While agnates are soon sloughed off into the general category of clansmen, the kinship system allows the recognition of affinal connections as broadly as is necessary to establish readily a frame for interaction: one's brother's father-in-law is one's father-in-law, and ultimately one's clansman's father-in-law can be addressed as one's father-in-law. Polygyny, the dispersal of agnates, a high rate of local endogamy, and the 'horizontal' extension of relationships through half-siblings, mothers' co-wives, etc., results in a network of relationships in which each homestead has many agnatic and affinal ties to others both near and far, and virtually all homesteads in each neighborhood are involved in local clusters of extremely dense interrelationship.

I find it useful to visualize the system by imagining each homestead as a neuron, connected to several of its neighbors and more idiosyncratically to many other neurons scattered more widely. Connections exist in a number of modes (affinity, agnation, initiation bonds, private cattle partnerships, etc.). The day to day processing of material and information, i.e. social behavior, will cause a constant 'twinkling' as neurons 'fire', with discernable areas of locally increased activity. Various occurrences will activate different patterns. Agricultural chores light up small knots of interrelated neighbors. Weddings activate two agnatic nets, and some of their immediate affines, spread quite widely. Annual circumcision celebrations appear as closely synchronized firings of groups of 20 or so neuron/homesteads all across the field. These clusters touch most, but not all, neurons and are centered slightly differently each year; their manifestations also last longer than most other activities. Over time repeated incidents will appear to indicate the presence of roughly localized clusters of 20 to 40 neurons that act together. Their joint firings are never exactly in the same pattern, at different times they trail off in different directions, and their edges interpenetrate. As entities they would have to be defined statistically over time. Such is the Kipsigis community, the kokwet.

Today the network comprises perhaps 50,000 homesteads. Any local portion is processing, and has encoded in its structure, only a small part of the total information of the system. Naturally, individual informants are unable to give a clear description of society-wide processes; everyone is working from limited knowledge. What is necessary for the system to function is general agreement on the major principles which define interaction, flexibility in their application to actual details, and mechanisms to allow the regeneration of the network through the establishment of new connections between 'neurons' previously involved in different sub-sets of the larger whole.

Transitions Reconsidered
In early 1967 I chanced to hear a newly initiated youth refer to himself as a member of Korongoro rather than Sawe age-set. Although I was deeply immersed in the affairs of the community, the remark came as a surprise. He explained that the initiates of December 1966 were told that publicly they should pretend to be tail-enders in the Sawe age-set, but privately they were to understand that they would soon be recognized as the first members of Korongoro. Their instructors had explained that an aged widow of the last Korongoro group was still living and that it would be unseemly to announce Korongoro reopened. Men have a certain freedom of access sexually to widows of their age-mates, and while this was obviously not a real consideration in this case, apparently any overlap of two manifestations of the same age-set, even to this slight extent, raises conceptual paradoxes. An elder commented at this point in the conversation that the same situation occurred with his initiation and when news of the transition leaked out, it caused the death of the old woman in question.

Whether there was a surviving widow of Korongoro in 1967 or whether the story was a mere pretext I cannot say, but the fact that local decisions are kept tentative until more information is received is highly significant. Further inquiries revealed that not only are such local decisions confirmed retroactively, but that there have been cases where a local community have announced a transition only to find that in the opinion of surrounding parts of the network they have jumped the gun.

To return to the neuron analogy, one might say that when a threshold is reached in one local area the information disperses into the directly connected points in the rest of the network. A few scattered events of this sort will have no overall effect, but when a significant number of such local events occur within the same period of time, they not only reinforce each other but sum to 'fire' the remaining local areas which were nearing the threshold. Pushing the analogy a bit further, one might say that the total network is convulsed into a coordinated transition of age-sets.

