The results obtained by the authors have two problems.

First, the data set in Supplement 1 seems to be in error with regard to the variable, Close Kin, defined as a dyad with "r" ? 0.125. For the Ache and the Hadza data, the values of 1 = Close Kin and 0 = Other only partially correspond to this definition. For example, for the Hadza data, while there are 10 dyads with both "r" ? 0.125 and Close Kin= 1, there are 6 dyads with "r" ? 0.125 but Close Kin= 0 and 19 dyads with "r" < 0.125 and Close Kin = 1, for a total of 25 mismatches.

Second, answering ?no? to the question ?Has subject ever spoken with target?? would seem to mean that the subject and the target have not interacted -- or at least in not any substantive manner. Thus these dyads should be excluded from the data analysis since the focus is on patterns regarding who interacts with whom. For the Hadza data, 521 of the 851 dyads are coded as responding ?no? to this question.

When I recoded the Close Kin variable to match the r values and excluded the dyads where the subject has never spoken with the target, I obtain substantially different results than reported by the authors. I also divided the Hadza data into M-M dyads and F-F dyads to see if interaction events are sexually biased. (I did not redo the Ache data.)

If we form a 2 x 2 contingency table using the Kin/Nonkin and Ritual/Nonritual dimensions, and sum over all the questions (excluding the above question) to get the total number of interaction events for all dyads and all questions, then we obtain X² = 1.93 ("p" > 0.10, df =1) for the F-F dyads and X² = 1.65 ("p" > 0.10, df =1) for the M-M dyads. Thus the interaction rates for dyads related by biological kinship are independent of interaction rates for dyads related by ritual behavior, which implies that the processes leading to forming ritual relations are independent of biological kinship. (The authors do not consider, however, behavior patterns with regard to the Hadza emic notions of kinship expressed through their kinship terminology.)

For the F-F dyads, the probability, "p", of a possible interaction event actually occurring among females who are close kin is "p" = 0.76 and among those who are ritually related, the probability is essentially the same: "p" = 0.77. For those who are not close kin, "p" = 0.67 and for those who are not ritually linked, "p" = 0.64. Thus the interaction events are biased towards female who are already socially linked, either through their emic concepts about kinship or through their ritual behavior. For M-M dyads, the pattern is similar, but with slightly smaller probabilities: "p" = 0.67 for the probability of a possible interaction event occurring between close kin and "p" = 0.71 for interaction between males who are ritually related. For those who are not close kin, "p" = 0.58 and for those who are not ritually linked, "p" = 0.45). In brief, it appears that interaction events are biased towards pairs of individuals who are already socially linked to each other.

Of course, these results assume that the only error in the data set lies in the coding of the Close Kin variable, rather than there being an error such as sorting a single column independently from the rest of the spreadsheet.