1.1. Cost of helping

An initial suggestion concerns the cost of help. Although KST does not rule out the evolution of nonkin altruism, it does imply that it cannot be selected unless there is a return benefit to the altruist or the altruist's kin (but see Fehr & Henrich, 2003, Gintis, 2000, Richerson et al., 2003, for discussion of how genetic or cultural group selection could produce altruism in the absence of any such benefit). There are various channels through which return benefits could come, e.g., through the reciprocation of help, an increase in mating opportunities, or an enhancement of the altruist's reputation (Gurven, 2004). However, it is never guaranteed that altruism will bring a return benefit, and the greater the cost of altruism, the greater the net direct fitness cost if it does not. This is less problematic when the recipient is a genetic relative, because the direct fitness cost may be compensated by the indirect fitness benefit associated with aiding relatives. So, although people may be altruistic toward nonkin, one would expect that this would be somewhat dependent on the cost of help. This leads to the following hypothesis:Hypothesis 1.

As the cost of helping increases, the share of help given to kin will increase, and the share given to nonkin will decrease.

1.2. Levels of reciprocal exchange

A second suggestion involves considering KST in tandem with Trivers's (1971) reciprocal altruism theory (RAT), according to which altruism can enhance direct fitness as long as there is sufficient probability that it will be reciprocated. Again, reciprocation is less important among kin than among nonkin, because the indirect fitness benefits of helping kin can outweigh the direct fitness costs of unreciprocated help. This has led to the suggestion that kin will exhibit a lower level of reciprocity than nonkin, a hypothesis that has found some support (Berté, 1988, Essock-Vitale et al., 1980, Essock-Vitale & McGuire, 1985, Hames, 1987, Gurven et al., 2001). There is a complication, however. Some commentators argue that RAT does not provide an adequate explanation for altruism among close friends (Roberts, 2005, Silk, 2003, Tooby & Cosmides, 1996). People are often angered by the suggestion that their friendships are founded on the exchange of favors, and deny that when they help their friends they do so with the expectation of immediate repayment. This rules out strict tit-for-tat reciprocity as a model for altruism among friends. However, equity and reciprocity are nonetheless important in friendships (Silk, 2003). It remains possible, therefore, that friendships are founded on reciprocity but that exchanges of help among friends take place within an extended timeframe, with friends tolerating longer periods of imbalance. This leads to the hypothesis that there will be a higher level of reciprocity among acquaintances than among friends. However, because kinship reduces the need for reciprocation, the level of reciprocity found among kin will be lower still than that among friends.Hypothesis 2.

The association between help given and help received will be larger for acquaintances than for friends, and larger for friends than for siblings or cousins.

2.1. Participants

Participants were 295 undergraduate psychology students: 146 (49.5%) males and 149 (50.5%) females. The minimum number of participants needed for each experimental condition was determined in advance using the computer program GPOWER, with alpha set at .05 and power at .8, and assuming a medium effect size. Ages ranged between 16 and 46 (mean=18.71; S.D.=2.58), with no significant age difference between the sexes (t293=1.46, p=.15).

2.2. Materials and procedure

Participants were recruited online from the pool of undergraduate psychology students and received course credit for their participation. Some were tested alone, but most were tested in groups of up to 30 people (median=25). Participants reported to a classroom and sat in forward-facing desks, spaced apart from one another. The materials for the study consisted of a booklet of questionnaires titled “Understanding Behavior” (available on request from the author). After providing general biographical information, participants completed the following questionnaires:

3.1. Help given as a function of relationship category and cost of help

Raw scores were converted to t-scores (standardized scores based on z-scores, but centered on 50 and with a standard deviation of 10 units). These scores were used because they permit one to compare the share of low-, medium- and high-cost help given to members of each relationship category. Scores were analyzed using an ANOVA with three between-group factors (Relationship Category, Sex of Participant, and Sex of Target) and one within-group factor (Cost of Help Given). This revealed an interaction between Relationship Category and Cost of Help Given (F6, 548=14.95, p<.001, ηp2=.14). Note that neither Participant Sex nor Target Sex interacted with any of the other variables of interest, and therefore that sex is not considered in the remainder of this article.

3.2. Relatedness, friendship and altruism

To locate the source of the Relationship Category×Cost of Help interaction, further between- and within-group analyses were conducted. Looking first at the between-group analyses, there were main effects of Relationship Category for low-cost help (F3, 279=33.35, p<.001, ηp2=.26), medium-cost help (F3, 275=24.69, p<.001, ηp2=.21) and high-cost help (F3, 278=53.21, p<.001, ηp2=.37). As Fig. 1 shows, at each level of help, for siblings, cousins and acquaintances, greater relatedness was associated with higher rates of helping. Fig. 1 also shows that, at each level of help, friends were an exception to this general rule. However, as the cost of help rose, the relative ranking of friends fell. Post hoc tests using Dunnett's C revealed that, for low-cost help, friends received significantly more help than did members of any other relationship category; for medium-cost help, friends fell to the level of siblings; and for high-cost help, friends fell to the level of cousins, receiving significantly less help than siblings.

Fig. 1. Help given as a function of relationship category and cost of helping.