Further Remarks
Although this model is adequate, I believe, to explain how a coordinated transition can occur without a centralized decision-making structure, and could be tested with further research, it seems highly unlikely that one or a few fieldworkers could collect enough information about any one transition to determine when and where the critical turning point was reached. More generally, I would like to suggest that the slightly indeterminate quality of many features of Kipsigis organization which seems to haunt all of the monographs should be appreciated not as an ethnographic shortcoming but as an essential feature of the culture.

It follows from the convulsion model that transitions are coordinated but not absolutely simultaneous events in different parts of the society. The minor, marginal sub-sets provide the necessary leeway. Men initiated around the time of the transition have the later option of declaring which age-set they identify with. In the one instance of this sort I know of, a man in the marginal Chumo/Sawe sub-set defined himself as Sawe in order to marry the daughter of a senior member of Chumo. The special rulings made in such cases, mentioned by Orchardson, are of course local decisions about individuals. There is no way to standardize these decisions; indeed it is important that they are free to vary. A similar flexibility is indicated by Orchardson's observation that the saget ap eito ceremony did not have to come exactly at the change of age-sets in any given community. Such things are sorted out as local events are compared.

Similarly Peristiany's comment that sub-set nicknames occasionally differ in widely distant areas is quite understandable. It is perhaps surprising that as Evans-Pritchard's student publishing shortly before The Nuer was issued, he did not pursue the implications of this observation in terms of structural relativity. But even Evans-Pritchard, in his introduction to Peristiany's book (1939) and in his comparative analysis of Kalenjin political structure (1940b), forces the descriptive data into a series of formal schema (for example a hierarchy of seven levels of regional organization) that simply do not exist. The analysis of local leadership, its functioning and its constituencies, is similarly obscured by the incomplete realization of the fact that in the social network centers of patterns are labeled, but these terms, whether spatial or genealogical, are frequently neither mutually exclusive nor unambiguous in their level of specificity.

Just as the separation of age-sets is critical yet fuzzy, so one can argue are the limits of exogamous descent groups. Knowledge and opinion concerning which ortinwek cannot intermarry vary from Kipsigis to Kipsigis; the number of descent groups is finite but intrinsically indeterminate. Indeed, the limit of that small group which shares rights in the family herd, on which all kin relationships are founded, is both formally declared and the subject of years of deeply interpersonal adjustment in all families. I suspect that any hard and fast building block of Kipsigis social structure when viewed closely will show a similar, non-categorical, processual nature.

The imposition of centralized authority over acephalous societies has often been lamented as ending the chance to study the dynamics of these societies in operation. From the consideration of Kipsigis age-sets, however, it would appear that in many cases uncentralized processes, lacking points of vulnerability and possessing informational resilience, can be highly adaptive, and that their continued study promises insights of general analytical relevance.

