MATRILATERAL BIASES IN THE INVESTMENT OF AUNTS AND UNCLES

Replication in a Population Presumed to Have High Paternity Certainty

Dona ld H. McBurney, Jessica Simon, Steven J. C. Gaul in University of Pittsburgh

Allan Gel iebter St. Luke's-Roosevelt Hospital, Columbia University, and Touro College,

New York City

Gaulin, McBurney, and Brakeman-Wartell (1997) found that college stu- dents reported both matrilateral and sex biases in the investment of aunts and uncles (aunts invested more than uncles). They interpreted the matri- lateral bias as a consequence of paternity uncertainty. We replicated that study with Orthodox Jewish college students, selected because they come from a population we presume to have higher paternity certainty than the general population. The Orthodox sample also showed matrilateral and sex biases. Comparing the two data sets, the Orthodox sample reported more investment, and slightly less matrilateral and sex biases, but the dif- ferences were not statistically significant. We did find an interaction be- tween sex of relative and group membership, resulting from greater investment by Orthodox uncles. We interpret the results as reflecting the operation of a facultative investment mechanism whose upper limit is tuned to the maximum levels of paternity certainty found in ancestral en- vironments. Lack of a difference in matrilateral bias between groups may result from levels of paternity certainty near to, or above, that maximum in both groups.

KEY WORDS: Kin investment; Matrilateral bias; Paternity uncertainty

Received March 30, 2001; accepted June 1, 2001.

Address all correspondence to Donald H. McBurney, Department of Psychology, Univer- sity of Pittsburgh, Pittsburgh, PA 15260. E-mail: mcburney@pitt.edu

Copyright 2002 by Walter de Gruyter, Inc., New York Human Nature, Vol. 13, No. 3, pp. 391-402. 1045-6767/01/$1.00+.10

391

392 Human Nature, Vol. 13, No. 3, 2002

Dear Grandmothers: Please don't play favorites. If you buy a gift for your daughter's children, please buy something of equal value for your son's chil- dren . . . . It is always the daughter's children who are favored.--Your Son's Wife (Letter to "Dear Abby," 18 December 2000, Pittsburgh Post-Gazette)

Dear Ann Landers: My husband and I have been married for 16 years and we have three wonderful children. His sister recently had her first baby. My mother-in-law treats this new little boy as if he were her only grandchild. Morn has never been close to my children but now she ignores them com- pletely. When I asked Morn about her lack of interest, she replied, "I know this baby is my true grandchild, but I have no proof that your children be- long to my son." (16 August 2001, Indianapolis Star)

Paternity uncertainty has been shown to affect the investment tendencies not only of putative fathers, but also of distal kin (Gaulin, McBurney, and Brakeman-Wartell 1997, and references therein). Euler and Weitzel (1996) asked German college students to rate the solicitude shown them by their grandparents. They found both an effect of laterality (maternal > paternal grandparents) and sex of grandparent (grandmothers > grandfathers). Gaulin and colleagues found the same effects in an American sample that rated the solicitude of their aunts and uncles; they further derived esti- mates of paternity uncertainty between 13% and 20% and from 9% and 17% in the two sets of data, respectively. These estimates agree well with the median estimate of paternal discrepancy of 9% given by Baker and Bel- lis (1995). 1 Hoier, Euler, and H~inze (2001) replicated the study with Ger- man college students. They found the laterality bias to be similar to that reported by Gaulin and colleagues, although the sex-of-investor bias was larger, and overall solicitude was less.

Gaulin and colleagues raised the question: "Is the observed matrilateral bias in kin investment by aunts and uncles a response to (pre-)historically typical levels of [paternal uncertainty] in human populations or is it more facultatively dependent on actual current values?" They noted that "The existence of societies with fully developed matrilineal inheritance (Hat- tung 1985) suggests more facultative control, but selection in past environ- ments can set limits on the range of responsiveness" (Gaulin et al. 1997:149).

The present s tudy pursued this question by replicating the earlier s tudy with a population expected to have high paternity certainty. We wished to test whether the level of maternal bias found in our previous study and the similar study by Euler and Weitzel was facultatively tuned to modern, western societies within an unknown range from zero certainty to some high level, or whether it represented the highest degree of paternity cer- tainty typically experienced in ancestral environments.

