Abstract
In a previous publication, we reported that on the whole, young children in four cultures predictably made same-sex role choices (e.g., girls preferring daughter to son). But contrary to prediction, we found that children in two patricentric societies (patrilineal descent, etc.) made fewer same-sex role choices than did those from two societies without gender-biased social principles (bilateral descent, etc.) This surprising finding, which constituted an acknowledged puzzle for interpretation, prompted the current reanalysis of the dataset. Our new analysis has led to two conclusions: (a) children from the two patricentric societies, besides scoring lower on same-sex role choices, also scored a good deal lower on cognitive-performance tests, so they were generally poor test-takers; and (b) children in one of the patricentric societies—though not in any of the other samples—expressed a strong preference for higher-generation roles (i.e., mother and father). Discussion centers on plausible interpretations of these new-found results.
Almost a decade ago, we reported in these pages a set of results about role choices that in certain ways were expected, but in one particular way was surprising (Munroe, 2004). Here we have three main purposes: to treat the unexpected result and the probable underlying reason for it; to reanalyze the dataset in terms of a quite different hypothesis; and to discuss and interpret a new outcome. Initially, however, for the sake of clarity, we shall briefly describe the first study and its more predictable findings.
That first study consisted of asking young children, aged 3 to 9, a set of questions about their simple sex-role preferences. Each question was phrased to the child according to a dichotomous, sex-differentiated choice—for example, “Would you rather be a son or a mother?” The sex roles asked about were father, mother, son, daughter, baby boy and baby girl. Therefore, each of the three same-sex roles, male or female, was paired with each of the three other-sex roles, for a total of nine choices. Given the choices, a random selection by children would have resulted in a mean score of approximately four-and-one-half same-sex preferences for each sex within each culture group and at each age (all sample children were at ages of 3, 5, 7, or 9.) Any mean score above this would indicate that children were choosing own-sex roles at a nonchance level, viz., were systematically choosing their own-sex roles. The responses showed that for all four culture groups, all four age groupings, and for both boys and girls—thus 32 mean scores altogether—the averages were always above the midlevel point of 4.5. Choosing own-sex roles was not only as expected but it also was pervasive among the sample children.
We further expected that older children would choose same-sex roles even more frequently than would the younger children, and such proved to be the case in three of the four societies. Additionally, we anticipated, somewhat less surely, that at each age and in each culture group, boys and girls would be similar in their degree of same-sex choice, and this also was correct.
The surprising outcome was as follows. Two of the societies, the Logoli of western Kenya and the Newars of the Kathmandu Valley, Nepal, were organized into what are commonly termed patricentric systems (patrilineal descent, virilocal residence, strong sex stereotypes), and the other two, the Garifuna of Belize and Samoans of Ta’u, American Samoa, were organized without strong sex stereotypes and without gender-biased social principles (bilateral descent, and neolocal and bilocal residence, respectively). We predicted that in the two culture groups explicitly favoring and recognizing a sex-biased organizing principle, the children there would more frequently choose same-sex roles than in the other two samples. The outcomes, however, were opposite in direction from prediction, and strongly so: The same-sex preference scores of the Logoli and Newar children were absolutely lower than those of the Garifuna and Samoan children in 15 of 16 culture-by-age comparisons, and significantly lower in seven of these comparisons. As one example, among 5-year-olds, the mean same-sex preference score for the Logoli children was 5.8 and that for the Newars 6.1, but those for the Garifuna and Samoans were higher, with both at 7.1 (with, to repeat, a ceiling of 9 on the measure).
We attempted to account for this decidedly unpredicted set of results with a few post hoc interpretive gestures, but the fact was that these findings were an unexplained puzzle. Moreover, if salient cultural gender differentiation was not a factor producing more frequent same-sex role choices in two of the culture groups, then one might have expected no cultural differences at all. That is, we might have anticipated that the cross-cultural ubiquity of differential male and female roles would elicit similar levels of (age increasing) same-sex preference among children everywhere in the world, including among the sample children in each of these four culture groups. Yet, obviously, this had not happened either.
A clue to understanding came when we analyzed further data from the same testing session in which the sample children had engaged. The testing phase for each child, which lasted approximately 1.5 to 2 hr, included not only the role-preference measure discussed above but also a set of cognitive-developmental measures. The results from the latter measures, seven in all, were reported recently (Gauvain & Munroe, 2009). These cognitive tests showed a clear pattern of superior performance by the Samoan and Garifuna children, who outscored the Newar and Logoli children in 25 of 28 specific comparisons. This set of cognitive outcomes thus matched the cultural-difference findings from the role-performance measure.