  1. 1 For a detailed discussion of the origin of the term Kalenjin see Kipkorir (1973:72-76).
  2. 2 The seven set cycle of the Kipsigis and Nandi is shared by the Okiek ('Dorobo') of North Tinderet Forest (Huntingford 1951:30), the Terik (Sangree 1966:69,75), the Sebei on the Ugandan side of Mount Elgon (Goldschmidt 1976:104) and, presumably, by the closely related Sabiny or Sabaot on the Kenyan side. For the Kalenjin groups east of the Uasin Gishu plateau, however, eight sets are reported, with Korongoro and Kipkoimet as separate sets rather than alternate terms for the same set (see footnote 15). The sources are, for the Keyo: Massam (1927:54) and Welbourn (1968:214), the latter based on information written by D. K. Kiprono; for the Marakwet: Kipkorir (1973:9); and for the Pokot, or 'Suk', Beech (1911:5) and Peristiany (1951:291). The only published reference to the Tugen age-set system is a brief remark by Huntingford (1953b:77) that it consists of seven age-sets including both Korongoro and Kipkoimet but lacking Nyongi.
  3. 3 In 1966 I located one survivor of Kebebuja, the minor sub-set ending Kablelach, who were initiated around 1891. Assuming the start of the recent Korongoro age-set, discussed in this paper, to have been generally accepted before his death in 1969, the adult male population briefly spanned seven age-sets, though it should be noted that the last survivor of Kimnyige in the area in which I worked died in 1966. While there may have been a few other men of extreme age who escaped my attention, in terms of the wider society these two age-sets were effectively empty.
  4. 4 Research was conducted among the Kipsigis from August 1965 to March 1968, and during June 1972. Research was supported by a grant from the National Science Foundation, and by the Cross-Cultural Study of Ethnocentrism Project and the Child Development Research Unit of the University of Nairobi, both funded by the Carnegie Corporation of New York.
  5. 5 Kenya National Archives, Lumbwa District Annual Report for 1906-07.
  6. 6 The census figures are as follows: 1848, 157,211 (Pilgrim 1961:33); 1962, 341,711 (Kenya Government 1968:15); 1969, 471,459 (Kenya Government 1971:15). These numbers include an increasing proportion of Kipsigis living outside Kericho District.
  7. 7 The census figures are as follows: 1848, 157,211 (Pilgrim 1961:33); 1962, 341,711 (Kenya Government 1968:15); 1969, 471,459 (Kenya Government 1971:15). These numbers include an increasing proportion of Kipsigis living outside Kericho District.
  8. 8 This goal is most clearly stated by Huntingford (1953a:54) but is implicit in the charts constructed by Huntingford (e.g. 1927:427), Hollis (1969:12), Orchardson (1961:125) and in many passages discussing precolonial events. Minor sources on Kipsigis age-sets include Dobbs (1921) and Barton (1923) and, for the Nandi, Hemsted (1923).
  9. 9 The publication dates of the main sources do not fully reflect the chronology of the debate. Huntingford first published his estimate of 15 year cycles in 1927 and maintained this position in subsequent writings on the Nandi-speaking peoples (e.g. Huntingford 1935, 1942, 1951). Although he first published the detailed arguments for his position, and his assessment of the Kipsigis situation, in 1953 (1953a and 1953b respectively), it is clear that he reached this conclusion much earlier. Orchardson first published on the Kipsigis in 1919, but aside from a brief mention of 'generations' (1931:100), did not discuss the age-set system in print in his several papers. His book length manuscript was first completed in 1929 and finally revised in 1937. It was 'abridged, edited, and partially re-written' by A. T. Matson and published in 1961. While Orchardson's book lacks any citations, his mention of the debate over age-set intervals and a discussion of setiot flowerings (1961:11,12) clearly refers to Huntingford's work. This passage appears to have been written in 1929 (the expected flowering of 1934 is not mentioned and the information on age-sets stops at 1929). I do not believe Huntingford discussed setiot in print prior to this date. It thus seems highly likely that Orchardson and Huntingford were in personal communication for many years. Peristiany conducted fieldwork in Kericho District from March to November, 1937 (Kenya National Archives, Kericho District Annual Report for 1937, p. 45). Peristiany read Orchardson's manuscript and used Orchardson's dictionary (still unpublished) while in the field (Peristiany 1939:xvi,231). Concerning his monograph, Peristiany comments: 'as this is a plain account of my field-work I do not want to introduce into it critical arguments and controversial subjects and to discuss Mr. Hollis's, Orchardson's and Huntingford's facts proving that one ibinda lasted anything from 7 1/2 to 21 years, but I think Mr. Huntingford is nearer the truth when he states that the ibinda lasted about 15 years. I hope to be able to discuss this question later in a separate paper' (1939:31).