Matrilateral Biases in the Investment of Aunts and Uncles 393

We chose Orthodox Jews as a category of college students deriving from a population with high paternity certainty. The traditional family patterns of this population strongly encourage fidelity in marriage (e.g., Biale 1984; Goldman 2000). Discovered adultery would result in divorce and os- tracism. The concept of modesty (tzniut) regulates dress and behavior (Donin 1972; Manolson 1997). Although applying to men as well, modesty requires that women especially do not dress or act provocatively. Interac- tions between husband and wife are governed by a specific corpus of Jew- ish law, known as Laws of Purity (Taharat Mishpacha). Sex within marriage is a holy act, albeit subject to certain important restrictions. It is the duty of the husband to satisfy his wife's sexual needs. Sex is not permitted from the start of menstruation until seven days after its end, thus concentrating sexual activity in the most fertile part of the cycle.

Our expectation of high paternity certainty in Orthodox Jews is strongly supported by genetic evidence. Jewish priests (Kohanim) from two geo- graphically separate lines (Ashkenazic and Sephardic) have recently been found to share Y-chromosome haplotypes that differ from those of other Jewish men (Skorecki et al. 1997). Boster, Hudson, and Gaulin (1999) demonstrate that such differences would require a very high degree of pa- ternity certainty in this line because this priesthood is hereditary through a male line that originated about 3,300 years ago. Their calculations lead to a paternity certainty between .988 and .996 per generation over the course of about 132 generations.

Other considerations leading to the choice of Orthodox Jews over some other groups include the tendency of Jewish identification to be stable over generations and their high college participation, the latter making for com- parability to our previous study.

METHODS

Participants

Participants were drawn from a medium-sized urban college in the northeastern United States, 95% of whose students are Orthodox Jews, 5% of whom are Hassidic. Two hundred seventy-eight students completed a survey in class. Data were analyzed from 251 who reported that both of their biological parents were living, and living together when the partici- pant began college. Of the 251, 132 were male and 119 were female. Fewer than 10 declined to complete the survey. Several weeks later the students were given written feedback about the purpose of the study. The study was approved by the college Institutional Review Board.

394 Human Nature, Vol. 13, No. 3, 2002

The Survey

The questionnaire was identical to that used in our previous study (Gaulin et al. 1997). The first part asked about the level of concern shown by matrilateral and patrilateral aunts and uncles on a seven-point scale. Half were asked first about the matrilateral relatives. Each class of relative was defined to distinguish biological relatives from genetically unrelated relatives. Thus, four separate questions asked, "How much concern does the maternal (paternal) uncle (aunt) show about your welfare?"

The second part of the questionnaire asked forced-choice questions. "If you have both a maternal and paternal (order was counterbalanced) uncle (aunt), which one shows more concern about your welfare?" Other ques- tions determined the ages of the relatives and the distances the relatives lived from the participant's home (defined as the place they left to go to college). Further details about the questionnaire and the methods can be found in the previous study.

The Data Set and Analyses

Seventy participants had both matrilateral and patrilateral aunts and matrilateral and patrilateral uncles. An additional 46 had matrilateral as well as patrilateral aunts but not a matched set of uncles, and 60 had ma- trilateral as well as patrilateral uncles but not a matched set of aunts. We performed analyses for those 70 who had all four relatives, and also for all 116 (70 + 46) who had matrilateral and patrilateral aunts and for all 130 (70 + 60) participants who had matrilateral and patrilateral uncles. Certain participants' data were therefore used for more than one analysis, so the analyses we report are not entirely independent, Because we ignored the data of participants who did not have matched pairs of relatives, all sub- jects served as their own controls, as in our previous study.

Since the Pittsburgh sample may have included a small number of Orthodox Jews, any differences between the two groups may slightly un- derestimate the actual difference between Orthodox Jews and the non- Orthodox (Pittsburgh) population.

RESULTS

All data were analyzed using BMDP mainframe programs 4F, 1R, 2R, and 2V.