Another similarity was that cognitive performance improved with age just as same-sex role preferences had increased among older children: For the cognitive tests, age-related improvement appeared on all measures (Gauvain & Munroe, 2009), and for the role preference, as noted above, the same was true except for one subsample (Newar children). Buttressing this assessment was the fact that at the individual level, the degree of same-sex role preference was significantly correlated with children’s scores on every one of the cognitive measures, and this pattern continued to hold, with one exception, 1 when age was introduced as a control variable.
Thus it seems clear that we had captured the sample children’s greater desire, at older ages, to adopt the gender role to which they had been assigned at birth. In the words of Parke and Buriel (1998), “[A]n individual’s standards . . . and behaviors change to conform to those regarded as desirable and appropriate for his or her . . . role in any particular society” (p. 463). But the coincidence of role-preference and cognitive outcomes prompted us to conclude that these findings reflected not a mysterious reversal of expectations but a mastery that differed in developmental timing for the Samoan/Garifuna children compared with the Newar/Logoli children.
Having conceded this, however, we wanted now to consider another element embedded in the role-preference measure. The structure of this simple measure was such that on many items, the paired concepts differed not only by sex but also by generation. 2 As described above, the choices were always sex-differentiated (e.g., mother vs. father), but in four of those choices there was also a built-in generational preference which the participants were obligated to make (e.g., mother vs. son, or baby girl vs. father), that is, upper generation versus lower. (For the five items that were not generationally differentiated, the child’s choice was between two roles of the same generation, e.g., mother vs. father, or son vs. daughter.)
It seemed to us that the patricentric bias we had earlier anticipated might still be detected if we rescored the role-preference items according to generational choices. The unilineal (patrilineal) organization of the Logoli and Newars might dispose children in these two culture groups to be conceptually biased toward choices that were hierarchically favored and that, ceteris paribus, these children would more often choose the roles embodying a higher-generational element. The standard ethnographies for the two culture groups explicitly set forth the importance of superior age. For the Logoli: “The authority of seniors over their juniors, the principle of seniority, prevails in all social relations among the Bantu Kavirondo” (Wagner, 1970, p. 77). For the Newars: “In a Newar family the gradation of age and generation is strictly adhered to in matters of mutual behaviour and privileges” (Nepali, 1965, p. 419). The new prediction therefore was made that Logoli and Newar participants would more frequently make upper-generation choices on the role measure than would Samoan and Garifuna children.
Given that cognitive-developmental performance had almost undoubtedly affected the degree of same-sex preference in role choices, it would be important in the current study to investigate whether cognitive levels also influenced the number of upper-generation choices. We take up this possibility in the analysis as a necessary control.
Method
Sample
Within each of the four sample communities, six boys and six girls at each of 3, 5, 7, and 9 years of age were selected for testing. The potential sample was reduced by the refusal of several 3-year-olds and two 5-year-olds to participate in all or parts of the testing session, and by the perseverating responses (all nine choices made according to either second-word offered or first-word offered) of several other children. The total number of children tested with analyzable responses was 173.
Procedure
The measures were administered to all 5- to 9-year-old children in a standard setting within each community. The 3-year-olds were tested in or near their own homes under conditions that provided as much privacy as possible. Each child was tested in the language he or she customarily spoke.
Role-Choice Measure
Sample children were presented with the following female versus male role combinations: mother-father, mother-son, mother-baby boy, father-daughter, father-baby girl, daughter-son, daughter-baby boy, son-baby girl, baby girl-baby boy. A child was asked, for example, “Would you rather be a daughter or a father?” Order of presentation and of pairing, not listed here, were the same for all children in all subsamples (as reported in Munroe, 2004). The original scoring had been based on the number of items on which the child chose the role corresponding to his or her own biological sex. The revised scoring for the current study counted the number of higher-generation choices made by each child for the following four “upper-” versus “lower-generation” pairs: mother-son, mother-baby boy, father-daughter, father-baby girl. Individual scores thus ranged from zero to four.
Plan of Analysis
Initial calculation illustrated that a child who had made all same-sex role choices (e.g., a boy always choosing father, son, baby boy) received a medium score of two on this new generation-preference measure. Because such children, totaling 64 across the four subsamples, had clearly favored same-sex roles and ignored generational possibilities, and had (unknowingly) chosen so that their preferences produced no bias along generational lines, they posed a conundrum for the reanalysis. Our solution was to test the hypothesis in both ways, first by excluding those 64 children who had made only own-sex role choices, then by including them. Thus the total sample size for the first reanalysis was 109, and for the second, 173.
