In my judgment, the results of the statistical analysis of the interview data, although mixed, offer sufficient support of the basic hypotheses to warrant their retention. In this final chapter I would like to go beyond the results, narrowly defined, and consider some of the reasoning underlying this study and some of the possible implications of the interpretation presented here and, finally, to comment briefly on the methodological position I have taken.

The easiest way to delineate the theoretical framework used here is to start with points of view that I find unacceptable. First, to my mind, the fact that cultural systems are bounded in very complex and incomplete ways is conceptually inseparable from the very nature of culture. Viewed from an evolutionary perspective, the development of culture can be seen as a shift in the determinants of behavior from physiological materials to learned, symbolic instructions. Cultural guides to behavior, being non-material, have been accumulated and altered in ways that have led to mankind's extraordinary biological success (range of radiation, population growth, amount of energy controlled, etc.) and the concomitant diversity of local patterns of behavior. With the separation of behavioral diversity from physiological diversity, mankind has avoided the fragmentation of speciation while developing differences between parts of the species (e.g., societies) that are exposed to the differential effects of selective pressures (including intraspecies competition). So far, the development of these differences has been of adaptive value to the species as a whole. It must be stressed that these selective forces operate on the level of cultural units (traits) rather than on the level of biological units (individuals). Indeed, the boundaries between different human societies, or pseudo-species, can be and frequently are crossed by genes, by individual organisms, and by cultural ideas and artifacts. Moreover, the types of boundaries maintained between different societies vary from place to place, and are open to constant revision.1 Although the maintenance of group identities and group boundaries is important to every society, it is not assumed that the present interpretation of Kipsigis stereotypes can serve as a useful model for understanding all types of ethnocentrism.

Secondly, I find it impossible to treat cultural processes as phenomena that can be understood without a consideration of individual psychological processes. Cultural patterns must be learned within the context of the life cycle. This learning process, whatever the specific content involved, is limited by the dynamics of psychological development. Diachronically, learning is limited by the cumulative effects at each stage of psychological growth of the patterns of earlier stages. Personality, in these terms, can be defined as the basic persistent patterns in an individual's mental organization seen over time and across growth stages. Synchronically, the cultural patterns involved in this organization must have some minimal level of coherence if the individual is to interact with other members of the society in predictable and appropriate ways. Thus certain aspects of a cultural system are integrated not only on an ecological or sociological level but, quite literally, integrated in people's heads, and an explanation of the relationship between childhood experiences and adult behavior in the individual requires the use of personality variables as they develop during the individual's life. This is the familiar "personality as a mediating factor" approach (Whiting 1961).

In this study I have tried to demonstrate a relationship between personality structure and the structure of group identity. Specifically, the central interpretation is that the typical Kipsigis male experiences include:

1. being raised in a setting that involves many limitations on the extent and quality of father-son interaction,
2. participating in initiation rites and daily patterns of social interaction that stress sex-role differences and demand overt compliance to a rather extreme model of manhood, and
3. having thus been sensitized on a personal level to questions of sex identity, exhibiting a tendency to view others in these terms.

This interpretation implies that the importance given to relative masculinity in Kipsigis culture (and other similar cultures in Kenya) would appear rather extreme if compared to a representative sample of the world's societies. At present this can only be supported with impressionistic data; nonetheless I doubt if many observers familiar with the Kipsigis would disagree.

The analysis is based on the fact that there is individual variation among Kipsigis men. The implication here is that for some men to content of Kipsigis tribal identity and the associated stereotypes of other societies are not simply cognitive "givens" of their culture, but are expressions of a theme that is intimately related to a basic dynamic tension in their own personality structures. Initiations are not simply passive symbolic definitions of manhood but are, for those boys who enter initiations with subconscious conflicts about their own identity, active training in how to convert this tension into motivations that are adaptive in adult roles. Similarly, stereotypes are not just traditional cognitive patterns to be internalized, but are also projections of the belief that the experience of initiation establishes manhood. They are, in part, attempts to structure what lies over the horizon in a way that confirms what one wants to feel is within the self. In the present analysis childhood experiences and initiation experiences are seen as "input" into the individual. The measures of sex identity are attempts to judge the resulting personality characteristics. An individual's ethnocentric opinions are seen, in part, as "output" motivated and structured by one such characteristic (cross-sex anxiety).