Forced-Choice Data

Those who had all four classes of relatives. The subset of subjects who had at least one matrilateral and one patrilateral aunt and, likewise, at least one

Matrilateral Biases in the Investment of Aunts and Uncles 395

matrilateral and one patrilateral uncle answered the forced-choice ques- tion twice, once for aunts and once for uncles. In our previously published analysis of the Pittsburgh data (Gaulin et al. 1997) we showed that the dis- tribution of choices (Both matrilateral: One matrilateral, One patrilateral: Both patrilateral) deviated significantly from the expected 1:2:1 ratio that would result from random distribution of two responses (greater solici- tude from matrilateral versus patrilateral aunts and uncles). For this analysis the question is: Does the Orthodox sample (n = 72) show a signif- icantly weaker matrilateral bias than the Pittsburgh sample (n = 73)? We used a contingency table approach to answer this question. The data are shown in Table 1. The populations do differ in the predicted direction (fewer matrilateral choices by the Orthodox) but only slightly and non- significantly (X2 = 2.3; df = 2; p = .32).

Those who had pairs of aunts or pairs of uncles. Larger samples can be drawn from the data sets if we insist only that a subject had a matrilateral and pa- trilateral aunt or a matrilateral and patrilateral uncle. This rule still en- forces the principle that each subject should evaluate an equal number of patrilateral and matrilateral relatives, and thus eliminates a major poten- tial source of bias. Similar contingency table analyses can be constructed for the forced-choice data regarding aunts and the parallel data regarding uncles.

For those subjects who had both patrilateral and matrilateral aunts, the Orthodox sample shows slightly less matrilateral bias (79/116 = 0.68) than does the Pittsburgh sample (99/138 = 0.72) but the difference fails to reach statistical significance (X2 = 0.85; df = 1; p = .36).

A very similar pattern emerges for those subjects who had both kinds of uncles. The orthodox sample shows somewhat less matrilateral bias (77/128 = 0.60) than does the Pittsburgh sample (84/132 = 0.64), but the difference is far from significant (~2 = 0.33; df = 1; p = .56).

None of the preceding analyses controls for possible factors moderating the investment of second-degree relatives. As in our previous analysis (Gaulin et al. 1997), and in other studies of investment by second-degree relatives (Euler and Weitzel 1996; Pashos 2000), we thought it appropriate

Table 1. Forced Choice Data by Study Population for Subjects Who Had All Four Classes of Relatives

Choices

Sample Both Matrilateral One Matrilateral, One Patrilateral Both Patrilateral

Pittsburgh 33 30 10 Orthodox 33 23 16

396 Human Nature, Vol. 13, No. 3, 2002

to control for any systematic age and distance effects. Older relatives may have more resources to invest; more-distant relatives may find investment opportunities scarcer. In considering these effects we analyzed the aunt and uncle data separately. In each analysis the choice of whether the pa- trilateral or matrilateral relative was more solicitous was regressed on both the age difference between the patrilateral and matrilateral relatives and the distance difference between them and the subject.

Among those subjects who had both patrilateral and matrilateral aunts the regression planes were not significantly different between the Pitts- burgh (n = 134) and Orthodox (n = 109) groups (F = 2.044; df = 3, 237; p = .108), although the y-intercept of the Orthodox group was slightly less ma- trilaterally biased (.336 vs..305). Because this effect was in the predicted direction, and because the (marginally significant) test for coincidence of planes might not be maximally sensitive to a difference in y-intercept---our key prediction--we undertook further analysis.

This involved building two separate regressions of the forced-choice variable on the age- and distance-difference variables. The reduced model considered only these three variables. The full model included these same three variables plus three additional independent variables: a dummy (0,1) variable for group membership (Pittsburgh vs. Orthodox) and two in- teraction terms that multiplied the group membership variable by each of the difference variables. A comparison of the sum of squares between the two models permits an F-test of the differences in y-intercept (Kleinbaum et al. 1988). This result also failed to reach significance (F = 0.27; df = 3, 237; p > .5).

Among those subjects who had both patrilateral and matrilateral uncles, the regression planes were also not significantly different between the Pittsburgh (n = 128) and Orthodox (n = 123) groups (F = 1.073; df = 3, 245; p = .361). In this case as well, the y-intercept of the Orthodox group was slightly less matrilaterally biased (.418 vs..352). Because the test for co- incidence of regression planes was not even marginally significant, we un- dertook no further analyses of these data.