Results
Upper-Generation Preferences
Table 1 indicates that mean scores for upper-generation choices among the patricentric Logoli of East Africa were easily the highest among the four communities. The Logoli mean of 3.07 was significantly higher than that of each of the other samples at the .01 level or better. As Table 1 also shows, none of the other two-culture comparisons revealed significant differences among themselves. Additionally, the Logoli children’s preference for upper-generation roles held for all age groups, that is, the Logoli means were the highest among each of the 3-, 5-, 7, and 9-year-olds subsets (data not shown here).
Higher-Generation Role Choices in Four Cultures
Note. Means are based on individual scores ranging from 0 to 4. Tests are one-tailed.
Interestingly, the mean for the other patrifocal culture-group, the Newars of the Kathmandu Valley, was the lowest of the four, though not significantly below any but that of the Logoli. At varying age levels, however, the Newar children’s scores moved in an irregular fashion and did not stay always at the bottom.
We can briefly report that when the 64 children who chose only same-sex roles were included in a followup analysis, the results looked very much the same: Two of the three Logoli-versus-”other” comparisons were somewhat weaker (Logoli vs. Samoans, Logoli vs. Newars), and one was a bit stronger (Logoli vs. Garifuna), but all three comparisons showed significant differences well beyond the .01 level. And again, none of the other two-culture comparisons (Garifuna vs. Newars, Garifuna vs. Samoans, Newars vs. Samoans) showed significant differences.
Inspection of individual scores showed clearly the Logoli children’s greater preference for upper-generation roles. Of the Logoli children who were above or below the mid-point score of 2, 21 of 22 made a preponderance of upper-generation preferences. Using the same criterion, we find that the three other culture-groups showed such preferences less often: 67% for the Garifuna, 57% for the Samoans, and 50% for the Newars. But among the Logoli, as we see, this figure was at 95%.
As to cognition, in none of the culture-groups, including the Logoli, did the scores on the varied cognitive tests (as reported in Gauvain & Munroe, 2009) show significant association with generational preferences (all runs were age-controlled). As well, nothing in the way of other potentially important variables (e.g., education, father absence, socioeconomic status) accounted for the decisive difference in generational preference between Logoli children and those in the Garifuna, Samoan, or Newar communities.
Sex-Role Preferences
Some unanticipated sex differences did appear in the midst of the generational choices. (These were unexpected because, as reported in the earlier article [Munroe, 2004], there were no significant differences between boys and girls in their overall preferences for same-sex roles.) Of the four upper-generation choices, two involved the possibility of choosing the father role (father vs. daughter, father vs. baby girl), and two the possibility of choosing the mother role (mother vs. son, mother vs. baby boy). To exemplify what such distinctions entailed, we give an instance: When the role of father was paired with that of baby girl or daughter, Logoli boys chose father 83% of the time and Logoli girls chose it at 65%, a nonsignificant difference; whereas when Garifuna boys were given these choices, they chose father 90% of the time, and Garifuna girls chose it at only 25%, a very strong difference. Altogether for these choices, among the Garifuna and Samoan children, boys more often than girls preferred the father role, respective t-values = 6.33 (df = 20) and 3.91 (df = 24); both p’s < .001, two-tailed; and among the Newar and Samoan children, girls more often than boys preferred the mother role, t-values = 4.28 (df = 24) and 2.14 (df = 32); p values at < .001 and < .05, respectively, two-tailed. Note that none of these differences occurred in Logoli children’s choices.
Discussion
In this reanalysis we have accounted for generational-preference scores in three of the communities: The Logoli, as a patricentric group, displayed a clear leaning toward upper-generation roles while the Garifuna and Samoans, nonpatricentric culture-groups, gave no indication of such preferences, with all of this as predicted. The one problem, then, resides in the scores of the Newar children. As noted in the introduction, a pattern of age-privilege has been explicitly delineated in the ethnographic description of the Newars (Nepali, 1965), but our evidence for them yielded none of the expected upper-generation bias.
We have no special insight regarding the anomaly of the Newar children’s responses. Their answers to the role-measure questions did reveal a somewhat higher variance than those of children in the other culture-groups (Table 1), and this heterogeneity might indicate that they understood the dichotomous choices offered them less well than did the children elsewhere. The supposition is supported by their generally lower cognitive-skills scores. As stated above, and as already reported in the earlier article analyzing this same dataset, the Newars were also the only subsample in which increasing age failed to be significantly associated with same-sex choices on the role measure (Munroe, 2004). But their low mean score does illustrate that generational-preference scores were not a simple inverse of the sex-preference scores reported earlier (Munroe, 2004). If this were so, low scores on the same-sex choices (Logoli, Newars) would have mathematically resulted in high scores on upper-generation choices, and then the Newar means and the Logoli means would have been similar to each other in the current study, which of course they were not.
