Measuring Cultural Sharing

There are two serious problems involved in the assessment of the extent of cultural sharing. They are doubtlessly part of the reason for the fact that only limited attempts to make this assessment have been carried out since Roberts's (1951) pioneering effort.

First, what it is, exactly, that actors do or do not share needs to be specified. If this difficulty is overcome, there remains the problem of determining whether this sharing is present. Put otherwise, how can we isolate the units that make up culture, and how can we determine whether these units are shared or not? Cognitive anthropologists have made progress in dealing with these two related problems (e.g., Berlin and Kay 1969). I follow one of their leads by keeping the immediate scope of investigation narrow and mainly limited to data that are readily quantified.

In the study focused on this issue (Swartz 1982a ), I took, of necessity, a provisional and partial approach to the two problems just noted. In this approach, the units of culture are identified with responses to items on a questionnaire (see Appendix and ibid., 335–338). Without doubt there are serious problems in taking responses to questionnaire items as equivalent to cultural elements. In my 1982 study, I used the same multiple-choice questionnaire (appropriately translated) for the Swahili and the other four European and American communities where the interviewing was done.

To take an obvious problem, it is quite possible that what I intended as different items are, in fact, reflections of a single cultural element and that what was a separate element of culture in one of the groups where we worked was not in another. The questionnaire was devoted entirely to questions concerned with nuclear family life. It asked informants to choose among alternatives concerned with issues—61 in all—such as who the family peacemaker was, how children should treat their aged parents, and where most children live after they marry. The status of these questions as cultural elements has

to remain in doubt, and the only basis for treating them as elements is pragmatic: if the problem of cultural sharing and cultural dynamics is to be studied directly, the effort must begin even if all the problems in making the study have not been solved.

A similar rationale was used to justify the approach to cultural sharing that was used. This approach is to take cultural sharing as present when two informants chose the same response to a question. A sharing coefficient was calculated based on the number of questionnaire items on which two informants made the same response as compared to the total number of responses from the two minus the number of identical responses according to the following formula:

[Full Size]

So, for example, if two informants answering sixty-one questions each chose the same responses on all of them (61/122–61), they would have a sharing coefficient of 1.00.

It is true, of course, that people can have some—even a good deal—of cultural sharing concerning the subject of a question without choosing the same response and that they can choose the same response without much cultural sharing. Nevertheless, the sharing coefficient as defined for the study gives some approximation of what two actors share on the issues they are asked about, and there is no reason to believe that the responses are systematically biased toward either error.

For the Swahili (and the other four groups studied), we interviewed only people who belonged to families that met the following conditions: (1) they contained a wife and a husband who lived together at the time of the interview; (2) the spouses considered themselves and were considered by their child or children as the natural parents of the child or children included in the study; (3) the spouses had at least one child twelve years of age or older who lived with them and who had never been married. We inteviewed three members of each family meeting the above conditions in each society. We chose this number because it was the largest number of interviews we found it practical to get from a particular family while at the same time being large enough to allow us some measure of intrafamilial sharing across generational and sex, as well as status, divisions. The five-society study aimed at dividing each community's families equally between families represented by two children (siblings) and a parent and families represented by two parents (spouses) and a child (see table 3).^[1]

The coefficient of sharing for the total sample was obtained by comparing the responses of each member of the sample from that society to those of every other and taking the mean of all those comparisons. This latter coefficient is not really representative of the whole Mombasa Swahili community,

even allowing for the nonrandom sample (see below), because only individuals belonging to the sorts of families meeting our criteria are included. Thus there is not representation of adults who live alone or with others who are neither parents nor spouses or married people having no children or whose children are all either below twelve or married.

Because of these exclusions, the total sample as constituted is probably more similar to the selected family groups than to a more fully representative sample of the society as a whole. Therefore, the differences in sharing found in the sample as a whole and sharing found within families are probably understatements of the true differences in sharing within families and sharing among community members from different families (i.e., community as a whole). Since the objective is to examine the hypothesis that the level of sharing of all cultural elements is high enough to serve as a (perhaps the ) main source of culture's ability to function as the basis for social life in functioning groups, an overstatement of sharing is more acceptable than an understatement that might wrongly falsify the hypothesis.

As indicated above, the families studied were not randomly selected. A sample based on such selection was not possible because of the demands interviewing made on the families and because of the widespread reluctance to discuss even the most distant family matters with nonmembers. We compensated for this as much as possible by choosing families so that there were representatives of the various mitaa (neighborhoods) and the range of socioeconomic and educational standings.

It should be noted that there was a difficulty in identifying families that considered themselves and were considered Swahili which was not properly resolved until the data had been collected. Because of complexities concerning the affiliations of the occupants of statuses of members of families whose rather recent forebears came from the Persian Gulf area as well as of members of families who are suspected of having slave ancestors, the initial decision (made before I had discovered the extent and profundity of the status problems as concerns group membership) to include members of families whom Swahili interviewers decided belonged to the Swahili community proved unacceptable. Among the families included by interviewers were some, I discov-

ered, that undoubted community members did not accept as "true" or "full" members.

To try to deal with this problem, I assembled a panel of four middle-aged Swahili informants well acquainted with all sections of the group. Each of these men (mature women are difficult to employ for private sessions with other male informants and a male researcher) was, without doubt, viewed by others as a member of the community and of one of its long-established constituent families. These informants examined the names and other demographic data (but nothing else) collected from families interviewed and eliminated families they considered noncommunity members (i.e., considered members of other ethnic groups) with the result that the Mombasa Swahili are represented by only seventeen families despite data having been originally collected from thirty. All results of interviews here draw only on the seventeen undoubted member families.

It may also be that community members descended from fairly recent Arab immigrants are underrepresented if they told interviewers they were not Swahili. This is unlikely, however, since the interviewing was done by four young Swahili (three men, one woman) who consistently erred on the side of including doubtful community members. Moreover, some of those included are known by me to claim descent from Omani and Yemeni forebears who emigrated relatively recently.

The four societies compared with the Swahili in the original study need not be examined or discussed at length here. The characteristics of these families and the data collected from them are reported in the original paper, which presents the findings regarding cultural sharing from all five societies, including the Swahili, and compares the results along the dimensions to be examined here (see Swartz 1982a ).