  10. 10 The initiation dates of Kipsigis age-sets in the twentieth century are (inclusively): Nyongi 1902-1921; Maina 1922-1929; Chumo 1930-1945; Sawe 1946-1965. The span of years for each is thus 20, 8, 16, and 20.
  11. 11 According to Huntingford (1953a:66) the next saget ap eito among the Nandi was due in 1938 but was postponed for a year due to uncertainty about the government's attitude, and then cancelled when World War II broke out.
  12. 12 Huntingford edited the second edition of Hollis's book. In doing so, Huntingford changed the text to read 15, rather than seven and a half years, per age-set and altered Hollis's dating of precolonial events accordingly (Huntingford 1969:xi). These 'corrections' produce some curious results: Hollis, writing in 1908, appears to have described Chumo age-set as 'men between 50 and 60, circumcised about 1836' (Hollis 1969:12), i.e. 12 to 22 years before their birth! The critical passage, with the dates Hollis gave in the first edition, is quoted in Huntingford (1953a:55-56).
  13. 13 Huntingford comments "it is possible he [Hollis] may have seen some boys who for some reason were circumcised during a closed period; or (what I think is more probable) he may have seen nothing more than boys dressed up for the occasion to show him what the circumcision dresses looked like....whatever he saw, he cannot have seen the Nyongi circumcision, because it was not opened til 1912" (Huntingford 1953a:57). Aside from the fact that Nyongi opened in 1902 among the Kipsigis, I find it difficult to believe that Nandi elders would sanction a mock enactment of part of their most serious ritual, especially for the benefit of the government which had 'pacified' them with such brutality a few years earlier, or to assume that an observer with such a keen interest in ceremony and material culture as Hollis could be thus mistaken.
  14. 14 The term orgoiik predates the absorption of Maasai laibons by the Nandi (and subsequently the Kipsigis), and applied to members of another clan. They were not nearly as influential as the laibons, however, and Huntingford does not mention them as having a role in initiations. Orgoiik of Maasai origin are not found in the Kalenjin groups other than Nandi and Kipsigis. Thus Huntingford's account of the role of orgoiik applies to only one example of the cyclical age-set system as it operated only since some time after 1860.
  15. 15 For discussion of the alternate use of these two terms see Peristiany (1939:42), Huntingford (1953a:74), and Orchardson (1961:125). In this paragraph I use Kipkoimet to maintain consistency with Peristiany's account. However, as described elsewhere, the term given to the Kipsigis initiates in my research area in 1966 was Korongoro.
  16. 16 The saget-ap-eito marked the retirement from warrior status of one age-set, the debut, as a group, of the next, whose initiations were being completed, and preceded the opening of initiations for the third. Peristiany's account fails to distinguish the second two groups and he wrongly identifies Kablelach in this case as awaiting initiation. Possibly the confusion arose because the emergence of the new senior warriors and the emergence of initiates from seclusion are ritually similar, both involving passage through, or around, a symbolic gate (ormarechet).
  17. 17 More recently, in the absence of closed periods between age-sets, the senior members of the group being initiated become increasingly unwilling with each year to see young initiates declared their age-mates.
  18. 18 It is said that Silobai, the nickname of one of the sub-sets of the current Maina age-set, is a veiled reference to an act of sodomy by two members of this group while in seclusion. Concerning the curtailment of Maina, Orchardson (1961:12) states that both Kipsigis and Nandi did so after the 1929 initiations. But while Peristiany reports that the initiations into Chumo started among the Kipsigis in 1930, Huntingford states that among the Nandi a closed period followed the last Maina initiations of 1930 until the opening of Chumo in 1940, maintaining the 15 year periodicity (1953a:Table II facing p. 74).
  19. 19 The figures in these two paragraphs are drawn from my data on one local area (Daniels 1970), and should be understood as suggestive rather than descriptive of exact rates of polygyny and local endogamy in other areas and at other times.