Rating-Scale Data

To analyze the investment ratings of the various classes of relatives we used repeated-measures analysis of covariance. In all these analyses the covariates were the ages and residential distances of the relatives in ques- tion, and the between-subjects (grouping) factor was membership in either the Pittsburgh or the Orthodox sample. For the subjects who had pairs of aunts or pairs of uncles but not both, laterality of the relative served as a within-subject factor. For subjects who had all four kinds of relatives, sex of relative (aunt vs. uncle) was added as a second within-subject factor. On

Matrilateral Biases in the Investment of Aunts and Uncles 397

the assumption of higher paterni ty certainty in the Orthodox sample, the prediction is a significant laterality-by-grouping-factor interaction.

Those who had all four classes of relatives. In this group there is a highly sig- nificant main effect for sex of relative (F = 26.05; df = 1,134; p < .0001) and a significant effect of laterality (F = 9.66; df = 1,134; p = .002). See Table 2. Adjusting for age and distance effects, relatives in the Orthodox sample were rated as investing more than those in the Pittsburgh sample (mean rating scale values of 4.32 vs. 3.96, respectively), but this difference be- tween groups failed to reach significance (F = 3.31; df = 1,134; p --- .071). More importantly, this grouping variable did not interact significantly wi th laterality (F = 0.10; df = 1, 134; p = .75). As an addit ional note, there was a significant sex-of-relative-by-group-membership interaction (F = 4.00; df = 1, 134; p < .047), resulting from the fact that virtually all of the in- crease in investment by Orthodox relatives was due to uncles (see below).

Those who had pairs of aunts or pairs of uncles. In terms of their structure these analyses precisely parallel the one reported above, with the excep- tion that the sex-of-relative level is omitted.

For those subjects who had patrilateral and matrilateral aunts there is a highly significant laterality effect (F = 30.82; df = 1,246; p < .0001). There was no main effect for group (Pittsburgh vs. Orthodox) (F = 0.12; df = 1, 246; p = .72), nor any group-by-laterality interaction (F = 0.10; df = 1, 246; p = .75).

Likewise, for those subjects who had patrilateral and matrilateral uncles there is a highly significant laterality effect (F = 11.31; df = 1,252; p = .0009). Among uncles the main effect for group (Pittsburgh vs. Orthodox) is highly

Table 2. Concern Rating (and Standard Deviation) as a Function of Sex, Laterality, and Sample, Adjusted for Age and Distance of Relative (distance between residential locations--see text)

Sample

Pittsburgh Orthodox

AUNTS

Matrilateral 4.77 (1.84) 4.83 (1.91) Patrilateral 4.02 (1.87) 4.20 (1.63)

UNCLES

Matrilateral 3.70 (1.93) 4.27 (2.04) Patrilateral 3.35 (1.71) 3.97 (1.76)

398 Human Nature, Vol. 13, No. 3, 2002

significant (F = 17.51; df = 1,252; p < .0001), but the group-by-laterality in- teraction is not significant (F = 0.11; df = 1,252; p = .75).

DISCUSSION

The Orthodox Jews did not show significantly less matrilateral bias than the Pittsburgh sample, although the difference was in the predicted direc- tion. Two explanations can be suggested. First, the Orthodox may not ac- tually have higher paternity certainty than the general population. This seems unlikely given the considerations discussed in the introduction. We favor a second explanation" The degree of bias found in these two studies, as well as in studies of German grandparents (Euler and Weitzel 1996; Pashos 2000), reflects the upper bound of the reaction range found in the ancestral environment of this trait.

One unexpected finding in the present s tudy was an interaction be- tween sex of relative and group membership, resulting from the fact that virtually all of the increase in investment by Orthodox relatives was due to uncles. This, in fact, was the only significant difference between the two studies. This finding could be taken as evidence of greater paternity cer- tainty in the Orthodox. Paternity certainty, however, should affect aunts and uncles equally, because laterality is independent of sex of investor in these data. The effect reflects a sex difference between the Pittsburgh and Orthodox data, not a laterality effect. We note, however, that if humans were perfectly monogamous there should be no sex effect, because the re- productive interests of men and women would be identical; thus the sex difference in investment would disappear. It seems more likely that the greater investment by Orthodox men is the result of some unknown cul- tural factor independent of the sex or laterality effects of concern in this paper.

