What We Know and What We Don’t Know About Variation in Social Organization: Melvin Ember’s Approach to the Study of Kinship

1. The idea that male provisioning per se was important in explaining bonding has been recently questioned by evolutionary anthropologists—see review of literature in Marlowe (2003). Early provisioning, as we shall see shortly is another matter.

2. Korotayev (2003) found that very low female contribution was related to residence.

3. I (Ember, 1975) first suggested that the positive relationship between division of labor and residence in North America might be accounted for by the large number of hunting and gathering societies there. A first set of analyses did suggest that division of labor and residence were correlated in North America. However, if division of labor were really a factor we would expect that more hunting would predict patrilocality and more gathering would predict more matrilocality. Only the latter is supported, which casts doubt on the division of labor explanation. Fishing, which is not related to division of labor, predicts patrilocality, perhaps because high fishing societies are more apt to fight internally. The division of labor argument among hunter-gatherers has also been questioned by Korotayev (2001).

4. Korotayev (2001) has suggested that nonsororal polygyny is the masking factor. But since high male mortality in warfare predicts polygyny (see polygyny section of this article), these two things may be picking up the same type of masking effect.

5. Marlowe (2004) has measured residence differently, counting initial residence patterns right after marriage. This is not the traditional way to look at residence, so it is hard to compare results. However, he has some other suggestions about why some foragers would be more multilocal than others. See also Otterbein (2005).

6. The suggested curvilinear relationship was supported by Blumberg and Winch (1972).

7. Dow and Eff (2009) argued that after correcting for missing data and adding a variable to measure “culture trait transmission,” male mortality in warfare and pathogen stress no longer significantly predict nonsororal polygyny as found by M. Ember et al., (2007). However, 43% of the data in that study were imputed by Dow. Although the multiple impution method Dow and Eff use may be fairly robust, imputation methods have to contain more error than nonimputed data. Error is going to lower correlations. A fairer test would have been to delete the same proportion of data for “culture trait transmission” and impute that before comparing all the variables.