Beech, M.
1911 The Suk: Their Language and Folklore. Oxford: Oxford University Press.
Barton, J.
1923 Notes on the Kipsikis or Lumbwa Tribe of Kenya Colony, Journal of the Royal Anthropological Institute 53:42-78.
Daniels, R. E.
1970 By Rites a Man: A Study of the Societal and Individual Foundations of Tribal Identity among the Kipsigis of Kenya. Unpublished Ph.D. dissertation, University of Chicago.
1980 Pastoral values among vulnerable peasants: can the Kipsigis of Kenya keep the home fires burning? , in S. Abbott and J. van Willigen (eds.) Predicting Sociocultural Change. Southern Anthropological Society Proceedings 13. Athens, Georgia: University of Georgia Press. 57-75.
Dobbs, C. M.
1921 Lumbwa circumcision ages, Journal of the East Africa and Uganda Natural History Society 16:55-77.
Evans-Pritchard, E. E.
1939 Introduction , in J. G. Peristiany The Social Institutions of the Kipsigis. London: George Routledge & Sons, Ltd. xix-xxxiv.
1940a The Nuer. Oxford: Oxford University Press.
1940b The political structure of the Nandi-speaking peoples , Africa 13:250-67.
Goldschmidt, W.
1976 Culture and Behavior of the Sebei: A Study in Continuity and Adaptation. Berkeley: University of California Press.
Gulliver, P. H.
1963 Social Control on an African Society. London: Routledge & Kegan Paul.
Hemsted, C.
1923 The tribal organisation of the Nandi , Journal of the East Africa and Uganda Natural History Society 18:3-15.
Hollis, A. C.
1969 The Nandi: Their Language and Folk-lore. Oxford: Oxford University Press. (first published 1909)
Huntingford, G. W. B.
1927 Miscellaneous records relating to the Nandi and Kony , Journal of the Royal Anthropological Institute 57:41761.
1935 The Nandi pororiet , Journal of the Royal Anthropological Institute 65:133-44.
1942 The social organisation of the Dorobo , African Studies 1:183-97.
1951 The social institutions of the Dorobo Anthropos 46:1-48.
1953a The Nandi of Kenya: Tribal Control in a Pastoral Society. London: Routledge & Kegan Paul.
1953b The Southern Nilo-Hamites. Ethnographic Survey of Africa, East Central Africa, Part VIII. London: International African Institute.
1969 Introduction [to the second edition], in A. C. Hollis The Nandi: Their Language and Folk-lore. Oxford: Oxford University Press. xi-xxiii.
Kenya Government
1968 Statistical Abstract. Nairobi: Ministry of Planning and Development.
1971 Statistical Abstract. Nairobi: Ministry of Planning and Development.
Kipkorir, B. E.
1973 The Marakwet of Kenya: A Preliminary Report. Nairobi: East African Literature Bureau.
Manners, R. A.
1967 The Kipsigis of Kenya: culture change in a "model" East African tribe , in J. H. Steward (ed.) Contemporary Change in Traditional Societies, Volume I, Introduction and African Tribes. Urbana: University of Illinois Press. 205 359.
Massam, J. A.
1927 The Cliff Dwellers of Kenya. London: Seeley, Service.
Matson, A. T.
1961 Introduction", in I. Q. Orchardson The Kipsigis. Nairobi: The Eagle Press.
Orchardson, I. Q.
1919 Supernatural beliefs of the Lumbwa , Journal of the East Africa and Uganda Natural History Society 14:417-20.
1931 Notes on the marriage customs of the Kipsigis, Journal of the East Africa and Uganda Natural History Society 40/41:99-112.
1961 The Kipsigis. Nairobi: The Eagle Press.
Peristiany, J. G.
1939 The Social Institutions of the Kipsigis. London: George Routledge & Sons, Ltd.
1951 The age-set system of the pastoral Pokot . Africa 21:188-206, 279-302.
Pilgrim, J. W.
1961 The social and economic consequences of land enclosure in the Kipsigis reserve , Unpublished manuscript. Kampala: East African Institute for Social Research.
Sangree, W. H.
1966 Age, Prayer and Politics in Tiriki, Kenya. London: Oxford University Press.
Stewart, F. H.
1977 Fundamentals of Age-Group Systems. New York: Academic Press.
van Gennep, A.
1960 The Rites of Passage. Chicago: University of Chicago Press. (First published in French 1908)
Welbourn, F. B.
1968 Keyo initiation. Journal of Religion in Africa 1:212-32.