The model is, of course, a highly simplified and incomplete representation of reality. This simplification should be seen as an analytical tactic, not a theoretical definition of all that is happening or all that is significant. The rationale for focusing on this particular directional aspect of the total system is the developmental nature of personality discussed above. The justification lies in the results gained from analytical tools which offer incisiveness at the cost of complexity.

In reality every cultural act invokes interaction between individuals directly or indirectly. Cultural behavior is transmitted across generations, between people in different stages of their lives. One person's "input" is another's "output" (again a simplification). Among the Kipsigis, most father-absent boys become fathers who maintain certain distances from their sons, some of the initiates become initiators, and some of the youths who assumed stereotypes congruent to their inner tensions become men who deal directly with people from other tribes and are faced with the problems of ordering and communicating these experiences.

The model used here would predict that the men for whom the issue of relative masculinity has deepest personal significance are the ones who would engage most deeply in the behavioral patterns that traditionally characterized manhood. Further, their behavior as adults would produce similar tensions about sex identity in succeeding generations. It must be stressed that this interpretation is concerned with the psychological factors that contribute to the the maintenance of these initiations and stereotypes. The interpretation has nothing to say about the origins of these practices.2

A few tests of this "feedback" idea have been presented in the preceding chapters, and it should be noted that the results (comparing child care scores to father-absence and comparing behavioral measures of adult participation in, and attitudes toward, initiation with measures of one's own initiation) are either negative or extremely restricted. In most cases the control variables, especially subject's age, are the only significant predictors of these kinds of adult behavior that can be found in the present data.

When we consider the basic interpretation in relation to the dramatic changes that have occurred in the twentieth century, and note that these changes have been much greater in the realm of intertribal relations than in intrafamily interaction patterns, we are led to the prediction -- which strikes me as rather unwelcome -- that to the extent that cross-sex anxiety supports traditional stereotypes it will hinder adaptive changes in group identity in the new, national setting. The lack of correlations between the measures of familiarity with other tribes and the measures of ethnocentrism do suggest that this is partially the case, but I feel that these data are much too crude to allow any conclusions on this important and complex question. A similar situation is found when one inspects the data for evidence to support each of the implications drawn here. Because of the limited degree of attention that was given to these intricate, secondary questions, the data are either too weak or are absent and no clear positive or negative results can be obtained. However, in my opinion it would be a mistake to consider this lack of conclusiveness to be a refutation of the original interpretation. To me, the positive support of the basic hypotheses, weak as it may be suggests that it would be profitable to consider these implications as propositions for further research.

Traditionally, most interpretations in social anthropology have been based on data collected by the participant-observer method in which the anthropologist uses himself as the control research instrument. While this method has produced excellent descriptive results, it has also posed serious problems for building theories based on a comparative consideration of such descriptions. One major problem is that of determining the empirical basis for interpretations made using this method. Because of the intensely personal, complex, and yet diffuse nature of the researcher's role, it is extremely difficult for him to record the process by which his interpretations were formulated. For the same reason, it is difficult for others to evaluate his findings with a reliable degree of objectivity.

In order to gain some control over the subjective factors inherent in the participant-observer method, a variety of data gathering techniques have been developed and quantified data, collected through some sort of systematic survey, have become a standard part of good ethnographies. Nevertheless, anthropological controversies still tend to degenerate into situations in which each party feels his own reading of the data to be correct, rather than leading to cumulative advances in understanding.