Specific preference by boys (vis-à-vis girls) for the father role (among Samoans and Garifuna) or by girls (vis-à-vis boys) for the mother role (among Samoans and Newars) did appear in our present analysis. This indicated that, despite an absence of overall gender-related sex-role preferences (i.e., for all male roles by boys vis-à-vis for all female roles by girls), there were some sex-specific interests in three of the samples. Nevertheless, as reported, the Logoli children did not display these biases, which reflected the fact that both their boys and girls overwhelmingly made upper-generation choices without respect to the gender distinction.
Degree of political hierarchy was not the basis for our results. The only culture group in our sample to possess a traditional chiefly system were the Samoans, whose matai custom meant that a titled person was represented in every extended family (Holmes & Holmes, 1974/1992). The Logoli did have a position of chief, a colonially imposed role that was not present in their traditional clan system (Wagner, 1970). If this latter role did somehow affect the scores of Logoli children, then the same should have been true for the responses of the Samoan participants. Yet the generational-preference measure yielded only a Samoan mean score of 2.12, which is basically neutral, and is much lower than the Logoli mean of 3.07.
Cultural differences in kin terminology pose an outside possibility for understanding the Logoli children’s strong preferences for the terms mother and father. The Garifuna, 3 Newars, and Samoans all possess Hawaiian kinship terminology systems, which lump together all upper-generation relatives by one term if female (this term includes ego’s mother) and by one other term if male (this term includes ego’s father; Ember & Ember, 2011). In contrast, the Logoli use an Iroquois terminology, which generalizes the mother and father terms less broadly, and which therefore might have made the terms more salient within the framework of our testing. Such a possibility could form a basis for further research.
“Data mining,” that is, the continual analysis of a dataset until findings finally emerge, is a concern in the current study, and it cannot be discounted. In this case, however, we have been dealing with only a single measure that tapped just two dimensions, sex and generation. Since we had analyzed the role preferences according to gender choices in our earlier study (Munroe, 2004), we were left with the remaining dimension of generation for this second round of analysis (and, as reported above, certain sex-role preferences also subtly reemerged in the current analysis). Data mining, then, may have been present, but it was less likely for our main result here (in Table 1) than for other data sets containing a multiplicity of variables.
We have repudiated our earlier conclusion that cultural differences in sex-role choices, reversed as they were from initial expectations given the structural characteristics of our sample societies, might still be understood in certain tortuous theoretical terms (Munroe, 2004). Besides our introductory reanalysis above, with results showing that the sex-role choices were in fact positively associated with overall cognitive performance, another reason for skepticism about our interpretation in the early study was that the findings were not convergent with what was already generally thought to be “known.” As Medin (2011) has put it, [Scientists] correctly infer that they will need a strong pattern of evidence to overturn convention, not because convention is good per se, but because convention may be based on hundreds, and perhaps thousands, of studies”
(p. 18). In the present analysis, though, we believe that the findings—for three of the four culture-groups, but not the Newars—do follow from the structural makeups of the societies in which the children were being reared. To that extent, ordinary information-processing activities would seem to have guided the choices of the majority of children in our study samples (Nisbett & Ross, 1980). So in a larger sense, the pattern we have found in this reanalysis tends to be consistent with general expectations.
Age and sex form a core basis for relationships in all societies (Linton, 1945), and our results illustrate one small way in which the play of these factors can be seen among young children. The strong preference for upper-generation roles among the Logoli, and the absence thereof among the Garifuna and Samoans, points to the significance of age-based distinctions as grounds for continuing developmental research. Although this preference did not appear among Newar children, unpatterned responses were also characteristic of the Newar sample in our earlier analysis (Munroe, 2004), and we would argue that the irregularity does not vitiate the thrust of the overall findings. Sex differences appeared as well when we considered preference for the specific roles of mother and father, with all samples except the Logoli showing a sex-based bias for one or the other (and, in the Samoan case, for both). Among Logoli children, mother and father were high in the choices of both girls and boys, thus diminishing the likelihood of any sex-based preference for these roles in that particular sample. Altogether, then, this reanalysis tends to show that a simple measure of preferences for generationally and sex-based roles can reveal meaningful cultural differences, and that such a measure, and similar ones, might usefully be applied to developmental research in other cultural contexts.
Footnotes
Acknowledgements
The author acknowledges with gratitude the cooperation of the villagers who participated in the investigation and the contributions of the research assistants in each of the sample societies. Ronald Macaulay, Susan Seymour, and Claudia Strauss made valuable suggestions for improvement of an earlier version of this article.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The field research was supported by a grant to Robert L. Munroe and Ruth H. Munroe from the National Science Foundation.