Pashos (2000) studied investment by German and Greek grandparents using methods similar to those of Euler and Weitzel (1996). He found that paternal grandfathers were seen as investing more than maternal grand- fathers among rural Greeks, unlike urban Greeks and Germans. All groups of subjects showed greater investment by grandmothers than grandfa- thers. As did Gaulin et al. (1997), Pashos argues that sex bias should be considered separately from the laterality bias, reflecting the factorial de- sign of all of these studies. He then argues that the matrilateral bias found by previous investigators and in his non-rural Greek data was not the re- sult of paternity uncertainty. Rather, it reflects "a socially engendered fa- voring of maternal relatives in Western industrial societies as opposed to the favoring of paternal [grandparents] seen in the patrilateral [sic (patriarchal)] culture of rural Greece" (Pashos 2000:97). It seems more rea-

Matrilateral Biases in the Investment of Aunts and Uncles 399

sonable to us to consider the rural Greek data to reflect the effect of a patriarchal system acting to override a (universal) matrilateral bias. Sup- porting this interpretation is the bias shown in Pashos's rural Greek data toward investing in grandsons over granddaughters. The patrilateral bias was significant toward grandsons only. Granddaughters reported a slight and nonsignificant patrilateral bias.

Further, Pashos interprets the proximate cause of the matrilateral bias in western cultures to be the stronger family bonds of women compared with those of men. Indeed, we believe that this phenomenon may well be the proximate cause of matrilateral bias; our question concerns the ultimate cause, for which paternity certainty is the only current contender.

In all studies to date using aunts and uncles (Gaulin et al. 1997; Hoier et al. 2001; and the present study), the matrilateral bias was somewhat stronger among aunts than uncles, though this sex difference (assessed in terms of a sex-by-laterality interaction) failed to reach significance in two of the three studies. Speculating on this trend, Gaulin and colleagues raised the possibility that accurate information about relatedness may be more available to females. Hoier and colleagues suggest that, whatever cues are available, men should have evolved greater sensitivity to them. They present data showing no sex difference in estimates of cuckoldry rates. However these data confound sensitivity and availability since the cuckoldry estimates offered by each sex are affected by some unmeasured combination of these two factors.

In a further attempt to explain the sex difference in the degree of matri- lateral bias, Euler and colleagues (2001) reanalyzed the data of Euler and Weitzel (1996). They hypothesize a same-sex bias (a preference for invest- ing in the children of the offspring of the same sex, i.e., grandmothers pre- ferring daughters' children and grandfathers preferring sons' children) and argue that such a bias would cause the matrilateral bias to be stronger among female relatives. Their argument is logically correct and it may well reflect the actual dynamics of these relationships. Nevertheless two mat- ters need to be kept in mind. First, the sex difference in the magnitude of the laterality effect often fails to reach statistical significance; for example, the sex-by-laterality interaction is statistically insignificant in our Pitts- burgh sample, our Orthodox sample, and even in our more powerful com- bined sample. Second, none of the ideas advanced to explain the sex difference can explain the overall matrilateral bias itself, neither in proxi- mate nor in ultimate terms.

In summary, the present study replicated the results of our earlier s tudy in finding both a matrilateral and a sex bias in investment by aunts and un- cles in an Orthodox Jewish population. Comparing the data from both studies, we found an interaction between sex of relative and group mem- bership, as a result of greater investment by Orthodox uncles than by the

400 Human Nature, Vol. 13, No. 3, 2002

uncles in the Pit tsburgh data set. Or thodox subjects repor ted more invest- ment than Pi t tsburgh subjects, but the result did not reach significance. Or- thodox subjects d id not perceive the investment of their aunts and uncles to be significantly less biased by laterality or sex than did the Pit tsburgh sample. We conclude that the degree of bias found in this and comparable studies reflects a mechanism of discriminative solicitude that is faculta- t ively dependen t on paterni ty certainty; the up p e r b o u n d of the reaction range of this mechanism, however , is l imited by the level of paterni ty cer- tainty that obtained in ancestral environments .

We thank Carol Baker for statistical advice and Carey D. Balaban for helpful dis- cussion.

Donald McBurney is a professor of psychology at the University of Pittsburgh. His principal interest is in the psychophysics of taste and smell. Current research con- cerns sensory adaptation as an evolutionary adaptation, with a focus on long-term adaptation to chili peppers.

Jessica Simon conducted this study as her undergraduate honors thesis at the Uni- versity of Pittsburgh. She is currently pursuing graduate work.