In my opinion, this situation will not improve unless anthropologists, recognizing the tentative and impressionistic nature of conclusions produced by the participant-observer method, conduct further research in the field using procedures approximating the scientific method, i.e., the development of testable hypotheses from these "conclusions," the operationalizing of concepts as variables and measures of these variables, and the systematic gathering of data from a number of observations sufficiently large to allow the use of statistical models in the subsequent analysis.

The central feature of the model of the scientific method is, of course, the emphasis given to "objectivity," the separation of observations from analysis, which contrasts strongly to the tautological thinking found in much of traditional anthropology. The scientific approach does not automatically insure a greater measure of objectivity, but it does offer specific ways to measure the degree of objectivity involved.

The application of this method in the present study has not been without considerable error. The logic of the statistical tests used in the analysis, for example, is based on the model of a random sample, while the actual sample was obtained quite differently. There are, however, ways to estimate the quality of the actual sample and these have been discussed at length in Chapter 6.

Perhaps more important is the general question of error at every level of the analysis. Random error of any kind weakens the results of statistical tests; biased (patterned) error falsely enhances the results. There is certainly reason to assume that the analysis presented here contains a large measure of error. Care was taken to avoid the most obvious types of bias, but almost all of the work, from the formulation of hypotheses to the recording of responses to the interpretation of results, was done by one person, and the possibilities of subjective bias are thus innumerable. Though this problem is unavoidable in anthropological fieldwork of any kind, it is possible to describe the procedures of data collection and analysis used in Part II (but not those used in Part I) in a way that would permit nearly full replication. More realistically, the method used in Part II has produced comparability between responses on certain key measures (the test measures of relative masculinity and the ethnocentrism measures) with those obtained by other researchers.

The formulation of hypotheses and the quantification of data, for all their procedural advantages, do not solve any of the problems inherent in the process of drawing inferences. Quite the contrary, the process of operationalizing intuitive interpretations given one an appreciation for the very limited nature of the concept currently in use and, while forcing one to consider which theory is most relevant to the actual research questions, also reveals how little is understood about the postulated relationships between the specific variables. Nor does the use of statistical tests of significance solve the problem of deciding which results are most meaningful. The statistical significance of a correlation is a measure of the probability that one will be wrong in rejecting the null hypothesis that whatever relationship is found between the variables is due to chance. It is obvious that a finding may be highly significant in this sense, but logically trivial. And a finding that does not achieve a recognized level of significance statistically may be highly significant analytically, i.e., in its contribution to building further (and hopefully more demonstrable) hypotheses. We should also note that if we accept a probability level of five per cent as "significant" then, with totally random data on a large number of variables, one out of twenty correlations can be expected to turn out statistically "significant." Thus in a study such as this, where a great number of intercorrelations have been tried, it is important that we distinguish between those which involve the basic hypotheses and those of a distinctly secondary nature. This can become a difficult judgment because it is rarely possible to achieve a one to one relationship between the conception of variables and their operationalization as measures.

Finally, we must recognize the larger implications of these problems, which might be summarized by saying that no single set of narrowly circumscribed observations can prove or disapprove a non-trivial theory; theories are statements about the interrelationships of patterns perceived in large numbers of observations, and several layers of inference are involved in judging the relevance of particular data to a general theory.

Thus the model of quantified methods presented here does not produce any royal road to anthropological answers. But I hope I have demonstrated, in at least one case, that such quantification is a necessary addition to descriptive analysis if anthropological interpretation is to be more than intuitive speculation in which a concept becomes a conclusion and a feeling is a finding.

To the Table of Centents

1 This flexibility of culturally defined units remains the stumbling block of attempts to apply models of biological evolution to human behavior, while the diversity in types of local group boundaries raises one of the major problems of cross-cultural studies, that of the comparability of units.

2 I was unable to elicit any information about (or even interest in) the origins of initiations. One informant did relate a legend describing how a woman and her children became separated from the rest of the people and thus their descendants (the Luo, of course) no longer know how to perform initiations. Another man, on hearing that clitoridectomy is not performed in America, commented "you've forgotten how."