3.3. Kinship and reciprocal exchange

Hypothesis 2 stated that a higher level of reciprocity would be found among acquaintances than among friends, and among friends than either among siblings or cousins. To assess this hypothesis, two new aggregate variables were formed: low/medium-cost help given and low/medium-cost help received. These were formed from the one low-cost help and five medium-cost help variables. The rationale for forming these variables was that reciprocation might not always be in kind; an instance of medium-cost help might, for example, be reciprocated with several instances of low-cost help. To ensure that the estimates of levels of reciprocal exchange were not artificially inflated, 35 participants who indicated that they had neither given help to nor received help from the target individual were excluded from the analysis. Simple linear regressions were performed for each relationship category, using help given as a predictor of help received. All β values were significant, suggesting a consistent pattern of reciprocity within each relationship category (see Fig. 2).

Fig. 2. Help given as a function of help received for each relationship category.

Having established that the β values for each relationship category differed significantly from 0, the next question was whether they differed significantly from one another. To answer this, 95% confidence intervals were calculated for each value. As Fig. 2 shows, the confidence intervals for siblings, cousins and friends overlapped with one another, indicating that there were no significant differences among the β values for these categories. Only the confidence interval for acquaintances did not overlap with those of any other category. Consistent with Hypothesis 2, then, the degree of reciprocity among acquaintances was significantly higher than that among friends and that among siblings or cousins. Contrary to Hypothesis 2, however, the degree of reciprocity among friends was no higher than that among siblings or cousins.

4.1. Alternative explanations

There are several alternative explanations for the present study's results. One is based on the work of Clark & Mills, 1993, Mills & Clark, 1982, who have drawn a distinction between communal and exchange relationships. In communal relationships, benefits are given based on recipient need and without expectation of immediate repayment. In exchange relationships, on the other hand, benefits are given with the expectation of prompt repayment. The communal/exchange distinction cuts across the kin/nonkin divide: Communal relationships include those with close friends and (typically) close kin; exchange relationships include those with acquaintances, but can also include those with kin.

It might be suggested that the communal/exchange approach provides a better account of the data than does an evolutionary approach, and especially of the fact that friends resemble kin in giving high levels of help without requiring immediate reciprocation. However, this conclusion would not be warranted. First, in itself, the communal/exchange distinction does not explain similarities in the treatment of kin and close friends; it simply labels these similarities. In contrast, an evolutionary approach aims to provide an ultimate explanation for the patterns observed. Of course, one might argue that the capacity to form communal vs. exchange relationships has an evolutionary origin (e.g., Korchmaros & Kenny, 2006). But then the communal/exchange approach would face the same challenge that the inclusive fitness approach faces: explaining why friends are treated in a similar way to kin, despite being unrelated. Beyond this, inclusive fitness theory provides reason to doubt that close relationships with kin and nonkin really are psychologically identical. It leads, for example, to the hypothesis that the share of help given to kin vs. nonkin will depend on the cost of help — a hypothesis that found support in the present study but that would not have emerged from the communal/exchange approach.

Another potential challenge is that the cost-of-help findings can be explained equally well in terms of cultural norms: People give more high-cost help to kin than to friends because they experience stronger normative pressure to do so. An initial response to this suggestion is that cultural norms could in principle be invoked to explain any pattern of findings. This does not mean such explanations are necessarily false, but it does highlight the need to provide independent evidence for any proposed norm. Korchmaros and Kenny (2006) claim to have such evidence. They found that people felt more obligated to help kin than nonkin in life-or-death situations than in everyday helping situations. The implicit assumption is that feelings of obligation are shaped only by cultural norms. This assumption seems unwarranted, however. Indeed, it seems entirely possible that questions about obligation tap into evolved psychological mechanisms that underlie the distribution of altruism. Thus, although possible in principle, there is no particular reason to think that the cost-of-help finding is a product of cultural norms.

4.2. Limitations

There were a number of potential weaknesses associated with the measurement strategy employed in the present study. One concerns the item dealing with willingness to donate a kidney. It is possible that scores on this item were influenced by people's belief that they would have to be genetically matched to the recipient, which could raise their scores for kin but lower them for nonkin. It is important to note, though, that high-cost help was also assessed using a second item (help during an emergency), and that exactly the same pattern of results emerged for that item. A second weakness concerns the assessment of levels of reciprocity. Although a strong association between help given and help received is consistent with a high level of reciprocity, it is somewhat ambiguous evidence. After all, the fact that two individuals give one another similar levels of help does not necessarily mean that either individual's help is contingent on the other's. As such, it would be desirable to corroborate the results of this study using an alternative method that would yield a less ambiguous indicator. A longitudinal field study would be one way to demonstrate more convincingly the extent to which help given is contingent on help received across different relationship categories.

A final limitation relates to the comparison of actual helping with hypothetical helping. Actual helping is presumably a product of both willingness and opportunity to help, whereas responses to hypothetical scenarios reflect only willingness to help. In defense of the validity of this comparison, however, the same pattern of results emerged even when controlling for variables related to opportunity to help, such as residential proximity, frequency of contact, and duration of contact. This suggests that actual helping provides a reasonably clear window on willingness to help, and thus can be meaningfully compared with responses to the hypothetical scenarios.

4.3. Conclusion

In summary, the present study uncovered several differences in patterns of helping among kin and nonkin that make good sense in light of inclusive fitness theory. Although the findings are not logically inconsistent with alternative approaches (such as the communal/exchange approach or explanations based on cultural norms), such approaches offer no a priori reason to expect these results. As such, the results constitute good support for an inclusive fitness approach to explaining altruism among humans.