Steven Gaulin is a professor of anthropology and psychology at the University of Pittsburgh. Trained at UC Berkeley and Harvard, his research focuses on the inter- section of sexual selection and evolutionary psychology. He is author, with Donald McBurney, of Psychology: An Evolutionary Approach.

Allan Geliebter is a psychologist at St. Luke's-Roosevelt Hospital, the University Hospital of Columbia University College of Physicians and Surgeons, and the De- partment of Psychology, Touro College, New York City. His research concerns food intake and obesity in humans.

N OTE

1. Three additional studies estimating paternity discrepancy have recently come to our attention. Sasse and colleagues (1994) found nonpaternity in a Swiss sample to be around 0.8%. Sykes and Irven (2000) found the genetic similarity in a sample of males bearing the name "Sykes" to be compatible with a nonpaternity rate of 1.3% per generation. Creda-Flores and colleagues (1999) estimated the non- paternity rate in a Mexican sample to be 11.8%. Thus the difference in nonpatemity between our Pittsburgh sample and the Orthodox sample may not be as great as we supposed. Our conclusion that the reaction range of the matrilateral bias may be limited by conditions that obtained in ancestral environments is unaffected by these contemporary data.

Matrilateral Biases in the Investment of Aunts and Uncles 401

REFERENCES

Baker, R. R., and M. A. Bellis 1995 Human Sperm Competition. London: Chapman and HaIL

Biale, R. 1984 Women and Jewish Law: An Exploration of Women's Issues in Halakhic Sources.

New York: Schocken Books. Boster J. S., R. R. Hudson, and S. J. C. Gaulin

1998 High Paternity Certainties of Jewish Priests. American Anthropologist 100: 967-971.

Creda-Flores, R. M., S. A. Barton, L. F. Marty-Gonzales, L. F. Rivas E, and R. Chakraborty

1999 Estimation of Nonpaternity in the Mexican Population of Nuevo Leon: A Validation Study with Blood Group Markers. American Journal of Physical An- thropology 109:281-293.

Donin, H. H. 1972 To be a Jew. New York: Basic Books.

Euler, H., and B. Weitzel 1996 Discriminative Grandparental Solicitude as Reproductive Strategy.

Human Nature 7:39-59. Euler, H. A., S. Hoier, and P. A. Rohde

2001 Relationship-Specific Closeness of Intergenerational Family Ties: Find- ings from Evolutionary Psychology and Implications for Models of Cultural Transmission. Journal of Cross-Cultural Psychology 32:147-158.

Gaulin, S. J. C., D. H. McBurney, and S. L. Brakeman-Wartell 1997 Matrilateral Biases in the Investment of Aunts and Uncles. Human Nature

8:139-151. Goldman, A. L.

2000 Being Jewish. New York: Simon and Schuster. Hartung, J.

1985 Matrilineal Inheritance: New Theory and Analysis. Behavioral and Brain Sciences 8:661-688.

Hoier, S., H. A. Euler, and M. H/inze 2001 Diskriminative Fiirsorglichkeit von Tanten und Onkeln: Eine empirische

Untersuchung aus evolutionspsychologischer Perspektive. Zeitschr~fiir D~:- ferentielle und Diagnostische Psychologie 22:205-215.

Kleinbaum, D. G., L. L. Kupper, and K. E. Miller 1988 Applied Regression Analysis and Other Multivariate Methods. Boston: PWS-

Kent. Manolson G.

1997 Outside~Inside: A Fresh Look at Tzniut. Southfield, Michigan: Targum. Pashos, A.

2000 Does Paternal Uncertainty Explain Discriminative Grandparental Solici- tude? A Cross-Cultural Study in Greece and Germany. Evolution and Human Behavior 21:97-109.

402 Human Nature, Vol. 13, No. 3, 2002

Sasse, G., H. Muller, R. Chakraborty, and J. Ott 1994 Estimating the Frequency of Nonpaternity in Switzerland. Human Hered-

ity 44:337-343. Skorecki K., S. SeUg, S. Blazer, R. Bradman, N. Bradman, P. J. Waburton, M. Ismajlowicz, and M. F. Hammer

1997 Y Chromosomes of Jewish Priests. Nature 385:32. Sykes, B., and C. Irven

2000 Surnames and the Y Chromosome. American Journal of Human Genetics 66:1417-1419.